branching: merge stable into default

contrib/heptapod-ci.yml

@@ -1,188 +1,188 b''
 stages:
   - tests
   - phabricator
 
 image: registry.heptapod.net/mercurial/ci-images/mercurial-core:$HG_CI_IMAGE_TAG
 
 variables:
     PYTHON: python
     TEST_HGMODULEPOLICY: "allow"
     HG_CI_IMAGE_TAG: "latest"
     TEST_HGTESTS_ALLOW_NETIO: "0"
 
 .all_template: &all
   when: always
 
 .runtests_template: &runtests
     <<: *all
     stage: tests
     # The runner made a clone as root.
     # We make a new clone owned by user used to run the step.
     before_script:
       - hg clone . /tmp/mercurial-ci/ --noupdate --config phases.publish=no
       - hg -R /tmp/mercurial-ci/ update `hg log --rev '.' --template '{node}'`
       - cd /tmp/mercurial-ci/
       - ls -1 tests/test-check-*.* > /tmp/check-tests.txt
       - black --version
       - clang-format --version
     script:
         - echo "python used, $PYTHON"
         - echo "$RUNTEST_ARGS"
         - HGTESTS_ALLOW_NETIO="$TEST_HGTESTS_ALLOW_NETIO" HGMODULEPOLICY="$TEST_HGMODULEPOLICY" "$PYTHON" tests/run-tests.py --color=always $RUNTEST_ARGS
 
 checks-py2:
     <<: *runtests
     variables:
         RUNTEST_ARGS: "--time --test-list /tmp/check-tests.txt"
 
 checks-py3:
     <<: *runtests
     variables:
         RUNTEST_ARGS: "--time --test-list /tmp/check-tests.txt"
         PYTHON: python3
 
 rust-cargo-test-py2: &rust_cargo_test
     <<: *all
     stage: tests
     script:
         - echo "python used, $PYTHON"
         - make rust-tests
 
 rust-cargo-test-py3:
     stage: tests
     <<: *rust_cargo_test
     variables:
         PYTHON: python3
 
 phabricator-refresh:
     stage: phabricator
     variables:
       DEFAULT_COMMENT: ":white_check_mark: refresh by Heptapod after a successful CI run (:octopus: :green_heart:)"
       STABLE_COMMENT: ":white_check_mark: refresh by Heptapod after a successful CI run (:octopus: :green_heart:)\n⚠ This patch is intended for stable ⚠\n{image https://media.giphy.com/media/nYI8SmmChYXK0/source.gif}"
     script:
       - |
         if [ `hg branch` == "stable" ]; then
             ./contrib/phab-refresh-stack.sh --comment "$STABLE_COMMENT";
         else
             ./contrib/phab-refresh-stack.sh --comment "$DEFAULT_COMMENT";
         fi
 
 test-py2:
     <<: *runtests
     variables:
         RUNTEST_ARGS: " --no-rust --blacklist /tmp/check-tests.txt"
         TEST_HGMODULEPOLICY: "c"
         TEST_HGTESTS_ALLOW_NETIO: "1"
 
 test-py3:
     <<: *runtests
     variables:
         RUNTEST_ARGS: " --no-rust --blacklist /tmp/check-tests.txt"
         PYTHON: python3
         TEST_HGMODULEPOLICY: "c"
         TEST_HGTESTS_ALLOW_NETIO: "1"
 
 test-py2-pure:
     <<: *runtests
     variables:
         RUNTEST_ARGS: "--pure --blacklist /tmp/check-tests.txt"
         TEST_HGMODULEPOLICY: "py"
 
 test-py3-pure:
     <<: *runtests
     variables:
         RUNTEST_ARGS: "--pure --blacklist /tmp/check-tests.txt"
         PYTHON: python3
         TEST_HGMODULEPOLICY: "py"
 
 test-py2-rust:
     <<: *runtests
     variables:
         HGWITHRUSTEXT: cpython
         RUNTEST_ARGS: "--rust --blacklist /tmp/check-tests.txt"
         TEST_HGMODULEPOLICY: "rust+c"
 
 test-py3-rust:
     <<: *runtests
     variables:
         HGWITHRUSTEXT: cpython
         RUNTEST_ARGS: "--rust --blacklist /tmp/check-tests.txt"
         PYTHON: python3
         TEST_HGMODULEPOLICY: "rust+c"
 
 test-py3-rhg:
     <<: *runtests
     variables:
         HGWITHRUSTEXT: cpython
         RUNTEST_ARGS: "--rust --rhg --blacklist /tmp/check-tests.txt"
         PYTHON: python3
         TEST_HGMODULEPOLICY: "rust+c"
 
 test-py2-chg:
     <<: *runtests
     variables:
         RUNTEST_ARGS: "--blacklist /tmp/check-tests.txt --chg"
         TEST_HGMODULEPOLICY: "c"
 
 test-py3-chg:
     <<: *runtests
     variables:
         PYTHON: python3
         RUNTEST_ARGS: "--blacklist /tmp/check-tests.txt --chg"
         TEST_HGMODULEPOLICY: "c"
 
 check-pytype-py3:
     extends: .runtests_template
     before_script:
       - hg clone . /tmp/mercurial-ci/ --noupdate --config phases.publish=no
       - hg -R /tmp/mercurial-ci/ update `hg log --rev '.' --template '{node}'`
       - cd /tmp/mercurial-ci/
       - make local PYTHON=$PYTHON
       - $PYTHON -m pip install --user -U pytype==2021.04.15
     variables:
         RUNTEST_ARGS: " --allow-slow-tests tests/test-check-pytype.t"
-        HGTEST_TIMEOUT: "3600"
+        HGTEST_SLOWTIMEOUT: "3600"
         PYTHON: python3
         TEST_HGMODULEPOLICY: "c"
 
 # `sh.exe --login` sets a couple of extra environment variables that are defined
 # in the MinGW shell, but switches CWD to /home/$username.  The previous value
 # is stored in OLDPWD.  Of the added variables, MSYSTEM is crucial to running
 # run-tests.py- it is needed to make run-tests.py generate a `python3` script
 # that satisfies the various shebang lines and delegates to `py -3`.
 .window_runtests_template: &windows_runtests
     <<: *all
     when: manual  # we don't have any Windows runners anymore at the moment
     stage: tests
     before_script:
       - C:/MinGW/msys/1.0/bin/sh.exe --login -c 'cd "$OLDPWD" && ls -1 tests/test-check-*.* > C:/Temp/check-tests.txt'
       # TODO: find/install cvs, bzr, perforce, gpg, sqlite3
 
     script:
         - echo "Entering script section"
         - echo "python used, $Env:PYTHON"
         - Invoke-Expression "$Env:PYTHON -V"
         - Invoke-Expression "$Env:PYTHON -m black --version"
         - echo "$Env:RUNTEST_ARGS"
         - echo "$Env:TMP"
         - echo "$Env:TEMP"
 
         - C:/MinGW/msys/1.0/bin/sh.exe --login -c 'cd "$OLDPWD" && HGTESTS_ALLOW_NETIO="$TEST_HGTESTS_ALLOW_NETIO" HGMODULEPOLICY="$TEST_HGMODULEPOLICY" $PYTHON tests/run-tests.py --color=always $RUNTEST_ARGS'
 
 windows-py3:
     <<: *windows_runtests
     tags:
       - windows
     variables:
         TEST_HGMODULEPOLICY: "c"
         RUNTEST_ARGS: "--blacklist C:/Temp/check-tests.txt"
         PYTHON: py -3
 
 windows-py3-pyox:
     <<: *windows_runtests
     tags:
       - windows
     variables:
         TEST_HGMODULEPOLICY: "c"
         RUNTEST_ARGS: "--blacklist C:/Temp/check-tests.txt --pyoxidized"
         PYTHON: py -3

mercurial/merge.py

@@ -1,2430 +1,2434 b''
 # merge.py - directory-level update/merge handling for Mercurial
 #
 # Copyright 2006, 2007 Olivia Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import collections
 import errno
 import struct
 
 from .i18n import _
 from .node import nullrev
 from .thirdparty import attr
 from .utils import stringutil
 from .dirstateutils import timestamp
 from . import (
     copies,
     encoding,
     error,
     filemerge,
     match as matchmod,
     mergestate as mergestatemod,
     obsutil,
     pathutil,
     pycompat,
     scmutil,
     subrepoutil,
     util,
     worker,
 )
 
 _pack = struct.pack
 _unpack = struct.unpack
 
 
 def _getcheckunknownconfig(repo, section, name):
     config = repo.ui.config(section, name)
     valid = [b'abort', b'ignore', b'warn']
     if config not in valid:
         validstr = b', '.join([b"'" + v + b"'" for v in valid])
         raise error.ConfigError(
             _(b"%s.%s not valid ('%s' is none of %s)")
             % (section, name, config, validstr)
         )
     return config
 
 
 def _checkunknownfile(repo, wctx, mctx, f, f2=None):
     if wctx.isinmemory():
         # Nothing to do in IMM because nothing in the "working copy" can be an
         # unknown file.
         #
         # Note that we should bail out here, not in ``_checkunknownfiles()``,
         # because that function does other useful work.
         return False
 
     if f2 is None:
         f2 = f
     return (
         repo.wvfs.audit.check(f)
         and repo.wvfs.isfileorlink(f)
         and repo.dirstate.normalize(f) not in repo.dirstate
         and mctx[f2].cmp(wctx[f])
     )
 
 
 class _unknowndirschecker(object):
     """
     Look for any unknown files or directories that may have a path conflict
     with a file.  If any path prefix of the file exists as a file or link,
     then it conflicts.  If the file itself is a directory that contains any
     file that is not tracked, then it conflicts.
 
     Returns the shortest path at which a conflict occurs, or None if there is
     no conflict.
     """
 
     def __init__(self):
         # A set of paths known to be good.  This prevents repeated checking of
         # dirs.  It will be updated with any new dirs that are checked and found
         # to be safe.
         self._unknowndircache = set()
 
         # A set of paths that are known to be absent.  This prevents repeated
         # checking of subdirectories that are known not to exist. It will be
         # updated with any new dirs that are checked and found to be absent.
         self._missingdircache = set()
 
     def __call__(self, repo, wctx, f):
         if wctx.isinmemory():
             # Nothing to do in IMM for the same reason as ``_checkunknownfile``.
             return False
 
         # Check for path prefixes that exist as unknown files.
         for p in reversed(list(pathutil.finddirs(f))):
             if p in self._missingdircache:
                 return
             if p in self._unknowndircache:
                 continue
             if repo.wvfs.audit.check(p):
                 if (
                     repo.wvfs.isfileorlink(p)
                     and repo.dirstate.normalize(p) not in repo.dirstate
                 ):
                     return p
                 if not repo.wvfs.lexists(p):
                     self._missingdircache.add(p)
                     return
                 self._unknowndircache.add(p)
 
         # Check if the file conflicts with a directory containing unknown files.
         if repo.wvfs.audit.check(f) and repo.wvfs.isdir(f):
             # Does the directory contain any files that are not in the dirstate?
             for p, dirs, files in repo.wvfs.walk(f):
                 for fn in files:
                     relf = util.pconvert(repo.wvfs.reljoin(p, fn))
                     relf = repo.dirstate.normalize(relf, isknown=True)
                     if relf not in repo.dirstate:
                         return f
         return None
 
 
 def _checkunknownfiles(repo, wctx, mctx, force, mresult, mergeforce):
     """
     Considers any actions that care about the presence of conflicting unknown
     files. For some actions, the result is to abort; for others, it is to
     choose a different action.
     """
     fileconflicts = set()
     pathconflicts = set()
     warnconflicts = set()
     abortconflicts = set()
     unknownconfig = _getcheckunknownconfig(repo, b'merge', b'checkunknown')
     ignoredconfig = _getcheckunknownconfig(repo, b'merge', b'checkignored')
     pathconfig = repo.ui.configbool(
         b'experimental', b'merge.checkpathconflicts'
     )
     if not force:
 
         def collectconflicts(conflicts, config):
             if config == b'abort':
                 abortconflicts.update(conflicts)
             elif config == b'warn':
                 warnconflicts.update(conflicts)
 
         checkunknowndirs = _unknowndirschecker()
         for f in mresult.files(
             (
                 mergestatemod.ACTION_CREATED,
                 mergestatemod.ACTION_DELETED_CHANGED,
             )
         ):
             if _checkunknownfile(repo, wctx, mctx, f):
                 fileconflicts.add(f)
             elif pathconfig and f not in wctx:
                 path = checkunknowndirs(repo, wctx, f)
                 if path is not None:
                     pathconflicts.add(path)
         for f, args, msg in mresult.getactions(
             [mergestatemod.ACTION_LOCAL_DIR_RENAME_GET]
         ):
             if _checkunknownfile(repo, wctx, mctx, f, args[0]):
                 fileconflicts.add(f)
 
         allconflicts = fileconflicts | pathconflicts
         ignoredconflicts = {c for c in allconflicts if repo.dirstate._ignore(c)}
         unknownconflicts = allconflicts - ignoredconflicts
         collectconflicts(ignoredconflicts, ignoredconfig)
         collectconflicts(unknownconflicts, unknownconfig)
     else:
         for f, args, msg in list(
             mresult.getactions([mergestatemod.ACTION_CREATED_MERGE])
         ):
             fl2, anc = args
             different = _checkunknownfile(repo, wctx, mctx, f)
             if repo.dirstate._ignore(f):
                 config = ignoredconfig
             else:
                 config = unknownconfig
 
             # The behavior when force is True is described by this table:
             #  config  different  mergeforce  |    action    backup
             #    *         n          *       |      get        n
             #    *         y          y       |     merge       -
             #   abort      y          n       |     merge       -   (1)
             #   warn       y          n       |  warn + get     y
             #  ignore      y          n       |      get        y
             #
             # (1) this is probably the wrong behavior here -- we should
             #     probably abort, but some actions like rebases currently
             #     don't like an abort happening in the middle of
             #     merge.update.
             if not different:
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_GET,
                     (fl2, False),
                     b'remote created',
                 )
             elif mergeforce or config == b'abort':
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_MERGE,
                     (f, f, None, False, anc),
                     b'remote differs from untracked local',
                 )
             elif config == b'abort':
                 abortconflicts.add(f)
             else:
                 if config == b'warn':
                     warnconflicts.add(f)
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_GET,
                     (fl2, True),
                     b'remote created',
                 )
 
     for f in sorted(abortconflicts):
         warn = repo.ui.warn
         if f in pathconflicts:
             if repo.wvfs.isfileorlink(f):
                 warn(_(b"%s: untracked file conflicts with directory\n") % f)
             else:
                 warn(_(b"%s: untracked directory conflicts with file\n") % f)
         else:
             warn(_(b"%s: untracked file differs\n") % f)
     if abortconflicts:
         raise error.StateError(
             _(
                 b"untracked files in working directory "
                 b"differ from files in requested revision"
             )
         )
 
     for f in sorted(warnconflicts):
         if repo.wvfs.isfileorlink(f):
             repo.ui.warn(_(b"%s: replacing untracked file\n") % f)
         else:
             repo.ui.warn(_(b"%s: replacing untracked files in directory\n") % f)
 
     for f, args, msg in list(
         mresult.getactions([mergestatemod.ACTION_CREATED])
     ):
         backup = (
             f in fileconflicts
             or f in pathconflicts
             or any(p in pathconflicts for p in pathutil.finddirs(f))
         )
         (flags,) = args
         mresult.addfile(f, mergestatemod.ACTION_GET, (flags, backup), msg)
 
 
 def _forgetremoved(wctx, mctx, branchmerge, mresult):
     """
     Forget removed files
 
     If we're jumping between revisions (as opposed to merging), and if
     neither the working directory nor the target rev has the file,
     then we need to remove it from the dirstate, to prevent the
     dirstate from listing the file when it is no longer in the
     manifest.
 
     If we're merging, and the other revision has removed a file
     that is not present in the working directory, we need to mark it
     as removed.
     """
 
     m = mergestatemod.ACTION_FORGET
     if branchmerge:
         m = mergestatemod.ACTION_REMOVE
     for f in wctx.deleted():
         if f not in mctx:
             mresult.addfile(f, m, None, b"forget deleted")
 
     if not branchmerge:
         for f in wctx.removed():
             if f not in mctx:
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_FORGET,
                     None,
                     b"forget removed",
                 )
 
 
 def _checkcollision(repo, wmf, mresult):
     """
     Check for case-folding collisions.
     """
     # If the repo is narrowed, filter out files outside the narrowspec.
     narrowmatch = repo.narrowmatch()
     if not narrowmatch.always():
         pmmf = set(wmf.walk(narrowmatch))
         if mresult:
             for f in list(mresult.files()):
                 if not narrowmatch(f):
                     mresult.removefile(f)
     else:
         # build provisional merged manifest up
         pmmf = set(wmf)
 
     if mresult:
         # KEEP and EXEC are no-op
         for f in mresult.files(
             (
                 mergestatemod.ACTION_ADD,
                 mergestatemod.ACTION_ADD_MODIFIED,
                 mergestatemod.ACTION_FORGET,
                 mergestatemod.ACTION_GET,
                 mergestatemod.ACTION_CHANGED_DELETED,
                 mergestatemod.ACTION_DELETED_CHANGED,
             )
         ):
             pmmf.add(f)
         for f in mresult.files((mergestatemod.ACTION_REMOVE,)):
             pmmf.discard(f)
         for f, args, msg in mresult.getactions(
             [mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL]
         ):
             f2, flags = args
             pmmf.discard(f2)
             pmmf.add(f)
         for f in mresult.files((mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,)):
             pmmf.add(f)
         for f, args, msg in mresult.getactions([mergestatemod.ACTION_MERGE]):
             f1, f2, fa, move, anc = args
             if move:
                 pmmf.discard(f1)
             pmmf.add(f)
 
     # check case-folding collision in provisional merged manifest
     foldmap = {}
     for f in pmmf:
         fold = util.normcase(f)
         if fold in foldmap:
             raise error.StateError(
                 _(b"case-folding collision between %s and %s")
                 % (f, foldmap[fold])
             )
         foldmap[fold] = f
 
     # check case-folding of directories
     foldprefix = unfoldprefix = lastfull = b''
     for fold, f in sorted(foldmap.items()):
         if fold.startswith(foldprefix) and not f.startswith(unfoldprefix):
             # the folded prefix matches but actual casing is different
             raise error.StateError(
                 _(b"case-folding collision between %s and directory of %s")
                 % (lastfull, f)
             )
         foldprefix = fold + b'/'
         unfoldprefix = f + b'/'
         lastfull = f
 
 
 def _filesindirs(repo, manifest, dirs):
     """
     Generator that yields pairs of all the files in the manifest that are found
     inside the directories listed in dirs, and which directory they are found
     in.
     """
     for f in manifest:
         for p in pathutil.finddirs(f):
             if p in dirs:
                 yield f, p
                 break
 
 
 def checkpathconflicts(repo, wctx, mctx, mresult):
     """
     Check if any actions introduce path conflicts in the repository, updating
     actions to record or handle the path conflict accordingly.
     """
     mf = wctx.manifest()
 
     # The set of local files that conflict with a remote directory.
     localconflicts = set()
 
     # The set of directories that conflict with a remote file, and so may cause
     # conflicts if they still contain any files after the merge.
     remoteconflicts = set()
 
     # The set of directories that appear as both a file and a directory in the
     # remote manifest.  These indicate an invalid remote manifest, which
     # can't be updated to cleanly.
     invalidconflicts = set()
 
     # The set of directories that contain files that are being created.
     createdfiledirs = set()
 
     # The set of files deleted by all the actions.
     deletedfiles = set()
 
     for f in mresult.files(
         (
             mergestatemod.ACTION_CREATED,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_MERGE,
             mergestatemod.ACTION_CREATED_MERGE,
         )
     ):
         # This action may create a new local file.
         createdfiledirs.update(pathutil.finddirs(f))
         if mf.hasdir(f):
             # The file aliases a local directory.  This might be ok if all
             # the files in the local directory are being deleted.  This
             # will be checked once we know what all the deleted files are.
             remoteconflicts.add(f)
     # Track the names of all deleted files.
     for f in mresult.files((mergestatemod.ACTION_REMOVE,)):
         deletedfiles.add(f)
     for (f, args, msg) in mresult.getactions((mergestatemod.ACTION_MERGE,)):
         f1, f2, fa, move, anc = args
         if move:
             deletedfiles.add(f1)
     for (f, args, msg) in mresult.getactions(
         (mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL,)
     ):
         f2, flags = args
         deletedfiles.add(f2)
 
     # Check all directories that contain created files for path conflicts.
     for p in createdfiledirs:
         if p in mf:
             if p in mctx:
                 # A file is in a directory which aliases both a local
                 # and a remote file.  This is an internal inconsistency
                 # within the remote manifest.
                 invalidconflicts.add(p)
             else:
                 # A file is in a directory which aliases a local file.
                 # We will need to rename the local file.
                 localconflicts.add(p)
         pd = mresult.getfile(p)
         if pd and pd[0] in (
             mergestatemod.ACTION_CREATED,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_MERGE,
             mergestatemod.ACTION_CREATED_MERGE,
         ):
             # The file is in a directory which aliases a remote file.
             # This is an internal inconsistency within the remote
             # manifest.
             invalidconflicts.add(p)
 
     # Rename all local conflicting files that have not been deleted.
     for p in localconflicts:
         if p not in deletedfiles:
             ctxname = bytes(wctx).rstrip(b'+')
             pnew = util.safename(p, ctxname, wctx, set(mresult.files()))
             porig = wctx[p].copysource() or p
             mresult.addfile(
                 pnew,
                 mergestatemod.ACTION_PATH_CONFLICT_RESOLVE,
                 (p, porig),
                 b'local path conflict',
             )
             mresult.addfile(
                 p,
                 mergestatemod.ACTION_PATH_CONFLICT,
                 (pnew, b'l'),
                 b'path conflict',
             )
 
     if remoteconflicts:
         # Check if all files in the conflicting directories have been removed.
         ctxname = bytes(mctx).rstrip(b'+')
         for f, p in _filesindirs(repo, mf, remoteconflicts):
             if f not in deletedfiles:
                 m, args, msg = mresult.getfile(p)
                 pnew = util.safename(p, ctxname, wctx, set(mresult.files()))
                 if m in (
                     mergestatemod.ACTION_DELETED_CHANGED,
                     mergestatemod.ACTION_MERGE,
                 ):
                     # Action was merge, just update target.
                     mresult.addfile(pnew, m, args, msg)
                 else:
                     # Action was create, change to renamed get action.
                     fl = args[0]
                     mresult.addfile(
                         pnew,
                         mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
                         (p, fl),
                         b'remote path conflict',
                     )
                 mresult.addfile(
                     p,
                     mergestatemod.ACTION_PATH_CONFLICT,
                     (pnew, mergestatemod.ACTION_REMOVE),
                     b'path conflict',
                 )
                 remoteconflicts.remove(p)
                 break
 
     if invalidconflicts:
         for p in invalidconflicts:
             repo.ui.warn(_(b"%s: is both a file and a directory\n") % p)
         raise error.StateError(
             _(b"destination manifest contains path conflicts")
         )
 
 
 def _filternarrowactions(narrowmatch, branchmerge, mresult):
     """
     Filters out actions that can ignored because the repo is narrowed.
 
     Raise an exception if the merge cannot be completed because the repo is
     narrowed.
     """
     # TODO: handle with nonconflicttypes
     nonconflicttypes = {
         mergestatemod.ACTION_ADD,
         mergestatemod.ACTION_ADD_MODIFIED,
         mergestatemod.ACTION_CREATED,
         mergestatemod.ACTION_CREATED_MERGE,
         mergestatemod.ACTION_FORGET,
         mergestatemod.ACTION_GET,
         mergestatemod.ACTION_REMOVE,
         mergestatemod.ACTION_EXEC,
     }
     # We mutate the items in the dict during iteration, so iterate
     # over a copy.
     for f, action in mresult.filemap():
         if narrowmatch(f):
             pass
         elif not branchmerge:
             mresult.removefile(f)  # just updating, ignore changes outside clone
         elif action[0] in mergestatemod.NO_OP_ACTIONS:
             mresult.removefile(f)  # merge does not affect file
         elif action[0] in nonconflicttypes:
             raise error.Abort(
                 _(
                     b'merge affects file \'%s\' outside narrow, '
                     b'which is not yet supported'
                 )
                 % f,
                 hint=_(b'merging in the other direction may work'),
             )
         else:
             raise error.Abort(
                 _(b'conflict in file \'%s\' is outside narrow clone') % f
             )
 
 
 class mergeresult(object):
     """An object representing result of merging manifests.
 
     It has information about what actions need to be performed on dirstate
     mapping of divergent renames and other such cases."""
 
     def __init__(self):
         """
         filemapping: dict of filename as keys and action related info as values
         diverge: mapping of source name -> list of dest name for
                  divergent renames
         renamedelete: mapping of source name -> list of destinations for files
                       deleted on one side and renamed on other.
         commitinfo: dict containing data which should be used on commit
                     contains a filename -> info mapping
         actionmapping: dict of action names as keys and values are dict of
                        filename as key and related data as values
         """
         self._filemapping = {}
         self._diverge = {}
         self._renamedelete = {}
         self._commitinfo = collections.defaultdict(dict)
         self._actionmapping = collections.defaultdict(dict)
 
     def updatevalues(self, diverge, renamedelete):
         self._diverge = diverge
         self._renamedelete = renamedelete
 
     def addfile(self, filename, action, data, message):
         """adds a new file to the mergeresult object
 
         filename: file which we are adding
         action: one of mergestatemod.ACTION_*
         data: a tuple of information like fctx and ctx related to this merge
         message: a message about the merge
         """
         # if the file already existed, we need to delete it's old
         # entry form _actionmapping too
         if filename in self._filemapping:
             a, d, m = self._filemapping[filename]
             del self._actionmapping[a][filename]
 
         self._filemapping[filename] = (action, data, message)
         self._actionmapping[action][filename] = (data, message)
 
     def getfile(self, filename, default_return=None):
         """returns (action, args, msg) about this file
 
         returns default_return if the file is not present"""
         if filename in self._filemapping:
             return self._filemapping[filename]
         return default_return
 
     def files(self, actions=None):
         """returns files on which provided action needs to perfromed
 
         If actions is None, all files are returned
         """
         # TODO: think whether we should return renamedelete and
         # diverge filenames also
         if actions is None:
             for f in self._filemapping:
                 yield f
 
         else:
             for a in actions:
                 for f in self._actionmapping[a]:
                     yield f
 
     def removefile(self, filename):
         """removes a file from the mergeresult object as the file might
         not merging anymore"""
         action, data, message = self._filemapping[filename]
         del self._filemapping[filename]
         del self._actionmapping[action][filename]
 
     def getactions(self, actions, sort=False):
         """get list of files which are marked with these actions
         if sort is true, files for each action is sorted and then added
 
         Returns a list of tuple of form (filename, data, message)
         """
         for a in actions:
             if sort:
                 for f in sorted(self._actionmapping[a]):
                     args, msg = self._actionmapping[a][f]
                     yield f, args, msg
             else:
                 for f, (args, msg) in pycompat.iteritems(
                     self._actionmapping[a]
                 ):
                     yield f, args, msg
 
     def len(self, actions=None):
         """returns number of files which needs actions
 
         if actions is passed, total of number of files in that action
         only is returned"""
 
         if actions is None:
             return len(self._filemapping)
 
         return sum(len(self._actionmapping[a]) for a in actions)
 
     def filemap(self, sort=False):
         if sorted:
             for key, val in sorted(pycompat.iteritems(self._filemapping)):
                 yield key, val
         else:
             for key, val in pycompat.iteritems(self._filemapping):
                 yield key, val
 
     def addcommitinfo(self, filename, key, value):
         """adds key-value information about filename which will be required
         while committing this merge"""
         self._commitinfo[filename][key] = value
 
     @property
     def diverge(self):
         return self._diverge
 
     @property
     def renamedelete(self):
         return self._renamedelete
 
     @property
     def commitinfo(self):
         return self._commitinfo
 
     @property
     def actionsdict(self):
         """returns a dictionary of actions to be perfomed with action as key
         and a list of files and related arguments as values"""
         res = collections.defaultdict(list)
         for a, d in pycompat.iteritems(self._actionmapping):
             for f, (args, msg) in pycompat.iteritems(d):
                 res[a].append((f, args, msg))
         return res
 
     def setactions(self, actions):
         self._filemapping = actions
         self._actionmapping = collections.defaultdict(dict)
         for f, (act, data, msg) in pycompat.iteritems(self._filemapping):
             self._actionmapping[act][f] = data, msg
 
     def hasconflicts(self):
         """tells whether this merge resulted in some actions which can
         result in conflicts or not"""
         for a in self._actionmapping.keys():
             if (
                 a
                 not in (
                     mergestatemod.ACTION_GET,
                     mergestatemod.ACTION_EXEC,
                     mergestatemod.ACTION_REMOVE,
                     mergestatemod.ACTION_PATH_CONFLICT_RESOLVE,
                 )
                 and self._actionmapping[a]
                 and a not in mergestatemod.NO_OP_ACTIONS
             ):
                 return True
 
         return False
 
 
 def manifestmerge(
     repo,
     wctx,
     p2,
     pa,
     branchmerge,
     force,
     matcher,
     acceptremote,
     followcopies,
     forcefulldiff=False,
 ):
     """
     Merge wctx and p2 with ancestor pa and generate merge action list
 
     branchmerge and force are as passed in to update
     matcher = matcher to filter file lists
     acceptremote = accept the incoming changes without prompting
 
     Returns an object of mergeresult class
     """
     mresult = mergeresult()
     if matcher is not None and matcher.always():
         matcher = None
 
     # manifests fetched in order are going to be faster, so prime the caches
     [
         x.manifest()
         for x in sorted(wctx.parents() + [p2, pa], key=scmutil.intrev)
     ]
 
     branch_copies1 = copies.branch_copies()
     branch_copies2 = copies.branch_copies()
     diverge = {}
     # information from merge which is needed at commit time
     # for example choosing filelog of which parent to commit
     # TODO: use specific constants in future for this mapping
     if followcopies:
         branch_copies1, branch_copies2, diverge = copies.mergecopies(
             repo, wctx, p2, pa
         )
 
     boolbm = pycompat.bytestr(bool(branchmerge))
     boolf = pycompat.bytestr(bool(force))
     boolm = pycompat.bytestr(bool(matcher))
     repo.ui.note(_(b"resolving manifests\n"))
     repo.ui.debug(
         b" branchmerge: %s, force: %s, partial: %s\n" % (boolbm, boolf, boolm)
     )
     repo.ui.debug(b" ancestor: %s, local: %s, remote: %s\n" % (pa, wctx, p2))
 
     m1, m2, ma = wctx.manifest(), p2.manifest(), pa.manifest()
     copied1 = set(branch_copies1.copy.values())
     copied1.update(branch_copies1.movewithdir.values())
     copied2 = set(branch_copies2.copy.values())
     copied2.update(branch_copies2.movewithdir.values())
 
     if b'.hgsubstate' in m1 and wctx.rev() is None:
         # Check whether sub state is modified, and overwrite the manifest
         # to flag the change. If wctx is a committed revision, we shouldn't
         # care for the dirty state of the working directory.
         if any(wctx.sub(s).dirty() for s in wctx.substate):
             m1[b'.hgsubstate'] = repo.nodeconstants.modifiednodeid
 
     # Don't use m2-vs-ma optimization if:
     # - ma is the same as m1 or m2, which we're just going to diff again later
     # - The caller specifically asks for a full diff, which is useful during bid
     #   merge.
     # - we are tracking salvaged files specifically hence should process all
     #   files
     if (
         pa not in ([wctx, p2] + wctx.parents())
         and not forcefulldiff
         and not (
             repo.ui.configbool(b'experimental', b'merge-track-salvaged')
             or repo.filecopiesmode == b'changeset-sidedata'
         )
     ):
         # Identify which files are relevant to the merge, so we can limit the
         # total m1-vs-m2 diff to just those files. This has significant
         # performance benefits in large repositories.
         relevantfiles = set(ma.diff(m2).keys())
 
         # For copied and moved files, we need to add the source file too.
         for copykey, copyvalue in pycompat.iteritems(branch_copies1.copy):
             if copyvalue in relevantfiles:
                 relevantfiles.add(copykey)
         for movedirkey in branch_copies1.movewithdir:
             relevantfiles.add(movedirkey)
         filesmatcher = scmutil.matchfiles(repo, relevantfiles)
         matcher = matchmod.intersectmatchers(matcher, filesmatcher)
 
     diff = m1.diff(m2, match=matcher)
 
     for f, ((n1, fl1), (n2, fl2)) in pycompat.iteritems(diff):
         if n1 and n2:  # file exists on both local and remote side
             if f not in ma:
                 # TODO: what if they're renamed from different sources?
                 fa = branch_copies1.copy.get(
                     f, None
                 ) or branch_copies2.copy.get(f, None)
                 args, msg = None, None
                 if fa is not None:
                     args = (f, f, fa, False, pa.node())
                     msg = b'both renamed from %s' % fa
                 else:
                     args = (f, f, None, False, pa.node())
                     msg = b'both created'
                 mresult.addfile(f, mergestatemod.ACTION_MERGE, args, msg)
             elif f in branch_copies1.copy:
                 fa = branch_copies1.copy[f]
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_MERGE,
                     (f, fa, fa, False, pa.node()),
                     b'local replaced from %s' % fa,
                 )
             elif f in branch_copies2.copy:
                 fa = branch_copies2.copy[f]
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_MERGE,
                     (fa, f, fa, False, pa.node()),
                     b'other replaced from %s' % fa,
                 )
             else:
                 a = ma[f]
                 fla = ma.flags(f)
                 nol = b'l' not in fl1 + fl2 + fla
                 if n2 == a and fl2 == fla:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_KEEP,
                         (),
                         b'remote unchanged',
                     )
                 elif n1 == a and fl1 == fla:  # local unchanged - use remote
                     if n1 == n2:  # optimization: keep local content
                         mresult.addfile(
                             f,
                             mergestatemod.ACTION_EXEC,
                             (fl2,),
                             b'update permissions',
                         )
                     else:
                         mresult.addfile(
                             f,
                             mergestatemod.ACTION_GET,
                             (fl2, False),
                             b'remote is newer',
                         )
                         if branchmerge:
                             mresult.addcommitinfo(
                                 f, b'filenode-source', b'other'
                             )
                 elif nol and n2 == a:  # remote only changed 'x'
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_EXEC,
                         (fl2,),
                         b'update permissions',
                     )
                 elif nol and n1 == a:  # local only changed 'x'
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_GET,
                         (fl1, False),
                         b'remote is newer',
                     )
                     if branchmerge:
                         mresult.addcommitinfo(f, b'filenode-source', b'other')
                 else:  # both changed something
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_MERGE,
                         (f, f, f, False, pa.node()),
                         b'versions differ',
                     )
         elif n1:  # file exists only on local side
             if f in copied2:
                 pass  # we'll deal with it on m2 side
             elif (
                 f in branch_copies1.movewithdir
             ):  # directory rename, move local
                 f2 = branch_copies1.movewithdir[f]
                 if f2 in m2:
                     mresult.addfile(
                         f2,
                         mergestatemod.ACTION_MERGE,
                         (f, f2, None, True, pa.node()),
                         b'remote directory rename, both created',
                     )
                 else:
                     mresult.addfile(
                         f2,
                         mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL,
                         (f, fl1),
                         b'remote directory rename - move from %s' % f,
                     )
             elif f in branch_copies1.copy:
                 f2 = branch_copies1.copy[f]
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_MERGE,
                     (f, f2, f2, False, pa.node()),
                     b'local copied/moved from %s' % f2,
                 )
             elif f in ma:  # clean, a different, no remote
                 if n1 != ma[f]:
                     if acceptremote:
                         mresult.addfile(
                             f,
                             mergestatemod.ACTION_REMOVE,
                             None,
                             b'remote delete',
                         )
                     else:
                         mresult.addfile(
                             f,
                             mergestatemod.ACTION_CHANGED_DELETED,
                             (f, None, f, False, pa.node()),
                             b'prompt changed/deleted',
                         )
                         if branchmerge:
                             mresult.addcommitinfo(
                                 f, b'merge-removal-candidate', b'yes'
                             )
                 elif n1 == repo.nodeconstants.addednodeid:
                     # This file was locally added. We should forget it instead of
                     # deleting it.
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_FORGET,
                         None,
                         b'remote deleted',
                     )
                 else:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_REMOVE,
                         None,
                         b'other deleted',
                     )
                     if branchmerge:
                         # the file must be absent after merging,
                         # howeber the user might make
                         # the file reappear using revert and if they does,
                         # we force create a new node
                         mresult.addcommitinfo(
                             f, b'merge-removal-candidate', b'yes'
                         )
 
             else:  # file not in ancestor, not in remote
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_KEEP_NEW,
                     None,
                     b'ancestor missing, remote missing',
                 )
 
         elif n2:  # file exists only on remote side
             if f in copied1:
                 pass  # we'll deal with it on m1 side
             elif f in branch_copies2.movewithdir:
                 f2 = branch_copies2.movewithdir[f]
                 if f2 in m1:
                     mresult.addfile(
                         f2,
                         mergestatemod.ACTION_MERGE,
                         (f2, f, None, False, pa.node()),
                         b'local directory rename, both created',
                     )
                 else:
                     mresult.addfile(
                         f2,
                         mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
                         (f, fl2),
                         b'local directory rename - get from %s' % f,
                     )
             elif f in branch_copies2.copy:
                 f2 = branch_copies2.copy[f]
                 msg, args = None, None
                 if f2 in m2:
                     args = (f2, f, f2, False, pa.node())
                     msg = b'remote copied from %s' % f2
                 else:
                     args = (f2, f, f2, True, pa.node())
                     msg = b'remote moved from %s' % f2
                 mresult.addfile(f, mergestatemod.ACTION_MERGE, args, msg)
             elif f not in ma:
                 # local unknown, remote created: the logic is described by the
                 # following table:
                 #
                 # force  branchmerge  different  |  action
                 #   n         *           *      |   create
                 #   y         n           *      |   create
                 #   y         y           n      |   create
                 #   y         y           y      |   merge
                 #
                 # Checking whether the files are different is expensive, so we
                 # don't do that when we can avoid it.
                 if not force:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_CREATED,
                         (fl2,),
                         b'remote created',
                     )
                 elif not branchmerge:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_CREATED,
                         (fl2,),
                         b'remote created',
                     )
                 else:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_CREATED_MERGE,
                         (fl2, pa.node()),
                         b'remote created, get or merge',
                     )
             elif n2 != ma[f]:
                 df = None
                 for d in branch_copies1.dirmove:
                     if f.startswith(d):
                         # new file added in a directory that was moved
                         df = branch_copies1.dirmove[d] + f[len(d) :]
                         break
                 if df is not None and df in m1:
                     mresult.addfile(
                         df,
                         mergestatemod.ACTION_MERGE,
                         (df, f, f, False, pa.node()),
                         b'local directory rename - respect move '
                         b'from %s' % f,
                     )
                 elif acceptremote:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_CREATED,
                         (fl2,),
                         b'remote recreating',
                     )
                 else:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_DELETED_CHANGED,
                         (None, f, f, False, pa.node()),
                         b'prompt deleted/changed',
                     )
                     if branchmerge:
                         mresult.addcommitinfo(
                             f, b'merge-removal-candidate', b'yes'
                         )
             else:
                 mresult.addfile(
                     f,
                     mergestatemod.ACTION_KEEP_ABSENT,
                     None,
                     b'local not present, remote unchanged',
                 )
                 if branchmerge:
                     # the file must be absent after merging
                     # however the user might make
                     # the file reappear using revert and if they does,
                     # we force create a new node
                     mresult.addcommitinfo(f, b'merge-removal-candidate', b'yes')
 
     if repo.ui.configbool(b'experimental', b'merge.checkpathconflicts'):
         # If we are merging, look for path conflicts.
         checkpathconflicts(repo, wctx, p2, mresult)
 
     narrowmatch = repo.narrowmatch()
     if not narrowmatch.always():
         # Updates "actions" in place
         _filternarrowactions(narrowmatch, branchmerge, mresult)
 
     renamedelete = branch_copies1.renamedelete
     renamedelete.update(branch_copies2.renamedelete)
 
     mresult.updatevalues(diverge, renamedelete)
     return mresult
 
 
 def _resolvetrivial(repo, wctx, mctx, ancestor, mresult):
     """Resolves false conflicts where the nodeid changed but the content
     remained the same."""
     # We force a copy of actions.items() because we're going to mutate
     # actions as we resolve trivial conflicts.
     for f in list(mresult.files((mergestatemod.ACTION_CHANGED_DELETED,))):
         if f in ancestor and not wctx[f].cmp(ancestor[f]):
             # local did change but ended up with same content
             mresult.addfile(
                 f, mergestatemod.ACTION_REMOVE, None, b'prompt same'
             )
 
     for f in list(mresult.files((mergestatemod.ACTION_DELETED_CHANGED,))):
         if f in ancestor and not mctx[f].cmp(ancestor[f]):
             # remote did change but ended up with same content
             mresult.removefile(f)  # don't get = keep local deleted
 
 
 def calculateupdates(
     repo,
     wctx,
     mctx,
     ancestors,
     branchmerge,
     force,
     acceptremote,
     followcopies,
     matcher=None,
     mergeforce=False,
 ):
     """
     Calculate the actions needed to merge mctx into wctx using ancestors
 
     Uses manifestmerge() to merge manifest and get list of actions required to
     perform for merging two manifests. If there are multiple ancestors, uses bid
     merge if enabled.
 
     Also filters out actions which are unrequired if repository is sparse.
 
     Returns mergeresult object same as manifestmerge().
     """
     # Avoid cycle.
     from . import sparse
 
     mresult = None
     if len(ancestors) == 1:  # default
         mresult = manifestmerge(
             repo,
             wctx,
             mctx,
             ancestors[0],
             branchmerge,
             force,
             matcher,
             acceptremote,
             followcopies,
         )
         _checkunknownfiles(repo, wctx, mctx, force, mresult, mergeforce)
 
     else:  # only when merge.preferancestor=* - the default
         repo.ui.note(
             _(b"note: merging %s and %s using bids from ancestors %s\n")
             % (
                 wctx,
                 mctx,
                 _(b' and ').join(pycompat.bytestr(anc) for anc in ancestors),
             )
         )
 
         # mapping filename to bids (action method to list af actions)
         # {FILENAME1 : BID1, FILENAME2 : BID2}
         # BID is another dictionary which contains
         # mapping of following form:
         # {ACTION_X : [info, ..], ACTION_Y : [info, ..]}
         fbids = {}
         mresult = mergeresult()
         diverge, renamedelete = None, None
         for ancestor in ancestors:
             repo.ui.note(_(b'\ncalculating bids for ancestor %s\n') % ancestor)
             mresult1 = manifestmerge(
                 repo,
                 wctx,
                 mctx,
                 ancestor,
                 branchmerge,
                 force,
                 matcher,
                 acceptremote,
                 followcopies,
                 forcefulldiff=True,
             )
             _checkunknownfiles(repo, wctx, mctx, force, mresult1, mergeforce)
 
             # Track the shortest set of warning on the theory that bid
             # merge will correctly incorporate more information
             if diverge is None or len(mresult1.diverge) < len(diverge):
                 diverge = mresult1.diverge
             if renamedelete is None or len(renamedelete) < len(
                 mresult1.renamedelete
             ):
                 renamedelete = mresult1.renamedelete
 
             # blindly update final mergeresult commitinfo with what we get
             # from mergeresult object for each ancestor
             # TODO: some commitinfo depends on what bid merge choose and hence
             # we will need to make commitinfo also depend on bid merge logic
             mresult._commitinfo.update(mresult1._commitinfo)
 
             for f, a in mresult1.filemap(sort=True):
                 m, args, msg = a
                 repo.ui.debug(b' %s: %s -> %s\n' % (f, msg, m))
                 if f in fbids:
                     d = fbids[f]
                     if m in d:
                         d[m].append(a)
                     else:
                         d[m] = [a]
                 else:
                     fbids[f] = {m: [a]}
 
         # Call for bids
         # Pick the best bid for each file
         repo.ui.note(
             _(b'\nauction for merging merge bids (%d ancestors)\n')
             % len(ancestors)
         )
         for f, bids in sorted(fbids.items()):
             if repo.ui.debugflag:
                 repo.ui.debug(b" list of bids for %s:\n" % f)
                 for m, l in sorted(bids.items()):
                     for _f, args, msg in l:
                         repo.ui.debug(b'   %s -> %s\n' % (msg, m))
             # bids is a mapping from action method to list af actions
             # Consensus?
             if len(bids) == 1:  # all bids are the same kind of method
                 m, l = list(bids.items())[0]
                 if all(a == l[0] for a in l[1:]):  # len(bids) is > 1
                     repo.ui.note(_(b" %s: consensus for %s\n") % (f, m))
                     mresult.addfile(f, *l[0])
                     continue
             # If keep is an option, just do it.
             if mergestatemod.ACTION_KEEP in bids:
                 repo.ui.note(_(b" %s: picking 'keep' action\n") % f)
                 mresult.addfile(f, *bids[mergestatemod.ACTION_KEEP][0])
                 continue
             # If keep absent is an option, just do that
             if mergestatemod.ACTION_KEEP_ABSENT in bids:
                 repo.ui.note(_(b" %s: picking 'keep absent' action\n") % f)
                 mresult.addfile(f, *bids[mergestatemod.ACTION_KEEP_ABSENT][0])
                 continue
             # ACTION_KEEP_NEW and ACTION_CHANGED_DELETED are conflicting actions
             # as one say that file is new while other says that file was present
             # earlier too and has a change delete conflict
             # Let's fall back to conflicting ACTION_CHANGED_DELETED and let user
             # do the right thing
             if (
                 mergestatemod.ACTION_CHANGED_DELETED in bids
                 and mergestatemod.ACTION_KEEP_NEW in bids
             ):
                 repo.ui.note(_(b" %s: picking 'changed/deleted' action\n") % f)
                 mresult.addfile(
                     f, *bids[mergestatemod.ACTION_CHANGED_DELETED][0]
                 )
                 continue
             # If keep new is an option, let's just do that
             if mergestatemod.ACTION_KEEP_NEW in bids:
                 repo.ui.note(_(b" %s: picking 'keep new' action\n") % f)
                 mresult.addfile(f, *bids[mergestatemod.ACTION_KEEP_NEW][0])
                 continue
             # ACTION_GET and ACTION_DELETE_CHANGED are conflicting actions as
             # one action states the file is newer/created on remote side and
             # other states that file is deleted locally and changed on remote
             # side. Let's fallback and rely on a conflicting action to let user
             # do the right thing
             if (
                 mergestatemod.ACTION_DELETED_CHANGED in bids
                 and mergestatemod.ACTION_GET in bids
             ):
                 repo.ui.note(_(b" %s: picking 'delete/changed' action\n") % f)
                 mresult.addfile(
                     f, *bids[mergestatemod.ACTION_DELETED_CHANGED][0]
                 )
                 continue
             # If there are gets and they all agree [how could they not?], do it.
             if mergestatemod.ACTION_GET in bids:
                 ga0 = bids[mergestatemod.ACTION_GET][0]
                 if all(a == ga0 for a in bids[mergestatemod.ACTION_GET][1:]):
                     repo.ui.note(_(b" %s: picking 'get' action\n") % f)
                     mresult.addfile(f, *ga0)
                     continue
             # TODO: Consider other simple actions such as mode changes
             # Handle inefficient democrazy.
             repo.ui.note(_(b' %s: multiple bids for merge action:\n') % f)
             for m, l in sorted(bids.items()):
                 for _f, args, msg in l:
                     repo.ui.note(b'  %s -> %s\n' % (msg, m))
             # Pick random action. TODO: Instead, prompt user when resolving
             m, l = list(bids.items())[0]
             repo.ui.warn(
                 _(b' %s: ambiguous merge - picked %s action\n') % (f, m)
             )
             mresult.addfile(f, *l[0])
             continue
         repo.ui.note(_(b'end of auction\n\n'))
         mresult.updatevalues(diverge, renamedelete)
 
     if wctx.rev() is None:
         _forgetremoved(wctx, mctx, branchmerge, mresult)
 
     sparse.filterupdatesactions(repo, wctx, mctx, branchmerge, mresult)
     _resolvetrivial(repo, wctx, mctx, ancestors[0], mresult)
 
     return mresult
 
 
 def _getcwd():
     try:
         return encoding.getcwd()
     except OSError as err:
         if err.errno == errno.ENOENT:
             return None
         raise
 
 
 def batchremove(repo, wctx, actions):
     """apply removes to the working directory
 
     yields tuples for progress updates
     """
     verbose = repo.ui.verbose
     cwd = _getcwd()
     i = 0
     for f, args, msg in actions:
         repo.ui.debug(b" %s: %s -> r\n" % (f, msg))
         if verbose:
             repo.ui.note(_(b"removing %s\n") % f)
         wctx[f].audit()
         try:
             wctx[f].remove(ignoremissing=True)
         except OSError as inst:
             repo.ui.warn(
                 _(b"update failed to remove %s: %s!\n")
                 % (f, stringutil.forcebytestr(inst.strerror))
             )
         if i == 100:
             yield i, f
             i = 0
         i += 1
     if i > 0:
         yield i, f
 
     if cwd and not _getcwd():
         # cwd was removed in the course of removing files; print a helpful
         # warning.
         repo.ui.warn(
             _(
                 b"current directory was removed\n"
                 b"(consider changing to repo root: %s)\n"
             )
             % repo.root
         )
 
 
 def batchget(repo, mctx, wctx, wantfiledata, actions):
     """apply gets to the working directory
 
     mctx is the context to get from
 
     Yields arbitrarily many (False, tuple) for progress updates, followed by
     exactly one (True, filedata). When wantfiledata is false, filedata is an
     empty dict. When wantfiledata is true, filedata[f] is a triple (mode, size,
     mtime) of the file f written for each action.
     """
     filedata = {}
     verbose = repo.ui.verbose
     fctx = mctx.filectx
     ui = repo.ui
     i = 0
     with repo.wvfs.backgroundclosing(ui, expectedcount=len(actions)):
         for f, (flags, backup), msg in actions:
             repo.ui.debug(b" %s: %s -> g\n" % (f, msg))
             if verbose:
                 repo.ui.note(_(b"getting %s\n") % f)
 
             if backup:
                 # If a file or directory exists with the same name, back that
                 # up.  Otherwise, look to see if there is a file that conflicts
                 # with a directory this file is in, and if so, back that up.
                 conflicting = f
                 if not repo.wvfs.lexists(f):
                     for p in pathutil.finddirs(f):
                         if repo.wvfs.isfileorlink(p):
                             conflicting = p
                             break
                 if repo.wvfs.lexists(conflicting):
                     orig = scmutil.backuppath(ui, repo, conflicting)
                     util.rename(repo.wjoin(conflicting), orig)
             wfctx = wctx[f]
             wfctx.clearunknown()
             atomictemp = ui.configbool(b"experimental", b"update.atomic-file")
             size = wfctx.write(
                 fctx(f).data(),
                 flags,
                 backgroundclose=True,
                 atomictemp=atomictemp,
             )
             if wantfiledata:
                 s = wfctx.lstat()
                 mode = s.st_mode
                 mtime = timestamp.mtime_of(s)
                 # for dirstate.update_file's parentfiledata argument:
                 filedata[f] = (mode, size, mtime)
             if i == 100:
                 yield False, (i, f)
                 i = 0
             i += 1
     if i > 0:
         yield False, (i, f)
     yield True, filedata
 
 
 def _prefetchfiles(repo, ctx, mresult):
     """Invoke ``scmutil.prefetchfiles()`` for the files relevant to the dict
     of merge actions.  ``ctx`` is the context being merged in."""
 
     # Skipping 'a', 'am', 'f', 'r', 'dm', 'e', 'k', 'p' and 'pr', because they
     # don't touch the context to be merged in.  'cd' is skipped, because
     # changed/deleted never resolves to something from the remote side.
     files = mresult.files(
         [
             mergestatemod.ACTION_GET,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
             mergestatemod.ACTION_MERGE,
         ]
     )
 
     prefetch = scmutil.prefetchfiles
     matchfiles = scmutil.matchfiles
     prefetch(
         repo,
         [
             (
                 ctx.rev(),
                 matchfiles(repo, files),
             )
         ],
     )
 
 
 @attr.s(frozen=True)
 class updateresult(object):
     updatedcount = attr.ib()
     mergedcount = attr.ib()
     removedcount = attr.ib()
     unresolvedcount = attr.ib()
 
     def isempty(self):
         return not (
             self.updatedcount
             or self.mergedcount
             or self.removedcount
             or self.unresolvedcount
         )
 
 
 def applyupdates(
     repo,
     mresult,
     wctx,
     mctx,
     overwrite,
     wantfiledata,
     labels=None,
 ):
     """apply the merge action list to the working directory
 
     mresult is a mergeresult object representing result of the merge
     wctx is the working copy context
     mctx is the context to be merged into the working copy
 
     Return a tuple of (counts, filedata), where counts is a tuple
     (updated, merged, removed, unresolved) that describes how many
     files were affected by the update, and filedata is as described in
     batchget.
     """
 
     _prefetchfiles(repo, mctx, mresult)
 
     updated, merged, removed = 0, 0, 0
     ms = wctx.mergestate(clean=True)
     ms.start(wctx.p1().node(), mctx.node(), labels)
 
     for f, op in pycompat.iteritems(mresult.commitinfo):
         # the other side of filenode was choosen while merging, store this in
         # mergestate so that it can be reused on commit
         ms.addcommitinfo(f, op)
 
     numupdates = mresult.len() - mresult.len(mergestatemod.NO_OP_ACTIONS)
     progress = repo.ui.makeprogress(
         _(b'updating'), unit=_(b'files'), total=numupdates
     )
 
     if b'.hgsubstate' in mresult._actionmapping[mergestatemod.ACTION_REMOVE]:
         subrepoutil.submerge(repo, wctx, mctx, wctx, overwrite, labels)
 
     # record path conflicts
     for f, args, msg in mresult.getactions(
         [mergestatemod.ACTION_PATH_CONFLICT], sort=True
     ):
         f1, fo = args
         s = repo.ui.status
         s(
             _(
                 b"%s: path conflict - a file or link has the same name as a "
                 b"directory\n"
             )
             % f
         )
         if fo == b'l':
             s(_(b"the local file has been renamed to %s\n") % f1)
         else:
             s(_(b"the remote file has been renamed to %s\n") % f1)
         s(_(b"resolve manually then use 'hg resolve --mark %s'\n") % f)
         ms.addpathconflict(f, f1, fo)
         progress.increment(item=f)
 
     # When merging in-memory, we can't support worker processes, so set the
     # per-item cost at 0 in that case.
     cost = 0 if wctx.isinmemory() else 0.001
 
     # remove in parallel (must come before resolving path conflicts and getting)
     prog = worker.worker(
         repo.ui,
         cost,
         batchremove,
         (repo, wctx),
         list(mresult.getactions([mergestatemod.ACTION_REMOVE], sort=True)),
     )
     for i, item in prog:
         progress.increment(step=i, item=item)
     removed = mresult.len((mergestatemod.ACTION_REMOVE,))
 
     # resolve path conflicts (must come before getting)
     for f, args, msg in mresult.getactions(
         [mergestatemod.ACTION_PATH_CONFLICT_RESOLVE], sort=True
     ):
         repo.ui.debug(b" %s: %s -> pr\n" % (f, msg))
         (f0, origf0) = args
         if wctx[f0].lexists():
             repo.ui.note(_(b"moving %s to %s\n") % (f0, f))
             wctx[f].audit()
             wctx[f].write(wctx.filectx(f0).data(), wctx.filectx(f0).flags())
             wctx[f0].remove()
         progress.increment(item=f)
 
     # get in parallel.
     threadsafe = repo.ui.configbool(
         b'experimental', b'worker.wdir-get-thread-safe'
     )
     prog = worker.worker(
         repo.ui,
         cost,
         batchget,
         (repo, mctx, wctx, wantfiledata),
         list(mresult.getactions([mergestatemod.ACTION_GET], sort=True)),
         threadsafe=threadsafe,
         hasretval=True,
     )
     getfiledata = {}
     for final, res in prog:
         if final:
             getfiledata = res
         else:
             i, item = res
             progress.increment(step=i, item=item)
 
     if b'.hgsubstate' in mresult._actionmapping[mergestatemod.ACTION_GET]:
         subrepoutil.submerge(repo, wctx, mctx, wctx, overwrite, labels)
 
     # forget (manifest only, just log it) (must come first)
     for f, args, msg in mresult.getactions(
         (mergestatemod.ACTION_FORGET,), sort=True
     ):
         repo.ui.debug(b" %s: %s -> f\n" % (f, msg))
         progress.increment(item=f)
 
     # re-add (manifest only, just log it)
     for f, args, msg in mresult.getactions(
         (mergestatemod.ACTION_ADD,), sort=True
     ):
         repo.ui.debug(b" %s: %s -> a\n" % (f, msg))
         progress.increment(item=f)
 
     # re-add/mark as modified (manifest only, just log it)
     for f, args, msg in mresult.getactions(
         (mergestatemod.ACTION_ADD_MODIFIED,), sort=True
     ):
         repo.ui.debug(b" %s: %s -> am\n" % (f, msg))
         progress.increment(item=f)
 
     # keep (noop, just log it)
     for a in mergestatemod.NO_OP_ACTIONS:
         for f, args, msg in mresult.getactions((a,), sort=True):
             repo.ui.debug(b" %s: %s -> %s\n" % (f, msg, a))
             # no progress
 
     # directory rename, move local
     for f, args, msg in mresult.getactions(
         (mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL,), sort=True
     ):
         repo.ui.debug(b" %s: %s -> dm\n" % (f, msg))
         progress.increment(item=f)
         f0, flags = args
         repo.ui.note(_(b"moving %s to %s\n") % (f0, f))
         wctx[f].audit()
         wctx[f].write(wctx.filectx(f0).data(), flags)
         wctx[f0].remove()
 
     # local directory rename, get
     for f, args, msg in mresult.getactions(
         (mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,), sort=True
     ):
         repo.ui.debug(b" %s: %s -> dg\n" % (f, msg))
         progress.increment(item=f)
         f0, flags = args
         repo.ui.note(_(b"getting %s to %s\n") % (f0, f))
         wctx[f].write(mctx.filectx(f0).data(), flags)
 
     # exec
     for f, args, msg in mresult.getactions(
         (mergestatemod.ACTION_EXEC,), sort=True
     ):
         repo.ui.debug(b" %s: %s -> e\n" % (f, msg))
         progress.increment(item=f)
         (flags,) = args
         wctx[f].audit()
         wctx[f].setflags(b'l' in flags, b'x' in flags)
 
     moves = []
 
     # 'cd' and 'dc' actions are treated like other merge conflicts
     mergeactions = list(
         mresult.getactions(
             [
                 mergestatemod.ACTION_CHANGED_DELETED,
                 mergestatemod.ACTION_DELETED_CHANGED,
                 mergestatemod.ACTION_MERGE,
             ],
             sort=True,
         )
     )
     for f, args, msg in mergeactions:
         f1, f2, fa, move, anc = args
         if f == b'.hgsubstate':  # merged internally
             continue
         if f1 is None:
             fcl = filemerge.absentfilectx(wctx, fa)
         else:
             repo.ui.debug(b" preserving %s for resolve of %s\n" % (f1, f))
             fcl = wctx[f1]
         if f2 is None:
             fco = filemerge.absentfilectx(mctx, fa)
         else:
             fco = mctx[f2]
         actx = repo[anc]
         if fa in actx:
             fca = actx[fa]
         else:
             # TODO: move to absentfilectx
             fca = repo.filectx(f1, fileid=nullrev)
         ms.add(fcl, fco, fca, f)
         if f1 != f and move:
             moves.append(f1)
 
     # remove renamed files after safely stored
     for f in moves:
         if wctx[f].lexists():
             repo.ui.debug(b"removing %s\n" % f)
             wctx[f].audit()
             wctx[f].remove()
 
     # these actions updates the file
     updated = mresult.len(
         (
             mergestatemod.ACTION_GET,
             mergestatemod.ACTION_EXEC,
             mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
             mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL,
         )
     )
 
     try:
         # premerge
         tocomplete = []
         for f, args, msg in mergeactions:
             repo.ui.debug(b" %s: %s -> m (premerge)\n" % (f, msg))
             ms.addcommitinfo(f, {b'merged': b'yes'})
             progress.increment(item=f)
             if f == b'.hgsubstate':  # subrepo states need updating
                 subrepoutil.submerge(
                     repo, wctx, mctx, wctx.ancestor(mctx), overwrite, labels
                 )
                 continue
             wctx[f].audit()
             complete, r = ms.preresolve(f, wctx)
             if not complete:
                 numupdates += 1
                 tocomplete.append((f, args, msg))
 
         # merge
         for f, args, msg in tocomplete:
             repo.ui.debug(b" %s: %s -> m (merge)\n" % (f, msg))
             ms.addcommitinfo(f, {b'merged': b'yes'})
             progress.increment(item=f, total=numupdates)
             ms.resolve(f, wctx)
 
+    except error.InterventionRequired:
+        # If the user has merge.on-failure=halt, catch the error and close the
+        # merge state "properly".
+        pass
     finally:
         ms.commit()
 
     unresolved = ms.unresolvedcount()
 
     msupdated, msmerged, msremoved = ms.counts()
     updated += msupdated
     merged += msmerged
     removed += msremoved
 
     extraactions = ms.actions()
 
     progress.complete()
     return (
         updateresult(updated, merged, removed, unresolved),
         getfiledata,
         extraactions,
     )
 
 
 def _advertisefsmonitor(repo, num_gets, p1node):
     # Advertise fsmonitor when its presence could be useful.
     #
     # We only advertise when performing an update from an empty working
     # directory. This typically only occurs during initial clone.
     #
     # We give users a mechanism to disable the warning in case it is
     # annoying.
     #
     # We only allow on Linux and MacOS because that's where fsmonitor is
     # considered stable.
     fsmonitorwarning = repo.ui.configbool(b'fsmonitor', b'warn_when_unused')
     fsmonitorthreshold = repo.ui.configint(
         b'fsmonitor', b'warn_update_file_count'
     )
     # avoid cycle dirstate -> sparse -> merge -> dirstate
     from . import dirstate
 
     if dirstate.rustmod is not None:
         # When using rust status, fsmonitor becomes necessary at higher sizes
         fsmonitorthreshold = repo.ui.configint(
             b'fsmonitor',
             b'warn_update_file_count_rust',
         )
 
     try:
         # avoid cycle: extensions -> cmdutil -> merge
         from . import extensions
 
         extensions.find(b'fsmonitor')
         fsmonitorenabled = repo.ui.config(b'fsmonitor', b'mode') != b'off'
         # We intentionally don't look at whether fsmonitor has disabled
         # itself because a) fsmonitor may have already printed a warning
         # b) we only care about the config state here.
     except KeyError:
         fsmonitorenabled = False
 
     if (
         fsmonitorwarning
         and not fsmonitorenabled
         and p1node == repo.nullid
         and num_gets >= fsmonitorthreshold
         and pycompat.sysplatform.startswith((b'linux', b'darwin'))
     ):
         repo.ui.warn(
             _(
                 b'(warning: large working directory being used without '
                 b'fsmonitor enabled; enable fsmonitor to improve performance; '
                 b'see "hg help -e fsmonitor")\n'
             )
         )
 
 
 UPDATECHECK_ABORT = b'abort'  # handled at higher layers
 UPDATECHECK_NONE = b'none'
 UPDATECHECK_LINEAR = b'linear'
 UPDATECHECK_NO_CONFLICT = b'noconflict'
 
 
 def _update(
     repo,
     node,
     branchmerge,
     force,
     ancestor=None,
     mergeancestor=False,
     labels=None,
     matcher=None,
     mergeforce=False,
     updatedirstate=True,
     updatecheck=None,
     wc=None,
 ):
     """
     Perform a merge between the working directory and the given node
 
     node = the node to update to
     branchmerge = whether to merge between branches
     force = whether to force branch merging or file overwriting
     matcher = a matcher to filter file lists (dirstate not updated)
     mergeancestor = whether it is merging with an ancestor. If true,
       we should accept the incoming changes for any prompts that occur.
       If false, merging with an ancestor (fast-forward) is only allowed
       between different named branches. This flag is used by rebase extension
       as a temporary fix and should be avoided in general.
     labels = labels to use for base, local and other
     mergeforce = whether the merge was run with 'merge --force' (deprecated): if
       this is True, then 'force' should be True as well.
 
     The table below shows all the behaviors of the update command given the
     -c/--check and -C/--clean or no options, whether the working directory is
     dirty, whether a revision is specified, and the relationship of the parent
     rev to the target rev (linear or not). Match from top first. The -n
     option doesn't exist on the command line, but represents the
     experimental.updatecheck=noconflict option.
 
     This logic is tested by test-update-branches.t.
 
     -c  -C  -n  -m  dirty  rev  linear  |  result
      y   y   *   *    *     *     *     |    (1)
      y   *   y   *    *     *     *     |    (1)
      y   *   *   y    *     *     *     |    (1)
      *   y   y   *    *     *     *     |    (1)
      *   y   *   y    *     *     *     |    (1)
      *   *   y   y    *     *     *     |    (1)
      *   *   *   *    *     n     n     |     x
      *   *   *   *    n     *     *     |    ok
      n   n   n   n    y     *     y     |   merge
      n   n   n   n    y     y     n     |    (2)
      n   n   n   y    y     *     *     |   merge
      n   n   y   n    y     *     *     |  merge if no conflict
      n   y   n   n    y     *     *     |  discard
      y   n   n   n    y     *     *     |    (3)
 
     x = can't happen
     * = don't-care
     1 = incompatible options (checked in commands.py)
     2 = abort: uncommitted changes (commit or update --clean to discard changes)
     3 = abort: uncommitted changes (checked in commands.py)
 
     The merge is performed inside ``wc``, a workingctx-like objects. It defaults
     to repo[None] if None is passed.
 
     Return the same tuple as applyupdates().
     """
     # Avoid cycle.
     from . import sparse
 
     # This function used to find the default destination if node was None, but
     # that's now in destutil.py.
     assert node is not None
     if not branchmerge and not force:
         # TODO: remove the default once all callers that pass branchmerge=False
         # and force=False pass a value for updatecheck. We may want to allow
         # updatecheck='abort' to better suppport some of these callers.
         if updatecheck is None:
             updatecheck = UPDATECHECK_LINEAR
         if updatecheck not in (
             UPDATECHECK_NONE,
             UPDATECHECK_LINEAR,
             UPDATECHECK_NO_CONFLICT,
         ):
             raise ValueError(
                 r'Invalid updatecheck %r (can accept %r)'
                 % (
                     updatecheck,
                     (
                         UPDATECHECK_NONE,
                         UPDATECHECK_LINEAR,
                         UPDATECHECK_NO_CONFLICT,
                     ),
                 )
             )
     if wc is not None and wc.isinmemory():
         maybe_wlock = util.nullcontextmanager()
     else:
         maybe_wlock = repo.wlock()
     with maybe_wlock:
         if wc is None:
             wc = repo[None]
         pl = wc.parents()
         p1 = pl[0]
         p2 = repo[node]
         if ancestor is not None:
             pas = [repo[ancestor]]
         else:
             if repo.ui.configlist(b'merge', b'preferancestor') == [b'*']:
                 cahs = repo.changelog.commonancestorsheads(p1.node(), p2.node())
                 pas = [repo[anc] for anc in (sorted(cahs) or [repo.nullid])]
             else:
                 pas = [p1.ancestor(p2, warn=branchmerge)]
 
         fp1, fp2, xp1, xp2 = p1.node(), p2.node(), bytes(p1), bytes(p2)
 
         overwrite = force and not branchmerge
         ### check phase
         if not overwrite:
             if len(pl) > 1:
                 raise error.StateError(_(b"outstanding uncommitted merge"))
             ms = wc.mergestate()
             if ms.unresolvedcount():
                 raise error.StateError(
                     _(b"outstanding merge conflicts"),
                     hint=_(b"use 'hg resolve' to resolve"),
                 )
         if branchmerge:
             if pas == [p2]:
                 raise error.Abort(
                     _(
                         b"merging with a working directory ancestor"
                         b" has no effect"
                     )
                 )
             elif pas == [p1]:
                 if not mergeancestor and wc.branch() == p2.branch():
                     raise error.Abort(
                         _(b"nothing to merge"),
                         hint=_(b"use 'hg update' or check 'hg heads'"),
                     )
             if not force and (wc.files() or wc.deleted()):
                 raise error.StateError(
                     _(b"uncommitted changes"),
                     hint=_(b"use 'hg status' to list changes"),
                 )
             if not wc.isinmemory():
                 for s in sorted(wc.substate):
                     wc.sub(s).bailifchanged()
 
         elif not overwrite:
             if p1 == p2:  # no-op update
                 # call the hooks and exit early
                 repo.hook(b'preupdate', throw=True, parent1=xp2, parent2=b'')
                 repo.hook(b'update', parent1=xp2, parent2=b'', error=0)
                 return updateresult(0, 0, 0, 0)
 
             if updatecheck == UPDATECHECK_LINEAR and pas not in (
                 [p1],
                 [p2],
             ):  # nonlinear
                 dirty = wc.dirty(missing=True)
                 if dirty:
                     # Branching is a bit strange to ensure we do the minimal
                     # amount of call to obsutil.foreground.
                     foreground = obsutil.foreground(repo, [p1.node()])
                     # note: the <node> variable contains a random identifier
                     if repo[node].node() in foreground:
                         pass  # allow updating to successors
                     else:
                         msg = _(b"uncommitted changes")
                         hint = _(b"commit or update --clean to discard changes")
                         raise error.UpdateAbort(msg, hint=hint)
                 else:
                     # Allow jumping branches if clean and specific rev given
                     pass
 
         if overwrite:
             pas = [wc]
         elif not branchmerge:
             pas = [p1]
 
         # deprecated config: merge.followcopies
         followcopies = repo.ui.configbool(b'merge', b'followcopies')
         if overwrite:
             followcopies = False
         elif not pas[0]:
             followcopies = False
         if not branchmerge and not wc.dirty(missing=True):
             followcopies = False
 
         ### calculate phase
         mresult = calculateupdates(
             repo,
             wc,
             p2,
             pas,
             branchmerge,
             force,
             mergeancestor,
             followcopies,
             matcher=matcher,
             mergeforce=mergeforce,
         )
 
         if updatecheck == UPDATECHECK_NO_CONFLICT:
             if mresult.hasconflicts():
                 msg = _(b"conflicting changes")
                 hint = _(b"commit or update --clean to discard changes")
                 raise error.StateError(msg, hint=hint)
 
         # Prompt and create actions. Most of this is in the resolve phase
         # already, but we can't handle .hgsubstate in filemerge or
         # subrepoutil.submerge yet so we have to keep prompting for it.
         vals = mresult.getfile(b'.hgsubstate')
         if vals:
             f = b'.hgsubstate'
             m, args, msg = vals
             prompts = filemerge.partextras(labels)
             prompts[b'f'] = f
             if m == mergestatemod.ACTION_CHANGED_DELETED:
                 if repo.ui.promptchoice(
                     _(
                         b"local%(l)s changed %(f)s which other%(o)s deleted\n"
                         b"use (c)hanged version or (d)elete?"
                         b"$$ &Changed $$ &Delete"
                     )
                     % prompts,
                     0,
                 ):
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_REMOVE,
                         None,
                         b'prompt delete',
                     )
                 elif f in p1:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_ADD_MODIFIED,
                         None,
                         b'prompt keep',
                     )
                 else:
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_ADD,
                         None,
                         b'prompt keep',
                     )
             elif m == mergestatemod.ACTION_DELETED_CHANGED:
                 f1, f2, fa, move, anc = args
                 flags = p2[f2].flags()
                 if (
                     repo.ui.promptchoice(
                         _(
                             b"other%(o)s changed %(f)s which local%(l)s deleted\n"
                             b"use (c)hanged version or leave (d)eleted?"
                             b"$$ &Changed $$ &Deleted"
                         )
                         % prompts,
                         0,
                     )
                     == 0
                 ):
                     mresult.addfile(
                         f,
                         mergestatemod.ACTION_GET,
                         (flags, False),
                         b'prompt recreating',
                     )
                 else:
                     mresult.removefile(f)
 
         if not util.fscasesensitive(repo.path):
             # check collision between files only in p2 for clean update
             if not branchmerge and (
                 force or not wc.dirty(missing=True, branch=False)
             ):
                 _checkcollision(repo, p2.manifest(), None)
             else:
                 _checkcollision(repo, wc.manifest(), mresult)
 
         # divergent renames
         for f, fl in sorted(pycompat.iteritems(mresult.diverge)):
             repo.ui.warn(
                 _(
                     b"note: possible conflict - %s was renamed "
                     b"multiple times to:\n"
                 )
                 % f
             )
             for nf in sorted(fl):
                 repo.ui.warn(b" %s\n" % nf)
 
         # rename and delete
         for f, fl in sorted(pycompat.iteritems(mresult.renamedelete)):
             repo.ui.warn(
                 _(
                     b"note: possible conflict - %s was deleted "
                     b"and renamed to:\n"
                 )
                 % f
             )
             for nf in sorted(fl):
                 repo.ui.warn(b" %s\n" % nf)
 
         ### apply phase
         if not branchmerge:  # just jump to the new rev
             fp1, fp2, xp1, xp2 = fp2, repo.nullid, xp2, b''
         # If we're doing a partial update, we need to skip updating
         # the dirstate.
         always = matcher is None or matcher.always()
         updatedirstate = updatedirstate and always and not wc.isinmemory()
         if updatedirstate:
             repo.hook(b'preupdate', throw=True, parent1=xp1, parent2=xp2)
             # note that we're in the middle of an update
             repo.vfs.write(b'updatestate', p2.hex())
 
         _advertisefsmonitor(
             repo, mresult.len((mergestatemod.ACTION_GET,)), p1.node()
         )
 
         wantfiledata = updatedirstate and not branchmerge
         stats, getfiledata, extraactions = applyupdates(
             repo,
             mresult,
             wc,
             p2,
             overwrite,
             wantfiledata,
             labels=labels,
         )
 
         if updatedirstate:
             if extraactions:
                 for k, acts in pycompat.iteritems(extraactions):
                     for a in acts:
                         mresult.addfile(a[0], k, *a[1:])
                     if k == mergestatemod.ACTION_GET and wantfiledata:
                         # no filedata until mergestate is updated to provide it
                         for a in acts:
                             getfiledata[a[0]] = None
 
             assert len(getfiledata) == (
                 mresult.len((mergestatemod.ACTION_GET,)) if wantfiledata else 0
             )
             with repo.dirstate.parentchange():
                 repo.setparents(fp1, fp2)
                 mergestatemod.recordupdates(
                     repo, mresult.actionsdict, branchmerge, getfiledata
                 )
                 # update completed, clear state
                 util.unlink(repo.vfs.join(b'updatestate'))
 
                 if not branchmerge:
                     repo.dirstate.setbranch(p2.branch())
 
                 # If we're updating to a location, clean up any stale temporary includes
                 # (ex: this happens during hg rebase --abort).
                 if not branchmerge:
                     sparse.prunetemporaryincludes(repo)
 
     if updatedirstate:
         repo.hook(
             b'update', parent1=xp1, parent2=xp2, error=stats.unresolvedcount
         )
     return stats
 
 
 def merge(ctx, labels=None, force=False, wc=None):
     """Merge another topological branch into the working copy.
 
     force = whether the merge was run with 'merge --force' (deprecated)
     """
 
     return _update(
         ctx.repo(),
         ctx.rev(),
         labels=labels,
         branchmerge=True,
         force=force,
         mergeforce=force,
         wc=wc,
     )
 
 
 def update(ctx, updatecheck=None, wc=None):
     """Do a regular update to the given commit, aborting if there are conflicts.
 
     The 'updatecheck' argument can be used to control what to do in case of
     conflicts.
 
     Note: This is a new, higher-level update() than the one that used to exist
     in this module. That function is now called _update(). You can hopefully
     replace your callers to use this new update(), or clean_update(), merge(),
     revert_to(), or graft().
     """
     return _update(
         ctx.repo(),
         ctx.rev(),
         branchmerge=False,
         force=False,
         labels=[b'working copy', b'destination'],
         updatecheck=updatecheck,
         wc=wc,
     )
 
 
 def clean_update(ctx, wc=None):
     """Do a clean update to the given commit.
 
     This involves updating to the commit and discarding any changes in the
     working copy.
     """
     return _update(ctx.repo(), ctx.rev(), branchmerge=False, force=True, wc=wc)
 
 
 def revert_to(ctx, matcher=None, wc=None):
     """Revert the working copy to the given commit.
 
     The working copy will keep its current parent(s) but its content will
     be the same as in the given commit.
     """
 
     return _update(
         ctx.repo(),
         ctx.rev(),
         branchmerge=False,
         force=True,
         updatedirstate=False,
         matcher=matcher,
         wc=wc,
     )
 
 
 def graft(
     repo,
     ctx,
     base=None,
     labels=None,
     keepparent=False,
     keepconflictparent=False,
     wctx=None,
 ):
     """Do a graft-like merge.
 
     This is a merge where the merge ancestor is chosen such that one
     or more changesets are grafted onto the current changeset. In
     addition to the merge, this fixes up the dirstate to include only
     a single parent (if keepparent is False) and tries to duplicate any
     renames/copies appropriately.
 
     ctx - changeset to rebase
     base - merge base, or ctx.p1() if not specified
     labels - merge labels eg ['local', 'graft']
     keepparent - keep second parent if any
     keepconflictparent - if unresolved, keep parent used for the merge
 
     """
     # If we're grafting a descendant onto an ancestor, be sure to pass
     # mergeancestor=True to update. This does two things: 1) allows the merge if
     # the destination is the same as the parent of the ctx (so we can use graft
     # to copy commits), and 2) informs update that the incoming changes are
     # newer than the destination so it doesn't prompt about "remote changed foo
     # which local deleted".
     # We also pass mergeancestor=True when base is the same revision as p1. 2)
     # doesn't matter as there can't possibly be conflicts, but 1) is necessary.
     wctx = wctx or repo[None]
     pctx = wctx.p1()
     base = base or ctx.p1()
     mergeancestor = (
         repo.changelog.isancestor(pctx.node(), ctx.node())
         or pctx.rev() == base.rev()
     )
 
     stats = _update(
         repo,
         ctx.node(),
         True,
         True,
         base.node(),
         mergeancestor=mergeancestor,
         labels=labels,
         wc=wctx,
     )
 
     if keepconflictparent and stats.unresolvedcount:
         pother = ctx.node()
     else:
         pother = repo.nullid
         parents = ctx.parents()
         if keepparent and len(parents) == 2 and base in parents:
             parents.remove(base)
             pother = parents[0].node()
     # Never set both parents equal to each other
     if pother == pctx.node():
         pother = repo.nullid
 
     if wctx.isinmemory():
         wctx.setparents(pctx.node(), pother)
         # fix up dirstate for copies and renames
         copies.graftcopies(wctx, ctx, base)
     else:
         with repo.dirstate.parentchange():
             repo.setparents(pctx.node(), pother)
             repo.dirstate.write(repo.currenttransaction())
             # fix up dirstate for copies and renames
             copies.graftcopies(wctx, ctx, base)
     return stats
 
 
 def back_out(ctx, parent=None, wc=None):
     if parent is None:
         if ctx.p2() is not None:
             raise error.ProgrammingError(
                 b"must specify parent of merge commit to back out"
             )
         parent = ctx.p1()
     return _update(
         ctx.repo(),
         parent,
         branchmerge=True,
         force=True,
         ancestor=ctx.node(),
         mergeancestor=False,
     )
 
 
 def purge(
     repo,
     matcher,
     unknown=True,
     ignored=False,
     removeemptydirs=True,
     removefiles=True,
     abortonerror=False,
     noop=False,
     confirm=False,
 ):
     """Purge the working directory of untracked files.
 
     ``matcher`` is a matcher configured to scan the working directory -
     potentially a subset.
 
     ``unknown`` controls whether unknown files should be purged.
 
     ``ignored`` controls whether ignored files should be purged.
 
     ``removeemptydirs`` controls whether empty directories should be removed.
 
     ``removefiles`` controls whether files are removed.
 
     ``abortonerror`` causes an exception to be raised if an error occurs
     deleting a file or directory.
 
     ``noop`` controls whether to actually remove files. If not defined, actions
     will be taken.
 
     ``confirm`` ask confirmation before actually removing anything.
 
     Returns an iterable of relative paths in the working directory that were
     or would be removed.
     """
 
     def remove(removefn, path):
         try:
             removefn(path)
         except OSError:
             m = _(b'%s cannot be removed') % path
             if abortonerror:
                 raise error.Abort(m)
             else:
                 repo.ui.warn(_(b'warning: %s\n') % m)
 
     # There's no API to copy a matcher. So mutate the passed matcher and
     # restore it when we're done.
     oldtraversedir = matcher.traversedir
 
     res = []
 
     try:
         if removeemptydirs:
             directories = []
             matcher.traversedir = directories.append
 
         status = repo.status(match=matcher, ignored=ignored, unknown=unknown)
 
         if confirm:
             nb_ignored = len(status.ignored)
             nb_unkown = len(status.unknown)
             if nb_unkown and nb_ignored:
                 msg = _(b"permanently delete %d unkown and %d ignored files?")
                 msg %= (nb_unkown, nb_ignored)
             elif nb_unkown:
                 msg = _(b"permanently delete %d unkown files?")
                 msg %= nb_unkown
             elif nb_ignored:
                 msg = _(b"permanently delete %d ignored files?")
                 msg %= nb_ignored
             elif removeemptydirs:
                 dir_count = 0
                 for f in directories:
                     if matcher(f) and not repo.wvfs.listdir(f):
                         dir_count += 1
                 if dir_count:
                     msg = _(
                         b"permanently delete at least %d empty directories?"
                     )
                     msg %= dir_count
                 else:
                     # XXX we might be missing directory there
                     return res
             msg += b" (yN)$$ &Yes $$ &No"
             if repo.ui.promptchoice(msg, default=1) == 1:
                 raise error.CanceledError(_(b'removal cancelled'))
 
         if removefiles:
             for f in sorted(status.unknown + status.ignored):
                 if not noop:
                     repo.ui.note(_(b'removing file %s\n') % f)
                     remove(repo.wvfs.unlink, f)
                 res.append(f)
 
         if removeemptydirs:
             for f in sorted(directories, reverse=True):
                 if matcher(f) and not repo.wvfs.listdir(f):
                     if not noop:
                         repo.ui.note(_(b'removing directory %s\n') % f)
                         remove(repo.wvfs.rmdir, f)
                     res.append(f)
 
         return res
 
     finally:
         matcher.traversedir = oldtraversedir

mercurial/revlogutils/rewrite.py

@@ -1,886 +1,888 b''
 # censor code related to censoring revision
 # coding: utf8
 #
 # Copyright 2021 Pierre-Yves David <[e-mail]>
 # Copyright 2015 Google, Inc <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 import binascii
 import contextlib
 import os
 import struct
 
 from ..node import (
     nullrev,
 )
 from .constants import (
     COMP_MODE_PLAIN,
     ENTRY_DATA_COMPRESSED_LENGTH,
     ENTRY_DATA_COMPRESSION_MODE,
     ENTRY_DATA_OFFSET,
     ENTRY_DATA_UNCOMPRESSED_LENGTH,
     ENTRY_DELTA_BASE,
     ENTRY_LINK_REV,
     ENTRY_NODE_ID,
     ENTRY_PARENT_1,
     ENTRY_PARENT_2,
     ENTRY_SIDEDATA_COMPRESSED_LENGTH,
     ENTRY_SIDEDATA_COMPRESSION_MODE,
     ENTRY_SIDEDATA_OFFSET,
     REVIDX_ISCENSORED,
     REVLOGV0,
     REVLOGV1,
 )
 from ..i18n import _
 
 from .. import (
     error,
     mdiff,
     pycompat,
     revlogutils,
     util,
 )
 from ..utils import (
     storageutil,
 )
 from . import (
     constants,
     deltas,
 )
 
 
 def v1_censor(rl, tr, censornode, tombstone=b''):
     """censors a revision in a "version 1" revlog"""
     assert rl._format_version == constants.REVLOGV1, rl._format_version
 
     # avoid cycle
     from .. import revlog
 
     censorrev = rl.rev(censornode)
     tombstone = storageutil.packmeta({b'censored': tombstone}, b'')
 
     # Rewriting the revlog in place is hard. Our strategy for censoring is
     # to create a new revlog, copy all revisions to it, then replace the
     # revlogs on transaction close.
     #
     # This is a bit dangerous. We could easily have a mismatch of state.
     newrl = revlog.revlog(
         rl.opener,
         target=rl.target,
         radix=rl.radix,
         postfix=b'tmpcensored',
         censorable=True,
     )
     newrl._format_version = rl._format_version
     newrl._format_flags = rl._format_flags
     newrl._generaldelta = rl._generaldelta
     newrl._parse_index = rl._parse_index
 
     for rev in rl.revs():
         node = rl.node(rev)
         p1, p2 = rl.parents(node)
 
         if rev == censorrev:
             newrl.addrawrevision(
                 tombstone,
                 tr,
                 rl.linkrev(censorrev),
                 p1,
                 p2,
                 censornode,
                 constants.REVIDX_ISCENSORED,
             )
 
             if newrl.deltaparent(rev) != nullrev:
                 m = _(b'censored revision stored as delta; cannot censor')
                 h = _(
                     b'censoring of revlogs is not fully implemented;'
                     b' please report this bug'
                 )
                 raise error.Abort(m, hint=h)
             continue
 
         if rl.iscensored(rev):
             if rl.deltaparent(rev) != nullrev:
                 m = _(
                     b'cannot censor due to censored '
                     b'revision having delta stored'
                 )
                 raise error.Abort(m)
             rawtext = rl._chunk(rev)
         else:
             rawtext = rl.rawdata(rev)
 
         newrl.addrawrevision(
             rawtext, tr, rl.linkrev(rev), p1, p2, node, rl.flags(rev)
         )
 
     tr.addbackup(rl._indexfile, location=b'store')
     if not rl._inline:
         tr.addbackup(rl._datafile, location=b'store')
 
     rl.opener.rename(newrl._indexfile, rl._indexfile)
     if not rl._inline:
         rl.opener.rename(newrl._datafile, rl._datafile)
 
     rl.clearcaches()
     rl._loadindex()
 
 
 def v2_censor(revlog, tr, censornode, tombstone=b''):
     """censors a revision in a "version 2" revlog"""
     assert revlog._format_version != REVLOGV0, revlog._format_version
     assert revlog._format_version != REVLOGV1, revlog._format_version
 
     censor_revs = {revlog.rev(censornode)}
     _rewrite_v2(revlog, tr, censor_revs, tombstone)
 
 
 def _rewrite_v2(revlog, tr, censor_revs, tombstone=b''):
     """rewrite a revlog to censor some of its content
 
     General principle
 
     We create new revlog files (index/data/sidedata) to copy the content of
     the existing data without the censored data.
 
     We need to recompute new delta for any revision that used the censored
     revision as delta base. As the cumulative size of the new delta may be
     large, we store them in a temporary file until they are stored in their
     final destination.
 
     All data before the censored data can be blindly copied. The rest needs
     to be copied as we go and the associated index entry needs adjustement.
     """
     assert revlog._format_version != REVLOGV0, revlog._format_version
     assert revlog._format_version != REVLOGV1, revlog._format_version
 
     old_index = revlog.index
     docket = revlog._docket
 
     tombstone = storageutil.packmeta({b'censored': tombstone}, b'')
 
     first_excl_rev = min(censor_revs)
 
     first_excl_entry = revlog.index[first_excl_rev]
     index_cutoff = revlog.index.entry_size * first_excl_rev
     data_cutoff = first_excl_entry[ENTRY_DATA_OFFSET] >> 16
     sidedata_cutoff = revlog.sidedata_cut_off(first_excl_rev)
 
     with pycompat.unnamedtempfile(mode=b"w+b") as tmp_storage:
         # rev → (new_base, data_start, data_end, compression_mode)
         rewritten_entries = _precompute_rewritten_delta(
             revlog,
             old_index,
             censor_revs,
             tmp_storage,
         )
 
         all_files = _setup_new_files(
             revlog,
             index_cutoff,
             data_cutoff,
             sidedata_cutoff,
         )
 
         # we dont need to open the old index file since its content already
         # exist in a usable form in `old_index`.
         with all_files() as open_files:
             (
                 old_data_file,
                 old_sidedata_file,
                 new_index_file,
                 new_data_file,
                 new_sidedata_file,
             ) = open_files
 
             # writing the censored revision
 
             # Writing all subsequent revisions
             for rev in range(first_excl_rev, len(old_index)):
                 if rev in censor_revs:
                     _rewrite_censor(
                         revlog,
                         old_index,
                         open_files,
                         rev,
                         tombstone,
                     )
                 else:
                     _rewrite_simple(
                         revlog,
                         old_index,
                         open_files,
                         rev,
                         rewritten_entries,
                         tmp_storage,
                     )
     docket.write(transaction=None, stripping=True)
 
 
 def _precompute_rewritten_delta(
     revlog,
     old_index,
     excluded_revs,
     tmp_storage,
 ):
     """Compute new delta for revisions whose delta is based on revision that
     will not survive as is.
 
     Return a mapping: {rev → (new_base, data_start, data_end, compression_mode)}
     """
     dc = deltas.deltacomputer(revlog)
     rewritten_entries = {}
     first_excl_rev = min(excluded_revs)
     with revlog._segmentfile._open_read() as dfh:
         for rev in range(first_excl_rev, len(old_index)):
             if rev in excluded_revs:
                 # this revision will be preserved as is, so we don't need to
                 # consider recomputing a delta.
                 continue
             entry = old_index[rev]
             if entry[ENTRY_DELTA_BASE] not in excluded_revs:
                 continue
             # This is a revision that use the censored revision as the base
             # for its delta. We need a need new deltas
             if entry[ENTRY_DATA_UNCOMPRESSED_LENGTH] == 0:
                 # this revision is empty, we can delta against nullrev
                 rewritten_entries[rev] = (nullrev, 0, 0, COMP_MODE_PLAIN)
             else:
 
                 text = revlog.rawdata(rev, _df=dfh)
                 info = revlogutils.revisioninfo(
                     node=entry[ENTRY_NODE_ID],
                     p1=revlog.node(entry[ENTRY_PARENT_1]),
                     p2=revlog.node(entry[ENTRY_PARENT_2]),
                     btext=[text],
                     textlen=len(text),
                     cachedelta=None,
                     flags=entry[ENTRY_DATA_OFFSET] & 0xFFFF,
                 )
                 d = dc.finddeltainfo(
                     info, dfh, excluded_bases=excluded_revs, target_rev=rev
                 )
                 default_comp = revlog._docket.default_compression_header
                 comp_mode, d = deltas.delta_compression(default_comp, d)
                 # using `tell` is a bit lazy, but we are not here for speed
                 start = tmp_storage.tell()
                 tmp_storage.write(d.data[1])
                 end = tmp_storage.tell()
                 rewritten_entries[rev] = (d.base, start, end, comp_mode)
     return rewritten_entries
 
 
 def _setup_new_files(
     revlog,
     index_cutoff,
     data_cutoff,
     sidedata_cutoff,
 ):
     """
 
     return a context manager to open all the relevant files:
     - old_data_file,
     - old_sidedata_file,
     - new_index_file,
     - new_data_file,
     - new_sidedata_file,
 
     The old_index_file is not here because it is accessed through the
     `old_index` object if the caller function.
     """
     docket = revlog._docket
     old_index_filepath = revlog.opener.join(docket.index_filepath())
     old_data_filepath = revlog.opener.join(docket.data_filepath())
     old_sidedata_filepath = revlog.opener.join(docket.sidedata_filepath())
 
     new_index_filepath = revlog.opener.join(docket.new_index_file())
     new_data_filepath = revlog.opener.join(docket.new_data_file())
     new_sidedata_filepath = revlog.opener.join(docket.new_sidedata_file())
 
     util.copyfile(old_index_filepath, new_index_filepath, nb_bytes=index_cutoff)
     util.copyfile(old_data_filepath, new_data_filepath, nb_bytes=data_cutoff)
     util.copyfile(
         old_sidedata_filepath,
         new_sidedata_filepath,
         nb_bytes=sidedata_cutoff,
     )
     revlog.opener.register_file(docket.index_filepath())
     revlog.opener.register_file(docket.data_filepath())
     revlog.opener.register_file(docket.sidedata_filepath())
 
     docket.index_end = index_cutoff
     docket.data_end = data_cutoff
     docket.sidedata_end = sidedata_cutoff
 
     # reload the revlog internal information
     revlog.clearcaches()
     revlog._loadindex(docket=docket)
 
     @contextlib.contextmanager
     def all_files_opener():
         # hide opening in an helper function to please check-code, black
         # and various python version at the same time
         with open(old_data_filepath, 'rb') as old_data_file:
             with open(old_sidedata_filepath, 'rb') as old_sidedata_file:
                 with open(new_index_filepath, 'r+b') as new_index_file:
                     with open(new_data_filepath, 'r+b') as new_data_file:
                         with open(
                             new_sidedata_filepath, 'r+b'
                         ) as new_sidedata_file:
                             new_index_file.seek(0, os.SEEK_END)
                             assert new_index_file.tell() == index_cutoff
                             new_data_file.seek(0, os.SEEK_END)
                             assert new_data_file.tell() == data_cutoff
                             new_sidedata_file.seek(0, os.SEEK_END)
                             assert new_sidedata_file.tell() == sidedata_cutoff
                             yield (
                                 old_data_file,
                                 old_sidedata_file,
                                 new_index_file,
                                 new_data_file,
                                 new_sidedata_file,
                             )
 
     return all_files_opener
 
 
 def _rewrite_simple(
     revlog,
     old_index,
     all_files,
     rev,
     rewritten_entries,
     tmp_storage,
 ):
     """append a normal revision to the index after the rewritten one(s)"""
     (
         old_data_file,
         old_sidedata_file,
         new_index_file,
         new_data_file,
         new_sidedata_file,
     ) = all_files
     entry = old_index[rev]
     flags = entry[ENTRY_DATA_OFFSET] & 0xFFFF
     old_data_offset = entry[ENTRY_DATA_OFFSET] >> 16
 
     if rev not in rewritten_entries:
         old_data_file.seek(old_data_offset)
         new_data_size = entry[ENTRY_DATA_COMPRESSED_LENGTH]
         new_data = old_data_file.read(new_data_size)
         data_delta_base = entry[ENTRY_DELTA_BASE]
         d_comp_mode = entry[ENTRY_DATA_COMPRESSION_MODE]
     else:
         (
             data_delta_base,
             start,
             end,
             d_comp_mode,
         ) = rewritten_entries[rev]
         new_data_size = end - start
         tmp_storage.seek(start)
         new_data = tmp_storage.read(new_data_size)
 
     # It might be faster to group continuous read/write operation,
     # however, this is censor, an operation that is not focussed
     # around stellar performance. So I have not written this
     # optimisation yet.
     new_data_offset = new_data_file.tell()
     new_data_file.write(new_data)
 
     sidedata_size = entry[ENTRY_SIDEDATA_COMPRESSED_LENGTH]
     new_sidedata_offset = new_sidedata_file.tell()
     if 0 < sidedata_size:
         old_sidedata_offset = entry[ENTRY_SIDEDATA_OFFSET]
         old_sidedata_file.seek(old_sidedata_offset)
         new_sidedata = old_sidedata_file.read(sidedata_size)
         new_sidedata_file.write(new_sidedata)
 
     data_uncompressed_length = entry[ENTRY_DATA_UNCOMPRESSED_LENGTH]
     sd_com_mode = entry[ENTRY_SIDEDATA_COMPRESSION_MODE]
     assert data_delta_base <= rev, (data_delta_base, rev)
 
     new_entry = revlogutils.entry(
         flags=flags,
         data_offset=new_data_offset,
         data_compressed_length=new_data_size,
         data_uncompressed_length=data_uncompressed_length,
         data_delta_base=data_delta_base,
         link_rev=entry[ENTRY_LINK_REV],
         parent_rev_1=entry[ENTRY_PARENT_1],
         parent_rev_2=entry[ENTRY_PARENT_2],
         node_id=entry[ENTRY_NODE_ID],
         sidedata_offset=new_sidedata_offset,
         sidedata_compressed_length=sidedata_size,
         data_compression_mode=d_comp_mode,
         sidedata_compression_mode=sd_com_mode,
     )
     revlog.index.append(new_entry)
     entry_bin = revlog.index.entry_binary(rev)
     new_index_file.write(entry_bin)
 
     revlog._docket.index_end = new_index_file.tell()
     revlog._docket.data_end = new_data_file.tell()
     revlog._docket.sidedata_end = new_sidedata_file.tell()
 
 
 def _rewrite_censor(
     revlog,
     old_index,
     all_files,
     rev,
     tombstone,
 ):
     """rewrite and append a censored revision"""
     (
         old_data_file,
         old_sidedata_file,
         new_index_file,
         new_data_file,
         new_sidedata_file,
     ) = all_files
     entry = old_index[rev]
 
     # XXX consider trying the default compression too
     new_data_size = len(tombstone)
     new_data_offset = new_data_file.tell()
     new_data_file.write(tombstone)
 
     # we are not adding any sidedata as they might leak info about the censored version
 
     link_rev = entry[ENTRY_LINK_REV]
 
     p1 = entry[ENTRY_PARENT_1]
     p2 = entry[ENTRY_PARENT_2]
 
     new_entry = revlogutils.entry(
         flags=constants.REVIDX_ISCENSORED,
         data_offset=new_data_offset,
         data_compressed_length=new_data_size,
         data_uncompressed_length=new_data_size,
         data_delta_base=rev,
         link_rev=link_rev,
         parent_rev_1=p1,
         parent_rev_2=p2,
         node_id=entry[ENTRY_NODE_ID],
         sidedata_offset=0,
         sidedata_compressed_length=0,
         data_compression_mode=COMP_MODE_PLAIN,
         sidedata_compression_mode=COMP_MODE_PLAIN,
     )
     revlog.index.append(new_entry)
     entry_bin = revlog.index.entry_binary(rev)
     new_index_file.write(entry_bin)
     revlog._docket.index_end = new_index_file.tell()
     revlog._docket.data_end = new_data_file.tell()
 
 
 def _get_filename_from_filelog_index(path):
     # Drop the extension and the `data/` prefix
     path_part = path.rsplit(b'.', 1)[0].split(b'/', 1)
     if len(path_part) < 2:
         msg = _(b"cannot recognize filelog from filename: '%s'")
         msg %= path
         raise error.Abort(msg)
 
     return path_part[1]
 
 
 def _filelog_from_filename(repo, path):
     """Returns the filelog for the given `path`. Stolen from `engine.py`"""
 
     from .. import filelog  # avoid cycle
 
     fl = filelog.filelog(repo.svfs, path)
     return fl
 
 
 def _write_swapped_parents(repo, rl, rev, offset, fp):
     """Swaps p1 and p2 and overwrites the revlog entry for `rev` in `fp`"""
     from ..pure import parsers  # avoid cycle
 
     if repo._currentlock(repo._lockref) is None:
         # Let's be paranoid about it
         msg = "repo needs to be locked to rewrite parents"
         raise error.ProgrammingError(msg)
 
     index_format = parsers.IndexObject.index_format
     entry = rl.index[rev]
     new_entry = list(entry)
     new_entry[5], new_entry[6] = entry[6], entry[5]
     packed = index_format.pack(*new_entry[:8])
     fp.seek(offset)
     fp.write(packed)
 
 
 def _reorder_filelog_parents(repo, fl, to_fix):
     """
     Swaps p1 and p2 for all `to_fix` revisions of filelog `fl` and writes the
     new version to disk, overwriting the old one with a rename.
     """
     from ..pure import parsers  # avoid cycle
 
     ui = repo.ui
     assert len(to_fix) > 0
     rl = fl._revlog
     if rl._format_version != constants.REVLOGV1:
         msg = "expected version 1 revlog, got version '%d'" % rl._format_version
         raise error.ProgrammingError(msg)
 
     index_file = rl._indexfile
     new_file_path = index_file + b'.tmp-parents-fix'
     repaired_msg = _(b"repaired revision %d of 'filelog %s'\n")
 
     with ui.uninterruptible():
         try:
             util.copyfile(
                 rl.opener.join(index_file),
                 rl.opener.join(new_file_path),
                 checkambig=rl._checkambig,
             )
 
             with rl.opener(new_file_path, mode=b"r+") as fp:
                 if rl._inline:
                     index = parsers.InlinedIndexObject(fp.read())
                     for rev in fl.revs():
                         if rev in to_fix:
                             offset = index._calculate_index(rev)
                             _write_swapped_parents(repo, rl, rev, offset, fp)
                             ui.write(repaired_msg % (rev, index_file))
                 else:
                     index_format = parsers.IndexObject.index_format
                     for rev in to_fix:
                         offset = rev * index_format.size
                         _write_swapped_parents(repo, rl, rev, offset, fp)
                         ui.write(repaired_msg % (rev, index_file))
 
             rl.opener.rename(new_file_path, index_file)
             rl.clearcaches()
             rl._loadindex()
         finally:
             util.tryunlink(new_file_path)
 
 
 def _is_revision_affected(fl, filerev, metadata_cache=None):
     full_text = lambda: fl._revlog.rawdata(filerev)
     parent_revs = lambda: fl._revlog.parentrevs(filerev)
     return _is_revision_affected_inner(
         full_text, parent_revs, filerev, metadata_cache
     )
 
 
 def _is_revision_affected_inner(
     full_text,
     parents_revs,
     filerev,
     metadata_cache=None,
 ):
     """Mercurial currently (5.9rc0) uses `p1 == nullrev and p2 != nullrev` as a
     special meaning compared to the reverse in the context of filelog-based
     copytracing. issue6528 exists because new code assumed that parent ordering
     didn't matter, so this detects if the revision contains metadata (since
     it's only used for filelog-based copytracing) and its parents are in the
     "wrong" order."""
     try:
         raw_text = full_text()
     except error.CensoredNodeError:
         # We don't care about censored nodes as they never carry metadata
         return False
-    has_meta = raw_text.startswith(b'\x01\n')
+
+    # raw text can be a `memoryview`, which doesn't implement `startswith`
+    has_meta = bytes(raw_text[:2]) == b'\x01\n'
     if metadata_cache is not None:
         metadata_cache[filerev] = has_meta
     if has_meta:
         (p1, p2) = parents_revs()
         if p1 != nullrev and p2 == nullrev:
             return True
     return False
 
 
 def _is_revision_affected_fast(repo, fl, filerev, metadata_cache):
     rl = fl._revlog
     is_censored = lambda: rl.iscensored(filerev)
     delta_base = lambda: rl.deltaparent(filerev)
     delta = lambda: rl._chunk(filerev)
     full_text = lambda: rl.rawdata(filerev)
     parent_revs = lambda: rl.parentrevs(filerev)
     return _is_revision_affected_fast_inner(
         is_censored,
         delta_base,
         delta,
         full_text,
         parent_revs,
         filerev,
         metadata_cache,
     )
 
 
 def _is_revision_affected_fast_inner(
     is_censored,
     delta_base,
     delta,
     full_text,
     parent_revs,
     filerev,
     metadata_cache,
 ):
     """Optimization fast-path for `_is_revision_affected`.
 
     `metadata_cache` is a dict of `{rev: has_metadata}` which allows any
     revision to check if its base has metadata, saving computation of the full
     text, instead looking at the current delta.
 
     This optimization only works if the revisions are looked at in order."""
 
     if is_censored():
         # Censored revisions don't contain metadata, so they cannot be affected
         metadata_cache[filerev] = False
         return False
 
     p1, p2 = parent_revs()
     if p1 == nullrev or p2 != nullrev:
         return False
 
     delta_parent = delta_base()
     parent_has_metadata = metadata_cache.get(delta_parent)
     if parent_has_metadata is None:
         return _is_revision_affected_inner(
             full_text,
             parent_revs,
             filerev,
             metadata_cache,
         )
 
     chunk = delta()
     if not len(chunk):
         # No diff for this revision
         return parent_has_metadata
 
     header_length = 12
     if len(chunk) < header_length:
         raise error.Abort(_(b"patch cannot be decoded"))
 
     start, _end, _length = struct.unpack(b">lll", chunk[:header_length])
 
     if start < 2:  # len(b'\x01\n') == 2
         # This delta does *something* to the metadata marker (if any).
         # Check it the slow way
         is_affected = _is_revision_affected_inner(
             full_text,
             parent_revs,
             filerev,
             metadata_cache,
         )
         return is_affected
 
     # The diff did not remove or add the metadata header, it's then in the same
     # situation as its parent
     metadata_cache[filerev] = parent_has_metadata
     return parent_has_metadata
 
 
 def _from_report(ui, repo, context, from_report, dry_run):
     """
     Fix the revisions given in the `from_report` file, but still checks if the
     revisions are indeed affected to prevent an unfortunate cyclic situation
     where we'd swap well-ordered parents again.
 
     See the doc for `debug_fix_issue6528` for the format documentation.
     """
     ui.write(_(b"loading report file '%s'\n") % from_report)
 
     with context(), open(from_report, mode='rb') as f:
         for line in f.read().split(b'\n'):
             if not line:
                 continue
             filenodes, filename = line.split(b' ', 1)
             fl = _filelog_from_filename(repo, filename)
             to_fix = set(
                 fl.rev(binascii.unhexlify(n)) for n in filenodes.split(b',')
             )
             excluded = set()
 
             for filerev in to_fix:
                 if _is_revision_affected(fl, filerev):
                     msg = b"found affected revision %d for filelog '%s'\n"
                     ui.warn(msg % (filerev, filename))
                 else:
                     msg = _(b"revision %s of file '%s' is not affected\n")
                     msg %= (binascii.hexlify(fl.node(filerev)), filename)
                     ui.warn(msg)
                     excluded.add(filerev)
 
             to_fix = to_fix - excluded
             if not to_fix:
                 msg = _(b"no affected revisions were found for '%s'\n")
                 ui.write(msg % filename)
                 continue
             if not dry_run:
                 _reorder_filelog_parents(repo, fl, sorted(to_fix))
 
 
 def filter_delta_issue6528(revlog, deltas_iter):
     """filter incomind deltas to repaire issue 6528 on the fly"""
     metadata_cache = {}
 
     deltacomputer = deltas.deltacomputer(revlog)
 
     for rev, d in enumerate(deltas_iter, len(revlog)):
         (
             node,
             p1_node,
             p2_node,
             linknode,
             deltabase,
             delta,
             flags,
             sidedata,
         ) = d
 
         if not revlog.index.has_node(deltabase):
             raise error.LookupError(
                 deltabase, revlog.radix, _(b'unknown parent')
             )
         base_rev = revlog.rev(deltabase)
         if not revlog.index.has_node(p1_node):
             raise error.LookupError(p1_node, revlog.radix, _(b'unknown parent'))
         p1_rev = revlog.rev(p1_node)
         if not revlog.index.has_node(p2_node):
             raise error.LookupError(p2_node, revlog.radix, _(b'unknown parent'))
         p2_rev = revlog.rev(p2_node)
 
         is_censored = lambda: bool(flags & REVIDX_ISCENSORED)
         delta_base = lambda: revlog.rev(delta_base)
         delta_base = lambda: base_rev
         parent_revs = lambda: (p1_rev, p2_rev)
 
         def full_text():
             # note: being able to reuse the full text computation in the
             # underlying addrevision would be useful however this is a bit too
             # intrusive the for the "quick" issue6528 we are writing before the
             # 5.8 release
             textlen = mdiff.patchedsize(revlog.size(base_rev), delta)
 
             revinfo = revlogutils.revisioninfo(
                 node,
                 p1_node,
                 p2_node,
                 [None],
                 textlen,
                 (base_rev, delta),
                 flags,
             )
             # cached by the global "writing" context
             assert revlog._writinghandles is not None
             if revlog._inline:
                 fh = revlog._writinghandles[0]
             else:
                 fh = revlog._writinghandles[1]
             return deltacomputer.buildtext(revinfo, fh)
 
         is_affected = _is_revision_affected_fast_inner(
             is_censored,
             delta_base,
             lambda: delta,
             full_text,
             parent_revs,
             rev,
             metadata_cache,
         )
         if is_affected:
             d = (
                 node,
                 p2_node,
                 p1_node,
                 linknode,
                 deltabase,
                 delta,
                 flags,
                 sidedata,
             )
         yield d
 
 
 def repair_issue6528(
     ui, repo, dry_run=False, to_report=None, from_report=None, paranoid=False
 ):
     from .. import store  # avoid cycle
 
     @contextlib.contextmanager
     def context():
         if dry_run or to_report:  # No need for locking
             yield
         else:
             with repo.wlock(), repo.lock():
                 yield
 
     if from_report:
         return _from_report(ui, repo, context, from_report, dry_run)
 
     report_entries = []
 
     with context():
         files = list(
             (file_type, path)
             for (file_type, path, _s) in repo.store.datafiles()
             if path.endswith(b'.i') and file_type & store.FILEFLAGS_FILELOG
         )
 
         progress = ui.makeprogress(
             _(b"looking for affected revisions"),
             unit=_(b"filelogs"),
             total=len(files),
         )
         found_nothing = True
 
         for file_type, path in files:
             if (
                 not path.endswith(b'.i')
                 or not file_type & store.FILEFLAGS_FILELOG
             ):
                 continue
             progress.increment()
             filename = _get_filename_from_filelog_index(path)
             fl = _filelog_from_filename(repo, filename)
 
             # Set of filerevs (or hex filenodes if `to_report`) that need fixing
             to_fix = set()
             metadata_cache = {}
             for filerev in fl.revs():
                 affected = _is_revision_affected_fast(
                     repo, fl, filerev, metadata_cache
                 )
                 if paranoid:
                     slow = _is_revision_affected(fl, filerev)
                     if slow != affected:
                         msg = _(b"paranoid check failed for '%s' at node %s")
                         node = binascii.hexlify(fl.node(filerev))
                         raise error.Abort(msg % (filename, node))
                 if affected:
                     msg = b"found affected revision %d for filelog '%s'\n"
                     ui.warn(msg % (filerev, path))
                     found_nothing = False
                     if not dry_run:
                         if to_report:
                             to_fix.add(binascii.hexlify(fl.node(filerev)))
                         else:
                             to_fix.add(filerev)
 
             if to_fix:
                 to_fix = sorted(to_fix)
                 if to_report:
                     report_entries.append((filename, to_fix))
                 else:
                     _reorder_filelog_parents(repo, fl, to_fix)
 
         if found_nothing:
             ui.write(_(b"no affected revisions were found\n"))
 
         if to_report and report_entries:
             with open(to_report, mode="wb") as f:
                 for path, to_fix in report_entries:
                     f.write(b"%s %s\n" % (b",".join(to_fix), path))
 
         progress.complete()

rust/hg-cpython/src/revlog.rs

@@ -1,529 +1,512 b''
 // revlog.rs
 //
 // Copyright 2019-2020 Georges Racinet <[e-mail]>
 //
 // This software may be used and distributed according to the terms of the
 // GNU General Public License version 2 or any later version.
 
 use crate::{
     cindex,
     utils::{node_from_py_bytes, node_from_py_object},
 };
 use cpython::{
     buffer::{Element, PyBuffer},
     exc::{IndexError, ValueError},
     ObjectProtocol, PyBytes, PyClone, PyDict, PyErr, PyInt, PyModule,
     PyObject, PyResult, PyString, PyTuple, Python, PythonObject, ToPyObject,
 };
 use hg::{
     nodemap::{Block, NodeMapError, NodeTree},
     revlog::{nodemap::NodeMap, NodePrefix, RevlogIndex},
     Revision,
 };
 use std::cell::RefCell;
 
 /// Return a Struct implementing the Graph trait
 pub(crate) fn pyindex_to_graph(
     py: Python,
     index: PyObject,
 ) -> PyResult<cindex::Index> {
     match index.extract::<MixedIndex>(py) {
         Ok(midx) => Ok(midx.clone_cindex(py)),
         Err(_) => cindex::Index::new(py, index),
     }
 }
 
 py_class!(pub class MixedIndex |py| {
     data cindex: RefCell<cindex::Index>;
     data nt: RefCell<Option<NodeTree>>;
     data docket: RefCell<Option<PyObject>>;
     // Holds a reference to the mmap'ed persistent nodemap data
     data mmap: RefCell<Option<PyBuffer>>;
 
     def __new__(_cls, cindex: PyObject) -> PyResult<MixedIndex> {
         Self::new(py, cindex)
     }
 
     /// Compatibility layer used for Python consumers needing access to the C index
     ///
     /// Only use case so far is `scmutil.shortesthexnodeidprefix`,
     /// that may need to build a custom `nodetree`, based on a specified revset.
     /// With a Rust implementation of the nodemap, we will be able to get rid of
     /// this, by exposing our own standalone nodemap class,
     /// ready to accept `MixedIndex`.
     def get_cindex(&self) -> PyResult<PyObject> {
         Ok(self.cindex(py).borrow().inner().clone_ref(py))
     }
 
     // Index API involving nodemap, as defined in mercurial/pure/parsers.py
 
     /// Return Revision if found, raises a bare `error.RevlogError`
     /// in case of ambiguity, same as C version does
-    def get_rev(&self, pynode: PyBytes) -> PyResult<Option<Revision>> {
+    def get_rev(&self, node: PyBytes) -> PyResult<Option<Revision>> {
         let opt = self.get_nodetree(py)?.borrow();
         let nt = opt.as_ref().unwrap();
         let idx = &*self.cindex(py).borrow();
-        let node = node_from_py_bytes(py, &pynode)?;
-        match nt.find_bin(idx, node.into())
-        {
-            Ok(None) =>
-                // fallback to C implementation, remove once
-                // https://bz.mercurial-scm.org/show_bug.cgi?id=6554
-                // is fixed (a simple backout should do)
-                self.call_cindex(py, "get_rev", &PyTuple::new(py, &[pynode.into_object()]), None)?
-                .extract(py),
-            Ok(Some(rev)) => Ok(Some(rev)),
-            Err(e) => Err(nodemap_error(py, e)),
-        }
+        let node = node_from_py_bytes(py, &node)?;
+        nt.find_bin(idx, node.into()).map_err(|e| nodemap_error(py, e))
     }
 
     /// same as `get_rev()` but raises a bare `error.RevlogError` if node
     /// is not found.
     ///
     /// No need to repeat `node` in the exception, `mercurial/revlog.py`
     /// will catch and rewrap with it
     def rev(&self, node: PyBytes) -> PyResult<Revision> {
         self.get_rev(py, node)?.ok_or_else(|| revlog_error(py))
     }
 
     /// return True if the node exist in the index
     def has_node(&self, node: PyBytes) -> PyResult<bool> {
         self.get_rev(py, node).map(|opt| opt.is_some())
     }
 
     /// find length of shortest hex nodeid of a binary ID
     def shortest(&self, node: PyBytes) -> PyResult<usize> {
         let opt = self.get_nodetree(py)?.borrow();
         let nt = opt.as_ref().unwrap();
         let idx = &*self.cindex(py).borrow();
         match nt.unique_prefix_len_node(idx, &node_from_py_bytes(py, &node)?)
         {
             Ok(Some(l)) => Ok(l),
             Ok(None) => Err(revlog_error(py)),
             Err(e) => Err(nodemap_error(py, e)),
         }
     }
 
-    def partialmatch(&self, pynode: PyObject) -> PyResult<Option<PyBytes>> {
+    def partialmatch(&self, node: PyObject) -> PyResult<Option<PyBytes>> {
         let opt = self.get_nodetree(py)?.borrow();
         let nt = opt.as_ref().unwrap();
         let idx = &*self.cindex(py).borrow();
 
         let node_as_string = if cfg!(feature = "python3-sys") {
-            pynode.cast_as::<PyString>(py)?.to_string(py)?.to_string()
+            node.cast_as::<PyString>(py)?.to_string(py)?.to_string()
         }
         else {
-            let node = pynode.extract::<PyBytes>(py)?;
+            let node = node.extract::<PyBytes>(py)?;
             String::from_utf8_lossy(node.data(py)).to_string()
         };
 
         let prefix = NodePrefix::from_hex(&node_as_string).map_err(|_| PyErr::new::<ValueError, _>(py, "Invalid node or prefix"))?;
 
-        match nt.find_bin(idx, prefix) {
-            Ok(None) =>
-                // fallback to C implementation, remove once
-                // https://bz.mercurial-scm.org/show_bug.cgi?id=6554
-                // is fixed (a simple backout should do)
-                self.call_cindex(
-                    py, "partialmatch",
-                    &PyTuple::new(py, &[pynode]), None
-                )?.extract(py),
-            Ok(Some(rev)) =>
-                Ok(Some(PyBytes::new(py, idx.node(rev).unwrap().as_bytes()))),
-            Err(e) => Err(nodemap_error(py, e)),
-        }
+        nt.find_bin(idx, prefix)
+            // TODO make an inner API returning the node directly
+            .map(|opt| opt.map(
+                |rev| PyBytes::new(py, idx.node(rev).unwrap().as_bytes())))
+            .map_err(|e| nodemap_error(py, e))
+
     }
 
     /// append an index entry
     def append(&self, tup: PyTuple) -> PyResult<PyObject> {
         if tup.len(py) < 8 {
             // this is better than the panic promised by tup.get_item()
             return Err(
                 PyErr::new::<IndexError, _>(py, "tuple index out of range"))
         }
         let node_bytes = tup.get_item(py, 7).extract(py)?;
         let node = node_from_py_object(py, &node_bytes)?;
 
         let mut idx = self.cindex(py).borrow_mut();
         let rev = idx.len() as Revision;
 
         idx.append(py, tup)?;
         self.get_nodetree(py)?.borrow_mut().as_mut().unwrap()
             .insert(&*idx, &node, rev)
             .map_err(|e| nodemap_error(py, e))?;
         Ok(py.None())
     }
 
     def __delitem__(&self, key: PyObject) -> PyResult<()> {
         // __delitem__ is both for `del idx[r]` and `del idx[r1:r2]`
         self.cindex(py).borrow().inner().del_item(py, key)?;
         let mut opt = self.get_nodetree(py)?.borrow_mut();
         let mut nt = opt.as_mut().unwrap();
         nt.invalidate_all();
         self.fill_nodemap(py, &mut nt)?;
         Ok(())
     }
 
     //
     // Reforwarded C index API
     //
 
     // index_methods (tp_methods). Same ordering as in revlog.c
 
     /// return the gca set of the given revs
     def ancestors(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "ancestors", args, kw)
     }
 
     /// return the heads of the common ancestors of the given revs
     def commonancestorsheads(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "commonancestorsheads", args, kw)
     }
 
     /// Clear the index caches and inner py_class data.
     /// It is Python's responsibility to call `update_nodemap_data` again.
     def clearcaches(&self, *args, **kw) -> PyResult<PyObject> {
         self.nt(py).borrow_mut().take();
         self.docket(py).borrow_mut().take();
         self.mmap(py).borrow_mut().take();
         self.call_cindex(py, "clearcaches", args, kw)
     }
 
     /// return the raw binary string representing a revision
     def entry_binary(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "entry_binary", args, kw)
     }
 
     /// return a binary packed version of the header
     def pack_header(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "pack_header", args, kw)
     }
 
     /// get an index entry
     def get(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "get", args, kw)
     }
 
     /// compute phases
     def computephasesmapsets(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "computephasesmapsets", args, kw)
     }
 
     /// reachableroots
     def reachableroots2(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "reachableroots2", args, kw)
     }
 
     /// get head revisions
     def headrevs(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "headrevs", args, kw)
     }
 
     /// get filtered head revisions
     def headrevsfiltered(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "headrevsfiltered", args, kw)
     }
 
     /// True if the object is a snapshot
     def issnapshot(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "issnapshot", args, kw)
     }
 
     /// Gather snapshot data in a cache dict
     def findsnapshots(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "findsnapshots", args, kw)
     }
 
     /// determine revisions with deltas to reconstruct fulltext
     def deltachain(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "deltachain", args, kw)
     }
 
     /// slice planned chunk read to reach a density threshold
     def slicechunktodensity(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "slicechunktodensity", args, kw)
     }
 
     /// stats for the index
     def stats(&self, *args, **kw) -> PyResult<PyObject> {
         self.call_cindex(py, "stats", args, kw)
     }
 
     // index_sequence_methods and index_mapping_methods.
     //
     // Since we call back through the high level Python API,
     // there's no point making a distinction between index_get
     // and index_getitem.
 
     def __len__(&self) -> PyResult<usize> {
         self.cindex(py).borrow().inner().len(py)
     }
 
     def __getitem__(&self, key: PyObject) -> PyResult<PyObject> {
         // this conversion seems needless, but that's actually because
         // `index_getitem` does not handle conversion from PyLong,
         // which expressions such as [e for e in index] internally use.
         // Note that we don't seem to have a direct way to call
         // PySequence_GetItem (does the job), which would possibly be better
         // for performance
         let key = match key.extract::<Revision>(py) {
             Ok(rev) => rev.to_py_object(py).into_object(),
             Err(_) => key,
         };
         self.cindex(py).borrow().inner().get_item(py, key)
     }
 
     def __setitem__(&self, key: PyObject, value: PyObject) -> PyResult<()> {
         self.cindex(py).borrow().inner().set_item(py, key, value)
     }
 
     def __contains__(&self, item: PyObject) -> PyResult<bool> {
         // ObjectProtocol does not seem to provide contains(), so
         // this is an equivalent implementation of the index_contains()
         // defined in revlog.c
         let cindex = self.cindex(py).borrow();
         match item.extract::<Revision>(py) {
             Ok(rev) => {
                 Ok(rev >= -1 && rev < cindex.inner().len(py)? as Revision)
             }
             Err(_) => {
                 cindex.inner().call_method(
                     py,
                     "has_node",
                     PyTuple::new(py, &[item]),
                     None)?
                 .extract(py)
             }
         }
     }
 
     def nodemap_data_all(&self) -> PyResult<PyBytes> {
         self.inner_nodemap_data_all(py)
     }
 
     def nodemap_data_incremental(&self) -> PyResult<PyObject> {
         self.inner_nodemap_data_incremental(py)
     }
     def update_nodemap_data(
         &self,
         docket: PyObject,
         nm_data: PyObject
     ) -> PyResult<PyObject> {
         self.inner_update_nodemap_data(py, docket, nm_data)
     }
 
     @property
     def entry_size(&self) -> PyResult<PyInt> {
         self.cindex(py).borrow().inner().getattr(py, "entry_size")?.extract::<PyInt>(py)
     }
 
     @property
     def rust_ext_compat(&self) -> PyResult<PyInt> {
         self.cindex(py).borrow().inner().getattr(py, "rust_ext_compat")?.extract::<PyInt>(py)
     }
 
 });
 
 impl MixedIndex {
     fn new(py: Python, cindex: PyObject) -> PyResult<MixedIndex> {
         Self::create_instance(
             py,
             RefCell::new(cindex::Index::new(py, cindex)?),
             RefCell::new(None),
             RefCell::new(None),
             RefCell::new(None),
         )
     }
 
     /// This is scaffolding at this point, but it could also become
     /// a way to start a persistent nodemap or perform a
     /// vacuum / repack operation
     fn fill_nodemap(
         &self,
         py: Python,
         nt: &mut NodeTree,
     ) -> PyResult<PyObject> {
         let index = self.cindex(py).borrow();
         for r in 0..index.len() {
             let rev = r as Revision;
             // in this case node() won't ever return None
             nt.insert(&*index, index.node(rev).unwrap(), rev)
                 .map_err(|e| nodemap_error(py, e))?
         }
         Ok(py.None())
     }
 
     fn get_nodetree<'a>(
         &'a self,
         py: Python<'a>,
     ) -> PyResult<&'a RefCell<Option<NodeTree>>> {
         if self.nt(py).borrow().is_none() {
             let readonly = Box::new(Vec::new());
             let mut nt = NodeTree::load_bytes(readonly, 0);
             self.fill_nodemap(py, &mut nt)?;
             self.nt(py).borrow_mut().replace(nt);
         }
         Ok(self.nt(py))
     }
 
     /// forward a method call to the underlying C index
     fn call_cindex(
         &self,
         py: Python,
         name: &str,
         args: &PyTuple,
         kwargs: Option<&PyDict>,
     ) -> PyResult<PyObject> {
         self.cindex(py)
             .borrow()
             .inner()
             .call_method(py, name, args, kwargs)
     }
 
     pub fn clone_cindex(&self, py: Python) -> cindex::Index {
         self.cindex(py).borrow().clone_ref(py)
     }
 
     /// Returns the full nodemap bytes to be written as-is to disk
     fn inner_nodemap_data_all(&self, py: Python) -> PyResult<PyBytes> {
         let nodemap = self.get_nodetree(py)?.borrow_mut().take().unwrap();
         let (readonly, bytes) = nodemap.into_readonly_and_added_bytes();
 
         // If there's anything readonly, we need to build the data again from
         // scratch
         let bytes = if readonly.len() > 0 {
             let mut nt = NodeTree::load_bytes(Box::new(vec![]), 0);
             self.fill_nodemap(py, &mut nt)?;
 
             let (readonly, bytes) = nt.into_readonly_and_added_bytes();
             assert_eq!(readonly.len(), 0);
 
             bytes
         } else {
             bytes
         };
 
         let bytes = PyBytes::new(py, &bytes);
         Ok(bytes)
     }
 
     /// Returns the last saved docket along with the size of any changed data
     /// (in number of blocks), and said data as bytes.
     fn inner_nodemap_data_incremental(
         &self,
         py: Python,
     ) -> PyResult<PyObject> {
         let docket = self.docket(py).borrow();
         let docket = match docket.as_ref() {
             Some(d) => d,
             None => return Ok(py.None()),
         };
 
         let node_tree = self.get_nodetree(py)?.borrow_mut().take().unwrap();
         let masked_blocks = node_tree.masked_readonly_blocks();
         let (_, data) = node_tree.into_readonly_and_added_bytes();
         let changed = masked_blocks * std::mem::size_of::<Block>();
 
         Ok((docket, changed, PyBytes::new(py, &data))
             .to_py_object(py)
             .into_object())
     }
 
     /// Update the nodemap from the new (mmaped) data.
     /// The docket is kept as a reference for later incremental calls.
     fn inner_update_nodemap_data(
         &self,
         py: Python,
         docket: PyObject,
         nm_data: PyObject,
     ) -> PyResult<PyObject> {
         let buf = PyBuffer::get(py, &nm_data)?;
         let len = buf.item_count();
 
         // Build a slice from the mmap'ed buffer data
         let cbuf = buf.buf_ptr();
         let bytes = if std::mem::size_of::<u8>() == buf.item_size()
             && buf.is_c_contiguous()
             && u8::is_compatible_format(buf.format())
         {
             unsafe { std::slice::from_raw_parts(cbuf as *const u8, len) }
         } else {
             return Err(PyErr::new::<ValueError, _>(
                 py,
                 "Nodemap data buffer has an invalid memory representation"
                     .to_string(),
             ));
         };
 
         // Keep a reference to the mmap'ed buffer, otherwise we get a dangling
         // pointer.
         self.mmap(py).borrow_mut().replace(buf);
 
         let mut nt = NodeTree::load_bytes(Box::new(bytes), len);
 
         let data_tip =
             docket.getattr(py, "tip_rev")?.extract::<Revision>(py)?;
         self.docket(py).borrow_mut().replace(docket.clone_ref(py));
         let idx = self.cindex(py).borrow();
         let current_tip = idx.len();
 
         for r in (data_tip + 1)..current_tip as Revision {
             let rev = r as Revision;
             // in this case node() won't ever return None
             nt.insert(&*idx, idx.node(rev).unwrap(), rev)
                 .map_err(|e| nodemap_error(py, e))?
         }
 
         *self.nt(py).borrow_mut() = Some(nt);
 
         Ok(py.None())
     }
 }
 
 fn revlog_error(py: Python) -> PyErr {
     match py
         .import("mercurial.error")
         .and_then(|m| m.get(py, "RevlogError"))
     {
         Err(e) => e,
         Ok(cls) => PyErr::from_instance(
             py,
             cls.call(py, (py.None(),), None).ok().into_py_object(py),
         ),
     }
 }
 
 fn rev_not_in_index(py: Python, rev: Revision) -> PyErr {
     PyErr::new::<ValueError, _>(
         py,
         format!(
             "Inconsistency: Revision {} found in nodemap \
              is not in revlog index",
             rev
         ),
     )
 }
 
 /// Standard treatment of NodeMapError
 fn nodemap_error(py: Python, err: NodeMapError) -> PyErr {
     match err {
         NodeMapError::MultipleResults => revlog_error(py),
         NodeMapError::RevisionNotInIndex(r) => rev_not_in_index(py, r),
     }
 }
 
 /// Create the module, with __package__ given from parent
 pub fn init_module(py: Python, package: &str) -> PyResult<PyModule> {
     let dotted_name = &format!("{}.revlog", package);
     let m = PyModule::new(py, dotted_name)?;
     m.add(py, "__package__", package)?;
     m.add(py, "__doc__", "RevLog - Rust implementations")?;
 
     m.add_class::<MixedIndex>(py)?;
 
     let sys = PyModule::import(py, "sys")?;
     let sys_modules: PyDict = sys.get(py, "modules")?.extract(py)?;
     sys_modules.set_item(py, dotted_name, &m)?;
 
     Ok(m)
 }

rust/hgcli/pyoxidizer.bzl

@@ -1,336 +1,336 b''
 # The following variables can be passed in as parameters:
 #
 # VERSION
 #   Version string of program being produced.
 #
 # MSI_NAME
 #   Root name of MSI installer.
 #
 # EXTRA_MSI_FEATURES
 #   ; delimited string of extra features to advertise in the built MSA.
 #
 # SIGNING_PFX_PATH
 #   Path to code signing certificate to use.
 #
 # SIGNING_PFX_PASSWORD
 #   Password to code signing PFX file defined by SIGNING_PFX_PATH.
 #
 # SIGNING_SUBJECT_NAME
 #   String fragment in code signing certificate subject name used to find
 #   code signing certificate in Windows certificate store.
 #
 # TIME_STAMP_SERVER_URL
 #   URL of time-stamp token authority (RFC 3161) servers to stamp code signatures.
 
 ROOT = CWD + "/../.."
 
 VERSION = VARS.get("VERSION", "0.0")
 MSI_NAME = VARS.get("MSI_NAME", "mercurial")
 EXTRA_MSI_FEATURES = VARS.get("EXTRA_MSI_FEATURES")
 SIGNING_PFX_PATH = VARS.get("SIGNING_PFX_PATH")
 SIGNING_PFX_PASSWORD = VARS.get("SIGNING_PFX_PASSWORD", "")
 SIGNING_SUBJECT_NAME = VARS.get("SIGNING_SUBJECT_NAME")
 TIME_STAMP_SERVER_URL = VARS.get("TIME_STAMP_SERVER_URL", "http://timestamp.digicert.com")
 
 IS_WINDOWS = "windows" in BUILD_TARGET_TRIPLE
 
 USE_IN_MEMORY_RESOURCES = False
 
 # Code to run in Python interpreter.
 RUN_CODE = """
 import os
 import sys
 extra_path = os.environ.get('PYTHONPATH')
 if extra_path is not None:
     # extensions and hooks expect a working python environment
     # We do not prepend the values because the Mercurial library wants to be in
     # the front of the sys.path to avoid picking up other installations.
     sys.path.extend(extra_path.split(os.pathsep))
 # Add user site to sys.path to load extensions without the full path
 if os.name == 'nt':
     vi = sys.version_info
     appdata = os.environ.get('APPDATA')
     if appdata:
         sys.path.append(
             os.path.join(
                 appdata,
                 'Python',
                 'Python%d%d' % (vi[0], vi[1]),
                 'site-packages',
             )
         )
 elif sys.platform == "darwin":
     vi = sys.version_info
 
     def joinuser(*args):
         return os.path.expanduser(os.path.join(*args))
 
     # Note: site.py uses `sys._framework` instead of hardcoding "Python" as the
     #   3rd arg, but that is set to an empty string in an oxidized binary.  It
     #   has a fallback to ~/.local when `sys._framework` isn't set, but we want
     #   to match what the system python uses, so it sees pip installed stuff.
     usersite = joinuser("~", "Library", "Python",
                         "%d.%d" % vi[:2], "lib/python/site-packages")
 
     sys.path.append(usersite)
 import hgdemandimport;
 hgdemandimport.enable();
 from mercurial import dispatch;
 dispatch.run();
 """
 
 set_build_path(ROOT + "/build/pyoxidizer")
 
 def make_distribution():
     return default_python_distribution(python_version = "3.9")
 
 def resource_callback(policy, resource):
     if USE_IN_MEMORY_RESOURCES:
         resource.add_location = "in-memory"
         return
 
     # We use a custom resource routing policy to influence where things are loaded
     # from.
     #
     # For Python modules and resources, we load from memory if they are in
     # the standard library and from the filesystem if not. This is because
     # parts of Mercurial and some 3rd party packages aren't yet compatible
     # with memory loading.
     #
     # For Python extension modules, we load from the filesystem because
     # this yields greatest compatibility.
     if type(resource) in ("PythonModuleSource", "PythonPackageResource", "PythonPackageDistributionResource"):
         if resource.is_stdlib:
             resource.add_location = "in-memory"
         else:
             resource.add_location = "filesystem-relative:lib"
 
     elif type(resource) == "PythonExtensionModule":
         resource.add_location = "filesystem-relative:lib"
 
 def make_exe(dist):
     """Builds a Rust-wrapped Mercurial binary."""
     packaging_policy = dist.make_python_packaging_policy()
 
     # Extension may depend on any Python functionality. Include all
     # extensions.
     packaging_policy.extension_module_filter = "all"
     packaging_policy.resources_location = "in-memory"
     if not USE_IN_MEMORY_RESOURCES:
         packaging_policy.resources_location_fallback = "filesystem-relative:lib"
     packaging_policy.register_resource_callback(resource_callback)
 
     config = dist.make_python_interpreter_config()
     config.allocator_backend = "default"
     config.run_command = RUN_CODE
 
     # We want to let the user load extensions from the file system
     config.filesystem_importer = True
 
     # We need this to make resourceutil happy, since it looks for sys.frozen.
     config.sys_frozen = True
     config.legacy_windows_stdio = True
 
     exe = dist.to_python_executable(
         name = "hg",
         packaging_policy = packaging_policy,
         config = config,
     )
 
     # Add Mercurial to resources.
-    exe.add_python_resources(exe.pip_install(["--verbose", ROOT]))
+    exe.add_python_resources(exe.pip_install(["--verbose", "--no-use-pep517", ROOT]))
 
     # On Windows, we install extra packages for convenience.
     if IS_WINDOWS:
         exe.add_python_resources(
             exe.pip_install(["-r", ROOT + "/contrib/packaging/requirements-windows-py3.txt"]),
         )
     extra_packages = VARS.get("extra_py_packages", "")
     if extra_packages:
         for extra in extra_packages.split(","):
             extra_src, pkgs = extra.split("=")
             pkgs = pkgs.split(":")
             exe.add_python_resources(exe.read_package_root(extra_src, pkgs))
 
     return exe
 
 def make_manifest(dist, exe):
     m = FileManifest()
     m.add_python_resource(".", exe)
 
     return m
 
 
 # This adjusts the InstallManifest produced from exe generation to provide
 # additional files found in a Windows install layout.
 def make_windows_install_layout(manifest):
     # Copy various files to new install locations. This can go away once
     # we're using the importlib resource reader.
     RECURSIVE_COPIES = {
         "lib/mercurial/locale/": "locale/",
         "lib/mercurial/templates/": "templates/",
     }
     for (search, replace) in RECURSIVE_COPIES.items():
         for path in manifest.paths():
             if path.startswith(search):
                 new_path = path.replace(search, replace)
                 print("copy %s to %s" % (path, new_path))
                 file = manifest.get_file(path)
                 manifest.add_file(file, path = new_path)
 
     # Similar to above, but with filename pattern matching.
     # lib/mercurial/helptext/**/*.txt -> helptext/
     # lib/mercurial/defaultrc/*.rc -> defaultrc/
     for path in manifest.paths():
         if path.startswith("lib/mercurial/helptext/") and path.endswith(".txt"):
             new_path = path[len("lib/mercurial/"):]
         elif path.startswith("lib/mercurial/defaultrc/") and path.endswith(".rc"):
             new_path = path[len("lib/mercurial/"):]
         else:
             continue
 
         print("copying %s to %s" % (path, new_path))
         manifest.add_file(manifest.get_file(path), path = new_path)
 
     extra_install_files = VARS.get("extra_install_files", "")
     if extra_install_files:
         for extra in extra_install_files.split(","):
             print("adding extra files from %s" % extra)
             # TODO: I expected a ** glob to work, but it didn't.
             #
             # TODO: I know this has forward-slash paths. As far as I can tell,
             # backslashes don't ever match glob() expansions in 
             # tugger-starlark, even on Windows.
             manifest.add_manifest(glob(include=[extra + "/*/*"], strip_prefix=extra+"/"))
 
     # We also install a handful of additional files.
     EXTRA_CONTRIB_FILES = [
         "bash_completion",
         "hgweb.fcgi",
         "hgweb.wsgi",
         "logo-droplets.svg",
         "mercurial.el",
         "mq.el",
         "tcsh_completion",
         "tcsh_completion_build.sh",
         "xml.rnc",
         "zsh_completion",
     ]
 
     for f in EXTRA_CONTRIB_FILES:
         manifest.add_file(FileContent(path = ROOT + "/contrib/" + f), directory = "contrib")
 
     # Individual files with full source to destination path mapping.
     EXTRA_FILES = {
         "contrib/hgk": "contrib/hgk.tcl",
         "contrib/win32/postinstall.txt": "ReleaseNotes.txt",
         "contrib/win32/ReadMe.html": "ReadMe.html",
         "doc/style.css": "doc/style.css",
         "COPYING": "Copying.txt",
     }
 
     for source, dest in EXTRA_FILES.items():
         print("adding extra file %s" % dest)
         manifest.add_file(FileContent(path = ROOT + "/" + source), path = dest)
 
     # And finally some wildcard matches.
     manifest.add_manifest(glob(
         include = [ROOT + "/contrib/vim/*"],
         strip_prefix = ROOT + "/"
     ))
     manifest.add_manifest(glob(
         include = [ROOT + "/doc/*.html"],
         strip_prefix = ROOT + "/"
     ))
 
     # But we don't ship hg-ssh on Windows, so exclude its documentation.
     manifest.remove("doc/hg-ssh.8.html")
 
     return manifest
 
 
 def make_msi(manifest):
     manifest = make_windows_install_layout(manifest)
 
     if "x86_64" in BUILD_TARGET_TRIPLE:
         platform = "x64"
     else:
         platform = "x86"
 
     manifest.add_file(
         FileContent(path = ROOT + "/contrib/packaging/wix/COPYING.rtf"),
         path = "COPYING.rtf",
     )
     manifest.remove("Copying.txt")
     manifest.add_file(
         FileContent(path = ROOT + "/contrib/win32/mercurial.ini"),
         path = "defaultrc/mercurial.rc",
     )
     manifest.add_file(
         FileContent(filename = "editor.rc", content = "[ui]\neditor = notepad\n"),
         path = "defaultrc/editor.rc",
     )
 
     wix = WiXInstaller(
         "hg",
         "%s-%s-%s.msi" % (MSI_NAME, VERSION, platform),
         arch = platform,
     )
 
     # Materialize files in the manifest to the install layout.
     wix.add_install_files(manifest)
 
     # From mercurial.wxs.
     wix.install_files_root_directory_id = "INSTALLDIR"
 
     # Pull in our custom .wxs files.
     defines = {
         "PyOxidizer": "1",
         "Platform": platform,
         "Version": VERSION,
         "Comments": "Installs Mercurial version %s" % VERSION,
         "PythonVersion": "3",
         "MercurialHasLib": "1",
     }
 
     if EXTRA_MSI_FEATURES:
         defines["MercurialExtraFeatures"] = EXTRA_MSI_FEATURES
 
     wix.add_wxs_file(
         ROOT + "/contrib/packaging/wix/mercurial.wxs",
         preprocessor_parameters=defines,
     )
 
     # Our .wxs references to other files. Pull those into the build environment.
     for f in ("defines.wxi", "guids.wxi", "COPYING.rtf"):
         wix.add_build_file(f, ROOT + "/contrib/packaging/wix/" + f)
 
     wix.add_build_file("mercurial.ico", ROOT + "/contrib/win32/mercurial.ico")
 
     return wix
 
 
 def register_code_signers():
     if not IS_WINDOWS:
         return
 
     if SIGNING_PFX_PATH:
         signer = code_signer_from_pfx_file(SIGNING_PFX_PATH, SIGNING_PFX_PASSWORD)
     elif SIGNING_SUBJECT_NAME:
         signer = code_signer_from_windows_store_subject(SIGNING_SUBJECT_NAME)
     else:
         signer = None
 
     if signer:
         signer.set_time_stamp_server(TIME_STAMP_SERVER_URL)
         signer.activate()
 
 
 register_code_signers()
 
 register_target("distribution", make_distribution)
 register_target("exe", make_exe, depends = ["distribution"])
 register_target("app", make_manifest, depends = ["distribution", "exe"], default = True)
 register_target("msi", make_msi, depends = ["app"])
 
 resolve_targets()

tests/basic_test_result.py

@@ -1,52 +1,58 b''
 from __future__ import absolute_import, print_function
 
+import sys
 import unittest
 
+if sys.version_info[0] < 3:
+    base_class = unittest._TextTestResult
+else:
+    base_class = unittest.TextTestResult
 
-class TestResult(unittest._TextTestResult):
+
+class TestResult(base_class):
     def __init__(self, options, *args, **kwargs):
         super(TestResult, self).__init__(*args, **kwargs)
         self._options = options
 
         # unittest.TestResult didn't have skipped until 2.7. We need to
         # polyfill it.
         self.skipped = []
 
         # We have a custom "ignored" result that isn't present in any Python
         # unittest implementation. It is very similar to skipped. It may make
         # sense to map it into skip some day.
         self.ignored = []
 
         self.times = []
         self._firststarttime = None
         # Data stored for the benefit of generating xunit reports.
         self.successes = []
         self.faildata = {}
 
     def addFailure(self, test, reason):
         print("FAILURE!", test, reason)
 
     def addSuccess(self, test):
         print("SUCCESS!", test)
 
     def addError(self, test, err):
         print("ERR!", test, err)
 
     # Polyfill.
     def addSkip(self, test, reason):
         print("SKIP!", test, reason)
 
     def addIgnore(self, test, reason):
         print("IGNORE!", test, reason)
 
     def onStart(self, test):
         print("ON_START!", test)
 
     def onEnd(self):
         print("ON_END!")
 
     def addOutputMismatch(self, test, ret, got, expected):
         return False
 
     def stopTest(self, test, interrupted=False):
         super(TestResult, self).stopTest(test)

tests/run-tests.py

@@ -1,4069 +1,4072 b''
 #!/usr/bin/env python3
 #
 # run-tests.py - Run a set of tests on Mercurial
 #
 # Copyright 2006 Olivia Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 # Modifying this script is tricky because it has many modes:
 #   - serial (default) vs parallel (-jN, N > 1)
 #   - no coverage (default) vs coverage (-c, -C, -s)
 #   - temp install (default) vs specific hg script (--with-hg, --local)
 #   - tests are a mix of shell scripts and Python scripts
 #
 # If you change this script, it is recommended that you ensure you
 # haven't broken it by running it in various modes with a representative
 # sample of test scripts.  For example:
 #
 #  1) serial, no coverage, temp install:
 #      ./run-tests.py test-s*
 #  2) serial, no coverage, local hg:
 #      ./run-tests.py --local test-s*
 #  3) serial, coverage, temp install:
 #      ./run-tests.py -c test-s*
 #  4) serial, coverage, local hg:
 #      ./run-tests.py -c --local test-s*      # unsupported
 #  5) parallel, no coverage, temp install:
 #      ./run-tests.py -j2 test-s*
 #  6) parallel, no coverage, local hg:
 #      ./run-tests.py -j2 --local test-s*
 #  7) parallel, coverage, temp install:
 #      ./run-tests.py -j2 -c test-s*          # currently broken
 #  8) parallel, coverage, local install:
 #      ./run-tests.py -j2 -c --local test-s*  # unsupported (and broken)
 #  9) parallel, custom tmp dir:
 #      ./run-tests.py -j2 --tmpdir /tmp/myhgtests
 #  10) parallel, pure, tests that call run-tests:
 #      ./run-tests.py --pure `grep -l run-tests.py *.t`
 #
 # (You could use any subset of the tests: test-s* happens to match
 # enough that it's worth doing parallel runs, few enough that it
 # completes fairly quickly, includes both shell and Python scripts, and
 # includes some scripts that run daemon processes.)
 
 from __future__ import absolute_import, print_function
 
 import argparse
 import collections
 import contextlib
 import difflib
 import distutils.version as version
 import errno
 import json
 import multiprocessing
 import os
 import platform
 import random
 import re
 import shutil
 import signal
 import socket
 import subprocess
 import sys
 import sysconfig
 import tempfile
 import threading
 import time
 import unittest
 import uuid
 import xml.dom.minidom as minidom
 
 WINDOWS = os.name == r'nt'
 
 try:
     import Queue as queue
 except ImportError:
     import queue
 
 try:
     import shlex
 
     shellquote = shlex.quote
 except (ImportError, AttributeError):
     import pipes
 
     shellquote = pipes.quote
 
 
 processlock = threading.Lock()
 
 pygmentspresent = False
 try:  # is pygments installed
     import pygments
     import pygments.lexers as lexers
     import pygments.lexer as lexer
     import pygments.formatters as formatters
     import pygments.token as token
     import pygments.style as style
 
     if WINDOWS:
         hgpath = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
         sys.path.append(hgpath)
         try:
             from mercurial import win32  # pytype: disable=import-error
 
             # Don't check the result code because it fails on heptapod, but
             # something is able to convert to color anyway.
             win32.enablevtmode()
         finally:
             sys.path = sys.path[:-1]
 
     pygmentspresent = True
     difflexer = lexers.DiffLexer()
     terminal256formatter = formatters.Terminal256Formatter()
 except ImportError:
     pass
 
 if pygmentspresent:
 
     class TestRunnerStyle(style.Style):
         default_style = ""
         skipped = token.string_to_tokentype("Token.Generic.Skipped")
         failed = token.string_to_tokentype("Token.Generic.Failed")
         skippedname = token.string_to_tokentype("Token.Generic.SName")
         failedname = token.string_to_tokentype("Token.Generic.FName")
         styles = {
             skipped: '#e5e5e5',
             skippedname: '#00ffff',
             failed: '#7f0000',
             failedname: '#ff0000',
         }
 
     class TestRunnerLexer(lexer.RegexLexer):
         testpattern = r'[\w-]+\.(t|py)(#[a-zA-Z0-9_\-\.]+)?'
         tokens = {
             'root': [
                 (r'^Skipped', token.Generic.Skipped, 'skipped'),
                 (r'^Failed ', token.Generic.Failed, 'failed'),
                 (r'^ERROR: ', token.Generic.Failed, 'failed'),
             ],
             'skipped': [
                 (testpattern, token.Generic.SName),
                 (r':.*', token.Generic.Skipped),
             ],
             'failed': [
                 (testpattern, token.Generic.FName),
                 (r'(:| ).*', token.Generic.Failed),
             ],
         }
 
     runnerformatter = formatters.Terminal256Formatter(style=TestRunnerStyle)
     runnerlexer = TestRunnerLexer()
 
 origenviron = os.environ.copy()
 
 
 if sys.version_info > (3, 5, 0):
     PYTHON3 = True
     xrange = range  # we use xrange in one place, and we'd rather not use range
 
     def _sys2bytes(p):
         if p is None:
             return p
         return p.encode('utf-8')
 
     def _bytes2sys(p):
         if p is None:
             return p
         return p.decode('utf-8')
 
     osenvironb = getattr(os, 'environb', None)
     if osenvironb is None:
         # Windows lacks os.environb, for instance.  A proxy over the real thing
         # instead of a copy allows the environment to be updated via bytes on
         # all platforms.
         class environbytes(object):
             def __init__(self, strenv):
                 self.__len__ = strenv.__len__
                 self.clear = strenv.clear
                 self._strenv = strenv
 
             def __getitem__(self, k):
                 v = self._strenv.__getitem__(_bytes2sys(k))
                 return _sys2bytes(v)
 
             def __setitem__(self, k, v):
                 self._strenv.__setitem__(_bytes2sys(k), _bytes2sys(v))
 
             def __delitem__(self, k):
                 self._strenv.__delitem__(_bytes2sys(k))
 
             def __contains__(self, k):
                 return self._strenv.__contains__(_bytes2sys(k))
 
             def __iter__(self):
                 return iter([_sys2bytes(k) for k in iter(self._strenv)])
 
             def get(self, k, default=None):
                 v = self._strenv.get(_bytes2sys(k), _bytes2sys(default))
                 return _sys2bytes(v)
 
             def pop(self, k, default=None):
                 v = self._strenv.pop(_bytes2sys(k), _bytes2sys(default))
                 return _sys2bytes(v)
 
         osenvironb = environbytes(os.environ)
 
     getcwdb = getattr(os, 'getcwdb')
     if not getcwdb or WINDOWS:
         getcwdb = lambda: _sys2bytes(os.getcwd())
 
 elif sys.version_info >= (3, 0, 0):
     print(
         '%s is only supported on Python 3.5+ and 2.7, not %s'
         % (sys.argv[0], '.'.join(str(v) for v in sys.version_info[:3]))
     )
     sys.exit(70)  # EX_SOFTWARE from `man 3 sysexit`
 else:
     PYTHON3 = False
 
     # In python 2.x, path operations are generally done using
     # bytestrings by default, so we don't have to do any extra
     # fiddling there. We define the wrapper functions anyway just to
     # help keep code consistent between platforms.
     def _sys2bytes(p):
         return p
 
     _bytes2sys = _sys2bytes
     osenvironb = os.environ
     getcwdb = os.getcwd
 
 if WINDOWS:
     _getcwdb = getcwdb
 
     def getcwdb():
         cwd = _getcwdb()
         if re.match(b'^[a-z]:', cwd):
             # os.getcwd() is inconsistent on the capitalization of the drive
             # letter, so adjust it. see https://bugs.python.org/issue40368
             cwd = cwd[0:1].upper() + cwd[1:]
         return cwd
 
 
 # For Windows support
 wifexited = getattr(os, "WIFEXITED", lambda x: False)
 
 # Whether to use IPv6
 def checksocketfamily(name, port=20058):
     """return true if we can listen on localhost using family=name
 
     name should be either 'AF_INET', or 'AF_INET6'.
     port being used is okay - EADDRINUSE is considered as successful.
     """
     family = getattr(socket, name, None)
     if family is None:
         return False
     try:
         s = socket.socket(family, socket.SOCK_STREAM)
         s.bind(('localhost', port))
         s.close()
         return True
     except socket.error as exc:
         if exc.errno == errno.EADDRINUSE:
             return True
         elif exc.errno in (errno.EADDRNOTAVAIL, errno.EPROTONOSUPPORT):
             return False
         else:
             raise
     else:
         return False
 
 
 # useipv6 will be set by parseargs
 useipv6 = None
 
 
 def checkportisavailable(port):
     """return true if a port seems free to bind on localhost"""
     if useipv6:
         family = socket.AF_INET6
     else:
         family = socket.AF_INET
     try:
         with contextlib.closing(socket.socket(family, socket.SOCK_STREAM)) as s:
             s.bind(('localhost', port))
         return True
     except socket.error as exc:
         if WINDOWS and exc.errno == errno.WSAEACCES:
             return False
         elif PYTHON3:
             # TODO: make a proper exception handler after dropping py2.  This
             #       works because socket.error is an alias for OSError on py3,
             #       which is also the baseclass of PermissionError.
             if isinstance(exc, PermissionError):
                 return False
         if exc.errno not in (
             errno.EADDRINUSE,
             errno.EADDRNOTAVAIL,
             errno.EPROTONOSUPPORT,
         ):
             raise
     return False
 
 
 closefds = os.name == 'posix'
 
 
 def Popen4(cmd, wd, timeout, env=None):
     processlock.acquire()
     p = subprocess.Popen(
         _bytes2sys(cmd),
         shell=True,
         bufsize=-1,
         cwd=_bytes2sys(wd),
         env=env,
         close_fds=closefds,
         stdin=subprocess.PIPE,
         stdout=subprocess.PIPE,
         stderr=subprocess.STDOUT,
     )
     processlock.release()
 
     p.fromchild = p.stdout
     p.tochild = p.stdin
     p.childerr = p.stderr
 
     p.timeout = False
     if timeout:
 
         def t():
             start = time.time()
             while time.time() - start < timeout and p.returncode is None:
                 time.sleep(0.1)
             p.timeout = True
             vlog('# Timout reached for process %d' % p.pid)
             if p.returncode is None:
                 terminate(p)
 
         threading.Thread(target=t).start()
 
     return p
 
 
 if sys.executable:
     sysexecutable = sys.executable
 elif os.environ.get('PYTHONEXECUTABLE'):
     sysexecutable = os.environ['PYTHONEXECUTABLE']
 elif os.environ.get('PYTHON'):
     sysexecutable = os.environ['PYTHON']
 else:
     raise AssertionError('Could not find Python interpreter')
 
 PYTHON = _sys2bytes(sysexecutable.replace('\\', '/'))
 IMPL_PATH = b'PYTHONPATH'
 if 'java' in sys.platform:
     IMPL_PATH = b'JYTHONPATH'
 
 default_defaults = {
     'jobs': ('HGTEST_JOBS', multiprocessing.cpu_count()),
     'timeout': ('HGTEST_TIMEOUT', 360),
     'slowtimeout': ('HGTEST_SLOWTIMEOUT', 1500),
     'port': ('HGTEST_PORT', 20059),
     'shell': ('HGTEST_SHELL', 'sh'),
 }
 
 defaults = default_defaults.copy()
 
 
 def canonpath(path):
     return os.path.realpath(os.path.expanduser(path))
 
 
 def which(exe):
     if PYTHON3:
         # shutil.which only accept bytes from 3.8
         cmd = _bytes2sys(exe)
         real_exec = shutil.which(cmd)
         return _sys2bytes(real_exec)
     else:
         # let us do the os work
         for p in osenvironb[b'PATH'].split(os.pathsep):
             f = os.path.join(p, exe)
             if os.path.isfile(f):
                 return f
         return None
 
 
 def parselistfiles(files, listtype, warn=True):
     entries = dict()
     for filename in files:
         try:
             path = os.path.expanduser(os.path.expandvars(filename))
             f = open(path, "rb")
         except IOError as err:
             if err.errno != errno.ENOENT:
                 raise
             if warn:
                 print("warning: no such %s file: %s" % (listtype, filename))
             continue
 
         for line in f.readlines():
             line = line.split(b'#', 1)[0].strip()
             if line:
                 # Ensure path entries are compatible with os.path.relpath()
                 entries[os.path.normpath(line)] = filename
 
         f.close()
     return entries
 
 
 def parsettestcases(path):
     """read a .t test file, return a set of test case names
 
     If path does not exist, return an empty set.
     """
     cases = []
     try:
         with open(path, 'rb') as f:
             for l in f:
                 if l.startswith(b'#testcases '):
                     cases.append(sorted(l[11:].split()))
     except IOError as ex:
         if ex.errno != errno.ENOENT:
             raise
     return cases
 
 
 def getparser():
     """Obtain the OptionParser used by the CLI."""
     parser = argparse.ArgumentParser(usage='%(prog)s [options] [tests]')
 
     selection = parser.add_argument_group('Test Selection')
     selection.add_argument(
         '--allow-slow-tests',
         action='store_true',
         help='allow extremely slow tests',
     )
     selection.add_argument(
         "--blacklist",
         action="append",
         help="skip tests listed in the specified blacklist file",
     )
     selection.add_argument(
         "--changed",
         help="run tests that are changed in parent rev or working directory",
     )
     selection.add_argument(
         "-k", "--keywords", help="run tests matching keywords"
     )
     selection.add_argument(
         "-r", "--retest", action="store_true", help="retest failed tests"
     )
     selection.add_argument(
         "--test-list",
         action="append",
         help="read tests to run from the specified file",
     )
     selection.add_argument(
         "--whitelist",
         action="append",
         help="always run tests listed in the specified whitelist file",
     )
     selection.add_argument(
         'tests', metavar='TESTS', nargs='*', help='Tests to run'
     )
 
     harness = parser.add_argument_group('Test Harness Behavior')
     harness.add_argument(
         '--bisect-repo',
         metavar='bisect_repo',
         help=(
             "Path of a repo to bisect. Use together with " "--known-good-rev"
         ),
     )
     harness.add_argument(
         "-d",
         "--debug",
         action="store_true",
         help="debug mode: write output of test scripts to console"
         " rather than capturing and diffing it (disables timeout)",
     )
     harness.add_argument(
         "-f",
         "--first",
         action="store_true",
         help="exit on the first test failure",
     )
     harness.add_argument(
         "-i",
         "--interactive",
         action="store_true",
         help="prompt to accept changed output",
     )
     harness.add_argument(
         "-j",
         "--jobs",
         type=int,
         help="number of jobs to run in parallel"
         " (default: $%s or %d)" % defaults['jobs'],
     )
     harness.add_argument(
         "--keep-tmpdir",
         action="store_true",
         help="keep temporary directory after running tests",
     )
     harness.add_argument(
         '--known-good-rev',
         metavar="known_good_rev",
         help=(
             "Automatically bisect any failures using this "
             "revision as a known-good revision."
         ),
     )
     harness.add_argument(
         "--list-tests",
         action="store_true",
         help="list tests instead of running them",
     )
     harness.add_argument(
         "--loop", action="store_true", help="loop tests repeatedly"
     )
     harness.add_argument(
         '--random', action="store_true", help='run tests in random order'
     )
     harness.add_argument(
         '--order-by-runtime',
         action="store_true",
         help='run slowest tests first, according to .testtimes',
     )
     harness.add_argument(
         "-p",
         "--port",
         type=int,
         help="port on which servers should listen"
         " (default: $%s or %d)" % defaults['port'],
     )
     harness.add_argument(
         '--profile-runner',
         action='store_true',
         help='run statprof on run-tests',
     )
     harness.add_argument(
         "-R", "--restart", action="store_true", help="restart at last error"
     )
     harness.add_argument(
         "--runs-per-test",
         type=int,
         dest="runs_per_test",
         help="run each test N times (default=1)",
         default=1,
     )
     harness.add_argument(
         "--shell", help="shell to use (default: $%s or %s)" % defaults['shell']
     )
     harness.add_argument(
         '--showchannels', action='store_true', help='show scheduling channels'
     )
     harness.add_argument(
         "--slowtimeout",
         type=int,
         help="kill errant slow tests after SLOWTIMEOUT seconds"
         " (default: $%s or %d)" % defaults['slowtimeout'],
     )
     harness.add_argument(
         "-t",
         "--timeout",
         type=int,
         help="kill errant tests after TIMEOUT seconds"
         " (default: $%s or %d)" % defaults['timeout'],
     )
     harness.add_argument(
         "--tmpdir",
         help="run tests in the given temporary directory"
         " (implies --keep-tmpdir)",
     )
     harness.add_argument(
         "-v", "--verbose", action="store_true", help="output verbose messages"
     )
 
     hgconf = parser.add_argument_group('Mercurial Configuration')
     hgconf.add_argument(
         "--chg",
         action="store_true",
         help="install and use chg wrapper in place of hg",
     )
     hgconf.add_argument(
         "--chg-debug",
         action="store_true",
         help="show chg debug logs",
     )
     hgconf.add_argument(
         "--rhg",
         action="store_true",
         help="install and use rhg Rust implementation in place of hg",
     )
     hgconf.add_argument(
         "--pyoxidized",
         action="store_true",
         help="build the hg binary using pyoxidizer",
     )
     hgconf.add_argument("--compiler", help="compiler to build with")
     hgconf.add_argument(
         '--extra-config-opt',
         action="append",
         default=[],
         help='set the given config opt in the test hgrc',
     )
     hgconf.add_argument(
         "-l",
         "--local",
         action="store_true",
         help="shortcut for --with-hg=<testdir>/../hg, "
         "--with-rhg=<testdir>/../rust/target/release/rhg if --rhg is set, "
         "and --with-chg=<testdir>/../contrib/chg/chg if --chg is set",
     )
     hgconf.add_argument(
         "--ipv6",
         action="store_true",
         help="prefer IPv6 to IPv4 for network related tests",
     )
     hgconf.add_argument(
         "--pure",
         action="store_true",
         help="use pure Python code instead of C extensions",
     )
     hgconf.add_argument(
         "--rust",
         action="store_true",
         help="use Rust code alongside C extensions",
     )
     hgconf.add_argument(
         "--no-rust",
         action="store_true",
         help="do not use Rust code even if compiled",
     )
     hgconf.add_argument(
         "--with-chg",
         metavar="CHG",
         help="use specified chg wrapper in place of hg",
     )
     hgconf.add_argument(
         "--with-rhg",
         metavar="RHG",
         help="use specified rhg Rust implementation in place of hg",
     )
     hgconf.add_argument(
         "--with-hg",
         metavar="HG",
         help="test using specified hg script rather than a "
         "temporary installation",
     )
 
     reporting = parser.add_argument_group('Results Reporting')
     reporting.add_argument(
         "-C",
         "--annotate",
         action="store_true",
         help="output files annotated with coverage",
     )
     reporting.add_argument(
         "--color",
         choices=["always", "auto", "never"],
         default=os.environ.get('HGRUNTESTSCOLOR', 'auto'),
         help="colorisation: always|auto|never (default: auto)",
     )
     reporting.add_argument(
         "-c",
         "--cover",
         action="store_true",
         help="print a test coverage report",
     )
     reporting.add_argument(
         '--exceptions',
         action='store_true',
         help='log all exceptions and generate an exception report',
     )
     reporting.add_argument(
         "-H",
         "--htmlcov",
         action="store_true",
         help="create an HTML report of the coverage of the files",
     )
     reporting.add_argument(
         "--json",
         action="store_true",
         help="store test result data in 'report.json' file",
     )
     reporting.add_argument(
         "--outputdir",
         help="directory to write error logs to (default=test directory)",
     )
     reporting.add_argument(
         "-n", "--nodiff", action="store_true", help="skip showing test changes"
     )
     reporting.add_argument(
         "-S",
         "--noskips",
         action="store_true",
         help="don't report skip tests verbosely",
     )
     reporting.add_argument(
         "--time", action="store_true", help="time how long each test takes"
     )
     reporting.add_argument("--view", help="external diff viewer")
     reporting.add_argument(
         "--xunit", help="record xunit results at specified path"
     )
 
     for option, (envvar, default) in defaults.items():
         defaults[option] = type(default)(os.environ.get(envvar, default))
     parser.set_defaults(**defaults)
 
     return parser
 
 
 def parseargs(args, parser):
     """Parse arguments with our OptionParser and validate results."""
     options = parser.parse_args(args)
 
     # jython is always pure
     if 'java' in sys.platform or '__pypy__' in sys.modules:
         options.pure = True
 
     if platform.python_implementation() != 'CPython' and options.rust:
         parser.error('Rust extensions are only available with CPython')
 
     if options.pure and options.rust:
         parser.error('--rust cannot be used with --pure')
 
     if options.rust and options.no_rust:
         parser.error('--rust cannot be used with --no-rust')
 
     if options.local:
         if options.with_hg or options.with_rhg or options.with_chg:
             parser.error(
                 '--local cannot be used with --with-hg or --with-rhg or --with-chg'
             )
         if options.pyoxidized:
             parser.error('--pyoxidized does not work with --local (yet)')
         testdir = os.path.dirname(_sys2bytes(canonpath(sys.argv[0])))
         reporootdir = os.path.dirname(testdir)
         pathandattrs = [(b'hg', 'with_hg')]
         if options.chg:
             pathandattrs.append((b'contrib/chg/chg', 'with_chg'))
         if options.rhg:
             pathandattrs.append((b'rust/target/release/rhg', 'with_rhg'))
         for relpath, attr in pathandattrs:
             binpath = os.path.join(reporootdir, relpath)
             if not (WINDOWS or os.access(binpath, os.X_OK)):
                 parser.error(
                     '--local specified, but %r not found or '
                     'not executable' % binpath
                 )
             setattr(options, attr, _bytes2sys(binpath))
 
     if options.with_hg:
         options.with_hg = canonpath(_sys2bytes(options.with_hg))
         if not (
             os.path.isfile(options.with_hg)
             and os.access(options.with_hg, os.X_OK)
         ):
             parser.error('--with-hg must specify an executable hg script')
         if os.path.basename(options.with_hg) not in [b'hg', b'hg.exe']:
             msg = 'warning: --with-hg should specify an hg script, not: %s\n'
             msg %= _bytes2sys(os.path.basename(options.with_hg))
             sys.stderr.write(msg)
             sys.stderr.flush()
 
     if (options.chg or options.with_chg) and WINDOWS:
         parser.error('chg does not work on %s' % os.name)
     if (options.rhg or options.with_rhg) and WINDOWS:
         parser.error('rhg does not work on %s' % os.name)
     if options.pyoxidized and not WINDOWS:
         parser.error('--pyoxidized is currently Windows only')
     if options.with_chg:
         options.chg = False  # no installation to temporary location
         options.with_chg = canonpath(_sys2bytes(options.with_chg))
         if not (
             os.path.isfile(options.with_chg)
             and os.access(options.with_chg, os.X_OK)
         ):
             parser.error('--with-chg must specify a chg executable')
     if options.with_rhg:
         options.rhg = False  # no installation to temporary location
         options.with_rhg = canonpath(_sys2bytes(options.with_rhg))
         if not (
             os.path.isfile(options.with_rhg)
             and os.access(options.with_rhg, os.X_OK)
         ):
             parser.error('--with-rhg must specify a rhg executable')
     if options.chg and options.with_hg:
         # chg shares installation location with hg
         parser.error(
             '--chg does not work when --with-hg is specified '
             '(use --with-chg instead)'
         )
     if options.rhg and options.with_hg:
         # rhg shares installation location with hg
         parser.error(
             '--rhg does not work when --with-hg is specified '
             '(use --with-rhg instead)'
         )
     if options.rhg and options.chg:
         parser.error('--rhg and --chg do not work together')
 
     if options.color == 'always' and not pygmentspresent:
         sys.stderr.write(
             'warning: --color=always ignored because '
             'pygments is not installed\n'
         )
 
     if options.bisect_repo and not options.known_good_rev:
         parser.error("--bisect-repo cannot be used without --known-good-rev")
 
     global useipv6
     if options.ipv6:
         useipv6 = checksocketfamily('AF_INET6')
     else:
         # only use IPv6 if IPv4 is unavailable and IPv6 is available
         useipv6 = (not checksocketfamily('AF_INET')) and checksocketfamily(
             'AF_INET6'
         )
 
     options.anycoverage = options.cover or options.annotate or options.htmlcov
     if options.anycoverage:
         try:
             import coverage
 
             covver = version.StrictVersion(coverage.__version__).version
             if covver < (3, 3):
                 parser.error('coverage options require coverage 3.3 or later')
         except ImportError:
             parser.error('coverage options now require the coverage package')
 
     if options.anycoverage and options.local:
         # this needs some path mangling somewhere, I guess
         parser.error(
             "sorry, coverage options do not work when --local " "is specified"
         )
 
     if options.anycoverage and options.with_hg:
         parser.error(
             "sorry, coverage options do not work when --with-hg " "is specified"
         )
 
     global verbose
     if options.verbose:
         verbose = ''
 
     if options.tmpdir:
         options.tmpdir = canonpath(options.tmpdir)
 
     if options.jobs < 1:
         parser.error('--jobs must be positive')
     if options.interactive and options.debug:
         parser.error("-i/--interactive and -d/--debug are incompatible")
     if options.debug:
         if options.timeout != defaults['timeout']:
             sys.stderr.write('warning: --timeout option ignored with --debug\n')
         if options.slowtimeout != defaults['slowtimeout']:
             sys.stderr.write(
                 'warning: --slowtimeout option ignored with --debug\n'
             )
         options.timeout = 0
         options.slowtimeout = 0
 
     if options.blacklist:
         options.blacklist = parselistfiles(options.blacklist, 'blacklist')
     if options.whitelist:
         options.whitelisted = parselistfiles(options.whitelist, 'whitelist')
     else:
         options.whitelisted = {}
 
     if options.showchannels:
         options.nodiff = True
 
     return options
 
 
 def rename(src, dst):
     """Like os.rename(), trade atomicity and opened files friendliness
     for existing destination support.
     """
     shutil.copy(src, dst)
     os.remove(src)
 
 
 def makecleanable(path):
     """Try to fix directory permission recursively so that the entire tree
     can be deleted"""
     for dirpath, dirnames, _filenames in os.walk(path, topdown=True):
         for d in dirnames:
             p = os.path.join(dirpath, d)
             try:
                 os.chmod(p, os.stat(p).st_mode & 0o777 | 0o700)  # chmod u+rwx
             except OSError:
                 pass
 
 
 _unified_diff = difflib.unified_diff
 if PYTHON3:
     import functools
 
     _unified_diff = functools.partial(difflib.diff_bytes, difflib.unified_diff)
 
 
 def getdiff(expected, output, ref, err):
     servefail = False
     lines = []
     for line in _unified_diff(expected, output, ref, err):
         if line.startswith(b'+++') or line.startswith(b'---'):
             line = line.replace(b'\\', b'/')
             if line.endswith(b' \n'):
                 line = line[:-2] + b'\n'
         lines.append(line)
         if not servefail and line.startswith(
             b'+  abort: child process failed to start'
         ):
             servefail = True
 
     return servefail, lines
 
 
 verbose = False
 
 
 def vlog(*msg):
     """Log only when in verbose mode."""
     if verbose is False:
         return
 
     return log(*msg)
 
 
 # Bytes that break XML even in a CDATA block: control characters 0-31
 # sans \t, \n and \r
 CDATA_EVIL = re.compile(br"[\000-\010\013\014\016-\037]")
 
 # Match feature conditionalized output lines in the form, capturing the feature
 # list in group 2, and the preceeding line output in group 1:
 #
 #   output..output (feature !)\n
 optline = re.compile(br'(.*) \((.+?) !\)\n$')
 
 
 def cdatasafe(data):
     """Make a string safe to include in a CDATA block.
 
     Certain control characters are illegal in a CDATA block, and
     there's no way to include a ]]> in a CDATA either. This function
     replaces illegal bytes with ? and adds a space between the ]] so
     that it won't break the CDATA block.
     """
     return CDATA_EVIL.sub(b'?', data).replace(b']]>', b'] ]>')
 
 
 def log(*msg):
     """Log something to stdout.
 
     Arguments are strings to print.
     """
     with iolock:
         if verbose:
             print(verbose, end=' ')
         for m in msg:
             print(m, end=' ')
         print()
         sys.stdout.flush()
 
 
 def highlightdiff(line, color):
     if not color:
         return line
     assert pygmentspresent
     return pygments.highlight(
         line.decode('latin1'), difflexer, terminal256formatter
     ).encode('latin1')
 
 
 def highlightmsg(msg, color):
     if not color:
         return msg
     assert pygmentspresent
     return pygments.highlight(msg, runnerlexer, runnerformatter)
 
 
 def terminate(proc):
     """Terminate subprocess"""
     vlog('# Terminating process %d' % proc.pid)
     try:
         proc.terminate()
     except OSError:
         pass
 
 
 def killdaemons(pidfile):
     import killdaemons as killmod
 
     return killmod.killdaemons(pidfile, tryhard=False, remove=True, logfn=vlog)
 
 
 class Test(unittest.TestCase):
     """Encapsulates a single, runnable test.
 
     While this class conforms to the unittest.TestCase API, it differs in that
     instances need to be instantiated manually. (Typically, unittest.TestCase
     classes are instantiated automatically by scanning modules.)
     """
 
     # Status code reserved for skipped tests (used by hghave).
     SKIPPED_STATUS = 80
 
     def __init__(
         self,
         path,
         outputdir,
         tmpdir,
         keeptmpdir=False,
         debug=False,
         first=False,
         timeout=None,
         startport=None,
         extraconfigopts=None,
         shell=None,
         hgcommand=None,
         slowtimeout=None,
         usechg=False,
         chgdebug=False,
         useipv6=False,
     ):
         """Create a test from parameters.
 
         path is the full path to the file defining the test.
 
         tmpdir is the main temporary directory to use for this test.
 
         keeptmpdir determines whether to keep the test's temporary directory
         after execution. It defaults to removal (False).
 
         debug mode will make the test execute verbosely, with unfiltered
         output.
 
         timeout controls the maximum run time of the test. It is ignored when
         debug is True. See slowtimeout for tests with #require slow.
 
         slowtimeout overrides timeout if the test has #require slow.
 
         startport controls the starting port number to use for this test. Each
         test will reserve 3 port numbers for execution. It is the caller's
         responsibility to allocate a non-overlapping port range to Test
         instances.
 
         extraconfigopts is an iterable of extra hgrc config options. Values
         must have the form "key=value" (something understood by hgrc). Values
         of the form "foo.key=value" will result in "[foo] key=value".
 
         shell is the shell to execute tests in.
         """
         if timeout is None:
             timeout = defaults['timeout']
         if startport is None:
             startport = defaults['port']
         if slowtimeout is None:
             slowtimeout = defaults['slowtimeout']
         self.path = path
         self.relpath = os.path.relpath(path)
         self.bname = os.path.basename(path)
         self.name = _bytes2sys(self.bname)
         self._testdir = os.path.dirname(path)
         self._outputdir = outputdir
         self._tmpname = os.path.basename(path)
         self.errpath = os.path.join(self._outputdir, b'%s.err' % self.bname)
 
         self._threadtmp = tmpdir
         self._keeptmpdir = keeptmpdir
         self._debug = debug
         self._first = first
         self._timeout = timeout
         self._slowtimeout = slowtimeout
         self._startport = startport
         self._extraconfigopts = extraconfigopts or []
         self._shell = _sys2bytes(shell)
         self._hgcommand = hgcommand or b'hg'
         self._usechg = usechg
         self._chgdebug = chgdebug
         self._useipv6 = useipv6
 
         self._aborted = False
         self._daemonpids = []
         self._finished = None
         self._ret = None
         self._out = None
         self._skipped = None
         self._testtmp = None
         self._chgsockdir = None
 
         self._refout = self.readrefout()
 
     def readrefout(self):
         """read reference output"""
         # If we're not in --debug mode and reference output file exists,
         # check test output against it.
         if self._debug:
             return None  # to match "out is None"
         elif os.path.exists(self.refpath):
             with open(self.refpath, 'rb') as f:
                 return f.read().splitlines(True)
         else:
             return []
 
     # needed to get base class __repr__ running
     @property
     def _testMethodName(self):
         return self.name
 
     def __str__(self):
         return self.name
 
     def shortDescription(self):
         return self.name
 
     def setUp(self):
         """Tasks to perform before run()."""
         self._finished = False
         self._ret = None
         self._out = None
         self._skipped = None
 
         try:
             os.mkdir(self._threadtmp)
         except OSError as e:
             if e.errno != errno.EEXIST:
                 raise
 
         name = self._tmpname
         self._testtmp = os.path.join(self._threadtmp, name)
         os.mkdir(self._testtmp)
 
         # Remove any previous output files.
         if os.path.exists(self.errpath):
             try:
                 os.remove(self.errpath)
             except OSError as e:
                 # We might have raced another test to clean up a .err
                 # file, so ignore ENOENT when removing a previous .err
                 # file.
                 if e.errno != errno.ENOENT:
                     raise
 
         if self._usechg:
             self._chgsockdir = os.path.join(
                 self._threadtmp, b'%s.chgsock' % name
             )
             os.mkdir(self._chgsockdir)
 
     def run(self, result):
         """Run this test and report results against a TestResult instance."""
         # This function is extremely similar to unittest.TestCase.run(). Once
         # we require Python 2.7 (or at least its version of unittest), this
         # function can largely go away.
         self._result = result
         result.startTest(self)
         try:
             try:
                 self.setUp()
             except (KeyboardInterrupt, SystemExit):
                 self._aborted = True
                 raise
             except Exception:
                 result.addError(self, sys.exc_info())
                 return
 
             success = False
             try:
                 self.runTest()
             except KeyboardInterrupt:
                 self._aborted = True
                 raise
             except unittest.SkipTest as e:
                 result.addSkip(self, str(e))
                 # The base class will have already counted this as a
                 # test we "ran", but we want to exclude skipped tests
                 # from those we count towards those run.
                 result.testsRun -= 1
             except self.failureException as e:
                 # This differs from unittest in that we don't capture
                 # the stack trace. This is for historical reasons and
                 # this decision could be revisited in the future,
                 # especially for PythonTest instances.
                 if result.addFailure(self, str(e)):
                     success = True
             except Exception:
                 result.addError(self, sys.exc_info())
             else:
                 success = True
 
             try:
                 self.tearDown()
             except (KeyboardInterrupt, SystemExit):
                 self._aborted = True
                 raise
             except Exception:
                 result.addError(self, sys.exc_info())
                 success = False
 
             if success:
                 result.addSuccess(self)
         finally:
             result.stopTest(self, interrupted=self._aborted)
 
     def runTest(self):
         """Run this test instance.
 
         This will return a tuple describing the result of the test.
         """
         env = self._getenv()
         self._genrestoreenv(env)
         self._daemonpids.append(env['DAEMON_PIDS'])
         self._createhgrc(env['HGRCPATH'])
 
         vlog('# Test', self.name)
 
         ret, out = self._run(env)
         self._finished = True
         self._ret = ret
         self._out = out
 
         def describe(ret):
             if ret < 0:
                 return 'killed by signal: %d' % -ret
             return 'returned error code %d' % ret
 
         self._skipped = False
 
         if ret == self.SKIPPED_STATUS:
             if out is None:  # Debug mode, nothing to parse.
                 missing = ['unknown']
                 failed = None
             else:
                 missing, failed = TTest.parsehghaveoutput(out)
 
             if not missing:
                 missing = ['skipped']
 
             if failed:
                 self.fail('hg have failed checking for %s' % failed[-1])
             else:
                 self._skipped = True
                 raise unittest.SkipTest(missing[-1])
         elif ret == 'timeout':
             self.fail('timed out')
         elif ret is False:
             self.fail('no result code from test')
         elif out != self._refout:
             # Diff generation may rely on written .err file.
             if (
                 (ret != 0 or out != self._refout)
                 and not self._skipped
                 and not self._debug
             ):
                 with open(self.errpath, 'wb') as f:
                     for line in out:
                         f.write(line)
 
             # The result object handles diff calculation for us.
             with firstlock:
                 if self._result.addOutputMismatch(self, ret, out, self._refout):
                     # change was accepted, skip failing
                     return
                 if self._first:
                     global firsterror
                     firsterror = True
 
             if ret:
                 msg = 'output changed and ' + describe(ret)
             else:
                 msg = 'output changed'
 
             self.fail(msg)
         elif ret:
             self.fail(describe(ret))
 
     def tearDown(self):
         """Tasks to perform after run()."""
         for entry in self._daemonpids:
             killdaemons(entry)
         self._daemonpids = []
 
         if self._keeptmpdir:
             log(
                 '\nKeeping testtmp dir: %s\nKeeping threadtmp dir: %s'
                 % (
                     _bytes2sys(self._testtmp),
                     _bytes2sys(self._threadtmp),
                 )
             )
         else:
             try:
                 shutil.rmtree(self._testtmp)
             except OSError:
                 # unreadable directory may be left in $TESTTMP; fix permission
                 # and try again
                 makecleanable(self._testtmp)
                 shutil.rmtree(self._testtmp, True)
             shutil.rmtree(self._threadtmp, True)
 
         if self._usechg:
             # chgservers will stop automatically after they find the socket
             # files are deleted
             shutil.rmtree(self._chgsockdir, True)
 
         if (
             (self._ret != 0 or self._out != self._refout)
             and not self._skipped
             and not self._debug
             and self._out
         ):
             with open(self.errpath, 'wb') as f:
                 for line in self._out:
                     f.write(line)
 
         vlog("# Ret was:", self._ret, '(%s)' % self.name)
 
     def _run(self, env):
         # This should be implemented in child classes to run tests.
         raise unittest.SkipTest('unknown test type')
 
     def abort(self):
         """Terminate execution of this test."""
         self._aborted = True
 
     def _portmap(self, i):
         offset = b'' if i == 0 else b'%d' % i
         return (br':%d\b' % (self._startport + i), b':$HGPORT%s' % offset)
 
     def _getreplacements(self):
         """Obtain a mapping of text replacements to apply to test output.
 
         Test output needs to be normalized so it can be compared to expected
         output. This function defines how some of that normalization will
         occur.
         """
         r = [
             # This list should be parallel to defineport in _getenv
             self._portmap(0),
             self._portmap(1),
             self._portmap(2),
             (br'([^0-9])%s' % re.escape(self._localip()), br'\1$LOCALIP'),
             (br'\bHG_TXNID=TXN:[a-f0-9]{40}\b', br'HG_TXNID=TXN:$ID$'),
         ]
         r.append((self._escapepath(self._testtmp), b'$TESTTMP'))
         if WINDOWS:
             # JSON output escapes backslashes in Windows paths, so also catch a
             # double-escape.
             replaced = self._testtmp.replace(b'\\', br'\\')
             r.append((self._escapepath(replaced), b'$STR_REPR_TESTTMP'))
 
         replacementfile = os.path.join(self._testdir, b'common-pattern.py')
 
         if os.path.exists(replacementfile):
             data = {}
             with open(replacementfile, mode='rb') as source:
                 # the intermediate 'compile' step help with debugging
                 code = compile(source.read(), replacementfile, 'exec')
                 exec(code, data)
                 for value in data.get('substitutions', ()):
                     if len(value) != 2:
                         msg = 'malformatted substitution in %s: %r'
                         msg %= (replacementfile, value)
                         raise ValueError(msg)
                     r.append(value)
         return r
 
     def _escapepath(self, p):
         if WINDOWS:
             return b''.join(
                 c.isalpha()
                 and b'[%s%s]' % (c.lower(), c.upper())
                 or c in b'/\\'
                 and br'[/\\]'
                 or c.isdigit()
                 and c
                 or b'\\' + c
                 for c in [p[i : i + 1] for i in range(len(p))]
             )
         else:
             return re.escape(p)
 
     def _localip(self):
         if self._useipv6:
             return b'::1'
         else:
             return b'127.0.0.1'
 
     def _genrestoreenv(self, testenv):
         """Generate a script that can be used by tests to restore the original
         environment."""
         # Put the restoreenv script inside self._threadtmp
         scriptpath = os.path.join(self._threadtmp, b'restoreenv.sh')
         testenv['HGTEST_RESTOREENV'] = _bytes2sys(scriptpath)
 
         # Only restore environment variable names that the shell allows
         # us to export.
         name_regex = re.compile('^[a-zA-Z][a-zA-Z0-9_]*$')
 
         # Do not restore these variables; otherwise tests would fail.
         reqnames = {'PYTHON', 'TESTDIR', 'TESTTMP'}
 
         with open(scriptpath, 'w') as envf:
             for name, value in origenviron.items():
                 if not name_regex.match(name):
                     # Skip environment variables with unusual names not
                     # allowed by most shells.
                     continue
                 if name in reqnames:
                     continue
                 envf.write('%s=%s\n' % (name, shellquote(value)))
 
             for name in testenv:
                 if name in origenviron or name in reqnames:
                     continue
                 envf.write('unset %s\n' % (name,))
 
     def _getenv(self):
         """Obtain environment variables to use during test execution."""
 
         def defineport(i):
             offset = '' if i == 0 else '%s' % i
             env["HGPORT%s" % offset] = '%s' % (self._startport + i)
 
         env = os.environ.copy()
         env['PYTHONUSERBASE'] = sysconfig.get_config_var('userbase') or ''
         env['HGEMITWARNINGS'] = '1'
         env['TESTTMP'] = _bytes2sys(self._testtmp)
         uid_file = os.path.join(_bytes2sys(self._testtmp), 'UID')
         env['HGTEST_UUIDFILE'] = uid_file
         env['TESTNAME'] = self.name
         env['HOME'] = _bytes2sys(self._testtmp)
         if WINDOWS:
             env['REALUSERPROFILE'] = env['USERPROFILE']
             # py3.8+ ignores HOME: https://bugs.python.org/issue36264
             env['USERPROFILE'] = env['HOME']
         formated_timeout = _bytes2sys(b"%d" % default_defaults['timeout'][1])
         env['HGTEST_TIMEOUT_DEFAULT'] = formated_timeout
         env['HGTEST_TIMEOUT'] = _bytes2sys(b"%d" % self._timeout)
         # This number should match portneeded in _getport
         for port in xrange(3):
             # This list should be parallel to _portmap in _getreplacements
             defineport(port)
         env["HGRCPATH"] = _bytes2sys(os.path.join(self._threadtmp, b'.hgrc'))
         env["DAEMON_PIDS"] = _bytes2sys(
             os.path.join(self._threadtmp, b'daemon.pids')
         )
         env["HGEDITOR"] = (
             '"' + sysexecutable + '"' + ' -c "import sys; sys.exit(0)"'
         )
         env["HGUSER"] = "test"
         env["HGENCODING"] = "ascii"
         env["HGENCODINGMODE"] = "strict"
         env["HGHOSTNAME"] = "test-hostname"
         env['HGIPV6'] = str(int(self._useipv6))
         # See contrib/catapipe.py for how to use this functionality.
         if 'HGTESTCATAPULTSERVERPIPE' not in env:
             # If we don't have HGTESTCATAPULTSERVERPIPE explicitly set, pull the
             # non-test one in as a default, otherwise set to devnull
             env['HGTESTCATAPULTSERVERPIPE'] = env.get(
                 'HGCATAPULTSERVERPIPE', os.devnull
             )
 
         extraextensions = []
         for opt in self._extraconfigopts:
             section, key = opt.split('.', 1)
             if section != 'extensions':
                 continue
             name = key.split('=', 1)[0]
             extraextensions.append(name)
 
         if extraextensions:
             env['HGTESTEXTRAEXTENSIONS'] = ' '.join(extraextensions)
 
         # LOCALIP could be ::1 or 127.0.0.1. Useful for tests that require raw
         # IP addresses.
         env['LOCALIP'] = _bytes2sys(self._localip())
 
         # This has the same effect as Py_LegacyWindowsStdioFlag in exewrapper.c,
         # but this is needed for testing python instances like dummyssh,
         # dummysmtpd.py, and dumbhttp.py.
         if PYTHON3 and WINDOWS:
             env['PYTHONLEGACYWINDOWSSTDIO'] = '1'
 
         # Modified HOME in test environment can confuse Rust tools. So set
         # CARGO_HOME and RUSTUP_HOME automatically if a Rust toolchain is
         # present and these variables aren't already defined.
         cargo_home_path = os.path.expanduser('~/.cargo')
         rustup_home_path = os.path.expanduser('~/.rustup')
 
         if os.path.exists(cargo_home_path) and b'CARGO_HOME' not in osenvironb:
             env['CARGO_HOME'] = cargo_home_path
         if (
             os.path.exists(rustup_home_path)
             and b'RUSTUP_HOME' not in osenvironb
         ):
             env['RUSTUP_HOME'] = rustup_home_path
 
         # Reset some environment variables to well-known values so that
         # the tests produce repeatable output.
         env['LANG'] = env['LC_ALL'] = env['LANGUAGE'] = 'C'
         env['TZ'] = 'GMT'
         env["EMAIL"] = "Foo Bar <[e-mail]>"
         env['COLUMNS'] = '80'
         env['TERM'] = 'xterm'
 
         dropped = [
             'CDPATH',
             'CHGDEBUG',
             'EDITOR',
             'GREP_OPTIONS',
             'HG',
             'HGMERGE',
             'HGPLAIN',
             'HGPLAINEXCEPT',
             'HGPROF',
             'http_proxy',
             'no_proxy',
             'NO_PROXY',
             'PAGER',
             'VISUAL',
         ]
 
         for k in dropped:
             if k in env:
                 del env[k]
 
         # unset env related to hooks
         for k in list(env):
             if k.startswith('HG_'):
                 del env[k]
 
         if self._usechg:
             env['CHGSOCKNAME'] = os.path.join(self._chgsockdir, b'server')
         if self._chgdebug:
             env['CHGDEBUG'] = 'true'
 
         return env
 
     def _createhgrc(self, path):
         """Create an hgrc file for this test."""
         with open(path, 'wb') as hgrc:
             hgrc.write(b'[ui]\n')
             hgrc.write(b'slash = True\n')
             hgrc.write(b'interactive = False\n')
             hgrc.write(b'detailed-exit-code = True\n')
             hgrc.write(b'merge = internal:merge\n')
             hgrc.write(b'mergemarkers = detailed\n')
             hgrc.write(b'promptecho = True\n')
             dummyssh = os.path.join(self._testdir, b'dummyssh')
             hgrc.write(b'ssh = "%s" "%s"\n' % (PYTHON, dummyssh))
             hgrc.write(b'timeout.warn=15\n')
             hgrc.write(b'[chgserver]\n')
             hgrc.write(b'idletimeout=60\n')
             hgrc.write(b'[defaults]\n')
             hgrc.write(b'[devel]\n')
             hgrc.write(b'all-warnings = true\n')
             hgrc.write(b'default-date = 0 0\n')
             hgrc.write(b'[largefiles]\n')
             hgrc.write(
                 b'usercache = %s\n'
                 % (os.path.join(self._testtmp, b'.cache/largefiles'))
             )
             hgrc.write(b'[lfs]\n')
             hgrc.write(
                 b'usercache = %s\n'
                 % (os.path.join(self._testtmp, b'.cache/lfs'))
             )
             hgrc.write(b'[web]\n')
             hgrc.write(b'address = localhost\n')
             hgrc.write(b'ipv6 = %r\n' % self._useipv6)
             hgrc.write(b'server-header = testing stub value\n')
 
             for opt in self._extraconfigopts:
                 section, key = _sys2bytes(opt).split(b'.', 1)
                 assert b'=' in key, (
                     'extra config opt %s must ' 'have an = for assignment' % opt
                 )
                 hgrc.write(b'[%s]\n%s\n' % (section, key))
 
     def fail(self, msg):
         # unittest differentiates between errored and failed.
         # Failed is denoted by AssertionError (by default at least).
         raise AssertionError(msg)
 
     def _runcommand(self, cmd, env, normalizenewlines=False):
         """Run command in a sub-process, capturing the output (stdout and
         stderr).
 
         Return a tuple (exitcode, output). output is None in debug mode.
         """
         if self._debug:
             proc = subprocess.Popen(
                 _bytes2sys(cmd),
                 shell=True,
                 close_fds=closefds,
                 cwd=_bytes2sys(self._testtmp),
                 env=env,
             )
             ret = proc.wait()
             return (ret, None)
 
         proc = Popen4(cmd, self._testtmp, self._timeout, env)
 
         def cleanup():
             terminate(proc)
             ret = proc.wait()
             if ret == 0:
                 ret = signal.SIGTERM << 8
             killdaemons(env['DAEMON_PIDS'])
             return ret
 
         proc.tochild.close()
 
         try:
             output = proc.fromchild.read()
         except KeyboardInterrupt:
             vlog('# Handling keyboard interrupt')
             cleanup()
             raise
 
         ret = proc.wait()
         if wifexited(ret):
             ret = os.WEXITSTATUS(ret)
 
         if proc.timeout:
             ret = 'timeout'
 
         if ret:
             killdaemons(env['DAEMON_PIDS'])
 
         for s, r in self._getreplacements():
             output = re.sub(s, r, output)
 
         if normalizenewlines:
             output = output.replace(b'\r\n', b'\n')
 
         return ret, output.splitlines(True)
 
 
 class PythonTest(Test):
     """A Python-based test."""
 
     @property
     def refpath(self):
         return os.path.join(self._testdir, b'%s.out' % self.bname)
 
     def _run(self, env):
         # Quote the python(3) executable for Windows
         cmd = b'"%s" "%s"' % (PYTHON, self.path)
         vlog("# Running", cmd.decode("utf-8"))
         result = self._runcommand(cmd, env, normalizenewlines=WINDOWS)
         if self._aborted:
             raise KeyboardInterrupt()
 
         return result
 
 
 # Some glob patterns apply only in some circumstances, so the script
 # might want to remove (glob) annotations that otherwise should be
 # retained.
 checkcodeglobpats = [
     # On Windows it looks like \ doesn't require a (glob), but we know
     # better.
     re.compile(br'^pushing to \$TESTTMP/.*[^)]$'),
     re.compile(br'^moving \S+/.*[^)]$'),
     re.compile(br'^pulling from \$TESTTMP/.*[^)]$'),
     # Not all platforms have 127.0.0.1 as loopback (though most do),
     # so we always glob that too.
     re.compile(br'.*\$LOCALIP.*$'),
 ]
 
 bchr = chr
 if PYTHON3:
     bchr = lambda x: bytes([x])
 
 WARN_UNDEFINED = 1
 WARN_YES = 2
 WARN_NO = 3
 
 MARK_OPTIONAL = b" (?)\n"
 
 
 def isoptional(line):
     return line.endswith(MARK_OPTIONAL)
 
 
 class TTest(Test):
     """A "t test" is a test backed by a .t file."""
 
     SKIPPED_PREFIX = b'skipped: '
     FAILED_PREFIX = b'hghave check failed: '
     NEEDESCAPE = re.compile(br'[\x00-\x08\x0b-\x1f\x7f-\xff]').search
 
     ESCAPESUB = re.compile(br'[\x00-\x08\x0b-\x1f\\\x7f-\xff]').sub
     ESCAPEMAP = {bchr(i): br'\x%02x' % i for i in range(256)}
     ESCAPEMAP.update({b'\\': b'\\\\', b'\r': br'\r'})
 
     def __init__(self, path, *args, **kwds):
         # accept an extra "case" parameter
         case = kwds.pop('case', [])
         self._case = case
         self._allcases = {x for y in parsettestcases(path) for x in y}
         super(TTest, self).__init__(path, *args, **kwds)
         if case:
             casepath = b'#'.join(case)
             self.name = '%s#%s' % (self.name, _bytes2sys(casepath))
             self.errpath = b'%s#%s.err' % (self.errpath[:-4], casepath)
             self._tmpname += b'-%s' % casepath.replace(b'#', b'-')
         self._have = {}
 
     @property
     def refpath(self):
         return os.path.join(self._testdir, self.bname)
 
     def _run(self, env):
         with open(self.path, 'rb') as f:
             lines = f.readlines()
 
         # .t file is both reference output and the test input, keep reference
         # output updated with the the test input. This avoids some race
         # conditions where the reference output does not match the actual test.
         if self._refout is not None:
             self._refout = lines
 
         salt, script, after, expected = self._parsetest(lines)
 
         # Write out the generated script.
         fname = b'%s.sh' % self._testtmp
         with open(fname, 'wb') as f:
             for l in script:
                 f.write(l)
 
         cmd = b'%s "%s"' % (self._shell, fname)
         vlog("# Running", cmd.decode("utf-8"))
 
         exitcode, output = self._runcommand(cmd, env)
 
         if self._aborted:
             raise KeyboardInterrupt()
 
         # Do not merge output if skipped. Return hghave message instead.
         # Similarly, with --debug, output is None.
         if exitcode == self.SKIPPED_STATUS or output is None:
             return exitcode, output
 
         return self._processoutput(exitcode, output, salt, after, expected)
 
     def _hghave(self, reqs):
         allreqs = b' '.join(reqs)
 
         self._detectslow(reqs)
 
         if allreqs in self._have:
             return self._have.get(allreqs)
 
         # TODO do something smarter when all other uses of hghave are gone.
         runtestdir = osenvironb[b'RUNTESTDIR']
         tdir = runtestdir.replace(b'\\', b'/')
         proc = Popen4(
             b'%s -c "%s/hghave %s"' % (self._shell, tdir, allreqs),
             self._testtmp,
             0,
             self._getenv(),
         )
         stdout, stderr = proc.communicate()
         ret = proc.wait()
         if wifexited(ret):
             ret = os.WEXITSTATUS(ret)
         if ret == 2:
             print(stdout.decode('utf-8'))
             sys.exit(1)
 
         if ret != 0:
             self._have[allreqs] = (False, stdout)
             return False, stdout
 
         self._have[allreqs] = (True, None)
         return True, None
 
     def _detectslow(self, reqs):
         """update the timeout of slow test when appropriate"""
         if b'slow' in reqs:
             self._timeout = self._slowtimeout
 
     def _iftest(self, args):
         # implements "#if"
         reqs = []
         for arg in args:
             if arg.startswith(b'no-') and arg[3:] in self._allcases:
                 if arg[3:] in self._case:
                     return False
             elif arg in self._allcases:
                 if arg not in self._case:
                     return False
             else:
                 reqs.append(arg)
         self._detectslow(reqs)
         return self._hghave(reqs)[0]
 
     def _parsetest(self, lines):
         # We generate a shell script which outputs unique markers to line
         # up script results with our source. These markers include input
         # line number and the last return code.
         salt = b"SALT%d" % time.time()
 
         def addsalt(line, inpython):
             if inpython:
                 script.append(b'%s %d 0\n' % (salt, line))
             else:
                 script.append(b'echo %s %d $?\n' % (salt, line))
 
         activetrace = []
         session = str(uuid.uuid4())
         if PYTHON3:
             session = session.encode('ascii')
         hgcatapult = os.getenv('HGTESTCATAPULTSERVERPIPE') or os.getenv(
             'HGCATAPULTSERVERPIPE'
         )
 
         def toggletrace(cmd=None):
             if not hgcatapult or hgcatapult == os.devnull:
                 return
 
             if activetrace:
                 script.append(
                     b'echo END %s %s >> "$HGTESTCATAPULTSERVERPIPE"\n'
                     % (session, activetrace[0])
                 )
             if cmd is None:
                 return
 
             if isinstance(cmd, str):
                 quoted = shellquote(cmd.strip())
             else:
                 quoted = shellquote(cmd.strip().decode('utf8')).encode('utf8')
             quoted = quoted.replace(b'\\', b'\\\\')
             script.append(
                 b'echo START %s %s >> "$HGTESTCATAPULTSERVERPIPE"\n'
                 % (session, quoted)
             )
             activetrace[0:] = [quoted]
 
         script = []
 
         # After we run the shell script, we re-unify the script output
         # with non-active parts of the source, with synchronization by our
         # SALT line number markers. The after table contains the non-active
         # components, ordered by line number.
         after = {}
 
         # Expected shell script output.
         expected = {}
 
         pos = prepos = -1
 
         # True or False when in a true or false conditional section
         skipping = None
 
         # We keep track of whether or not we're in a Python block so we
         # can generate the surrounding doctest magic.
         inpython = False
 
         if self._debug:
             script.append(b'set -x\n')
         if os.getenv('MSYSTEM'):
             script.append(b'alias pwd="pwd -W"\n')
 
         if hgcatapult and hgcatapult != os.devnull:
             if PYTHON3:
                 hgcatapult = hgcatapult.encode('utf8')
                 cataname = self.name.encode('utf8')
             else:
                 cataname = self.name
 
             # Kludge: use a while loop to keep the pipe from getting
             # closed by our echo commands. The still-running file gets
             # reaped at the end of the script, which causes the while
             # loop to exit and closes the pipe. Sigh.
             script.append(
                 b'rtendtracing() {\n'
                 b'  echo END %(session)s %(name)s >> %(catapult)s\n'
                 b'  rm -f "$TESTTMP/.still-running"\n'
                 b'}\n'
                 b'trap "rtendtracing" 0\n'
                 b'touch "$TESTTMP/.still-running"\n'
                 b'while [ -f "$TESTTMP/.still-running" ]; do sleep 1; done '
                 b'> %(catapult)s &\n'
                 b'HGCATAPULTSESSION=%(session)s ; export HGCATAPULTSESSION\n'
                 b'echo START %(session)s %(name)s >> %(catapult)s\n'
                 % {
                     b'name': cataname,
                     b'session': session,
                     b'catapult': hgcatapult,
                 }
             )
 
         if self._case:
             casestr = b'#'.join(self._case)
             if isinstance(casestr, str):
                 quoted = shellquote(casestr)
             else:
                 quoted = shellquote(casestr.decode('utf8')).encode('utf8')
             script.append(b'TESTCASE=%s\n' % quoted)
             script.append(b'export TESTCASE\n')
 
         n = 0
         for n, l in enumerate(lines):
             if not l.endswith(b'\n'):
                 l += b'\n'
             if l.startswith(b'#require'):
                 lsplit = l.split()
                 if len(lsplit) < 2 or lsplit[0] != b'#require':
                     after.setdefault(pos, []).append(
                         b'  !!! invalid #require\n'
                     )
                 if not skipping:
                     haveresult, message = self._hghave(lsplit[1:])
                     if not haveresult:
                         script = [b'echo "%s"\nexit 80\n' % message]
                         break
                 after.setdefault(pos, []).append(l)
             elif l.startswith(b'#if'):
                 lsplit = l.split()
                 if len(lsplit) < 2 or lsplit[0] != b'#if':
                     after.setdefault(pos, []).append(b'  !!! invalid #if\n')
                 if skipping is not None:
                     after.setdefault(pos, []).append(b'  !!! nested #if\n')
                 skipping = not self._iftest(lsplit[1:])
                 after.setdefault(pos, []).append(l)
             elif l.startswith(b'#else'):
                 if skipping is None:
                     after.setdefault(pos, []).append(b'  !!! missing #if\n')
                 skipping = not skipping
                 after.setdefault(pos, []).append(l)
             elif l.startswith(b'#endif'):
                 if skipping is None:
                     after.setdefault(pos, []).append(b'  !!! missing #if\n')
                 skipping = None
                 after.setdefault(pos, []).append(l)
             elif skipping:
                 after.setdefault(pos, []).append(l)
             elif l.startswith(b'  >>> '):  # python inlines
                 after.setdefault(pos, []).append(l)
                 prepos = pos
                 pos = n
                 if not inpython:
                     # We've just entered a Python block. Add the header.
                     inpython = True
                     addsalt(prepos, False)  # Make sure we report the exit code.
                     script.append(b'"%s" -m heredoctest <<EOF\n' % PYTHON)
                 addsalt(n, True)
                 script.append(l[2:])
             elif l.startswith(b'  ... '):  # python inlines
                 after.setdefault(prepos, []).append(l)
                 script.append(l[2:])
             elif l.startswith(b'  $ '):  # commands
                 if inpython:
                     script.append(b'EOF\n')
                     inpython = False
                 after.setdefault(pos, []).append(l)
                 prepos = pos
                 pos = n
                 addsalt(n, False)
                 rawcmd = l[4:]
                 cmd = rawcmd.split()
                 toggletrace(rawcmd)
                 if len(cmd) == 2 and cmd[0] == b'cd':
                     rawcmd = b'cd %s || exit 1\n' % cmd[1]
                 script.append(rawcmd)
             elif l.startswith(b'  > '):  # continuations
                 after.setdefault(prepos, []).append(l)
                 script.append(l[4:])
             elif l.startswith(b'  '):  # results
                 # Queue up a list of expected results.
                 expected.setdefault(pos, []).append(l[2:])
             else:
                 if inpython:
                     script.append(b'EOF\n')
                     inpython = False
                 # Non-command/result. Queue up for merged output.
                 after.setdefault(pos, []).append(l)
 
         if inpython:
             script.append(b'EOF\n')
         if skipping is not None:
             after.setdefault(pos, []).append(b'  !!! missing #endif\n')
         addsalt(n + 1, False)
         # Need to end any current per-command trace
         if activetrace:
             toggletrace()
         return salt, script, after, expected
 
     def _processoutput(self, exitcode, output, salt, after, expected):
         # Merge the script output back into a unified test.
         warnonly = WARN_UNDEFINED  # 1: not yet; 2: yes; 3: for sure not
         if exitcode != 0:
             warnonly = WARN_NO
 
         pos = -1
         postout = []
         for out_rawline in output:
             out_line, cmd_line = out_rawline, None
             if salt in out_rawline:
                 out_line, cmd_line = out_rawline.split(salt, 1)
 
             pos, postout, warnonly = self._process_out_line(
                 out_line, pos, postout, expected, warnonly
             )
             pos, postout = self._process_cmd_line(cmd_line, pos, postout, after)
 
         if pos in after:
             postout += after.pop(pos)
 
         if warnonly == WARN_YES:
             exitcode = False  # Set exitcode to warned.
 
         return exitcode, postout
 
     def _process_out_line(self, out_line, pos, postout, expected, warnonly):
         while out_line:
             if not out_line.endswith(b'\n'):
                 out_line += b' (no-eol)\n'
 
             # Find the expected output at the current position.
             els = [None]
             if expected.get(pos, None):
                 els = expected[pos]
 
             optional = []
             for i, el in enumerate(els):
                 r = False
                 if el:
                     r, exact = self.linematch(el, out_line)
                 if isinstance(r, str):
                     if r == '-glob':
                         out_line = ''.join(el.rsplit(' (glob)', 1))
                         r = ''  # Warn only this line.
                     elif r == "retry":
                         postout.append(b'  ' + el)
                     else:
                         log('\ninfo, unknown linematch result: %r\n' % r)
                         r = False
                 if r:
                     els.pop(i)
                     break
                 if el:
                     if isoptional(el):
                         optional.append(i)
                     else:
                         m = optline.match(el)
                         if m:
                             conditions = [c for c in m.group(2).split(b' ')]
 
                             if not self._iftest(conditions):
                                 optional.append(i)
                     if exact:
                         # Don't allow line to be matches against a later
                         # line in the output
                         els.pop(i)
                         break
 
             if r:
                 if r == "retry":
                     continue
                 # clean up any optional leftovers
                 for i in optional:
                     postout.append(b'  ' + els[i])
                 for i in reversed(optional):
                     del els[i]
                 postout.append(b'  ' + el)
             else:
                 if self.NEEDESCAPE(out_line):
                     out_line = TTest._stringescape(
                         b'%s (esc)\n' % out_line.rstrip(b'\n')
                     )
                 postout.append(b'  ' + out_line)  # Let diff deal with it.
                 if r != '':  # If line failed.
                     warnonly = WARN_NO
                 elif warnonly == WARN_UNDEFINED:
                     warnonly = WARN_YES
             break
         else:
             # clean up any optional leftovers
             while expected.get(pos, None):
                 el = expected[pos].pop(0)
                 if el:
                     if not isoptional(el):
                         m = optline.match(el)
                         if m:
                             conditions = [c for c in m.group(2).split(b' ')]
 
                             if self._iftest(conditions):
                                 # Don't append as optional line
                                 continue
                         else:
                             continue
                 postout.append(b'  ' + el)
         return pos, postout, warnonly
 
     def _process_cmd_line(self, cmd_line, pos, postout, after):
         """process a "command" part of a line from unified test output"""
         if cmd_line:
             # Add on last return code.
             ret = int(cmd_line.split()[1])
             if ret != 0:
                 postout.append(b'  [%d]\n' % ret)
             if pos in after:
                 # Merge in non-active test bits.
                 postout += after.pop(pos)
             pos = int(cmd_line.split()[0])
         return pos, postout
 
     @staticmethod
     def rematch(el, l):
         try:
             # parse any flags at the beginning of the regex. Only 'i' is
             # supported right now, but this should be easy to extend.
             flags, el = re.match(br'^(\(\?i\))?(.*)', el).groups()[0:2]
             flags = flags or b''
             el = flags + b'(?:' + el + b')'
             # use \Z to ensure that the regex matches to the end of the string
             if WINDOWS:
                 return re.match(el + br'\r?\n\Z', l)
             return re.match(el + br'\n\Z', l)
         except re.error:
             # el is an invalid regex
             return False
 
     @staticmethod
     def globmatch(el, l):
         # The only supported special characters are * and ? plus / which also
         # matches \ on windows. Escaping of these characters is supported.
         if el + b'\n' == l:
             if os.altsep:
                 # matching on "/" is not needed for this line
                 for pat in checkcodeglobpats:
                     if pat.match(el):
                         return True
                 return b'-glob'
             return True
         el = el.replace(b'$LOCALIP', b'*')
         i, n = 0, len(el)
         res = b''
         while i < n:
             c = el[i : i + 1]
             i += 1
             if c == b'\\' and i < n and el[i : i + 1] in b'*?\\/':
                 res += el[i - 1 : i + 1]
                 i += 1
             elif c == b'*':
                 res += b'.*'
             elif c == b'?':
                 res += b'.'
             elif c == b'/' and os.altsep:
                 res += b'[/\\\\]'
             else:
                 res += re.escape(c)
         return TTest.rematch(res, l)
 
     def linematch(self, el, l):
         if el == l:  # perfect match (fast)
             return True, True
         retry = False
         if isoptional(el):
             retry = "retry"
             el = el[: -len(MARK_OPTIONAL)] + b"\n"
         else:
             m = optline.match(el)
             if m:
                 conditions = [c for c in m.group(2).split(b' ')]
 
                 el = m.group(1) + b"\n"
                 if not self._iftest(conditions):
                     # listed feature missing, should not match
                     return "retry", False
 
         if el.endswith(b" (esc)\n"):
             if PYTHON3:
                 el = el[:-7].decode('unicode_escape') + '\n'
                 el = el.encode('latin-1')
             else:
                 el = el[:-7].decode('string-escape') + '\n'
         if el == l or WINDOWS and el[:-1] + b'\r\n' == l:
             return True, True
         if el.endswith(b" (re)\n"):
             return (TTest.rematch(el[:-6], l) or retry), False
         if el.endswith(b" (glob)\n"):
             # ignore '(glob)' added to l by 'replacements'
             if l.endswith(b" (glob)\n"):
                 l = l[:-8] + b"\n"
             return (TTest.globmatch(el[:-8], l) or retry), False
         if os.altsep:
             _l = l.replace(b'\\', b'/')
             if el == _l or WINDOWS and el[:-1] + b'\r\n' == _l:
                 return True, True
         return retry, True
 
     @staticmethod
     def parsehghaveoutput(lines):
         """Parse hghave log lines.
 
         Return tuple of lists (missing, failed):
           * the missing/unknown features
           * the features for which existence check failed"""
         missing = []
         failed = []
         for line in lines:
             if line.startswith(TTest.SKIPPED_PREFIX):
                 line = line.splitlines()[0]
                 missing.append(_bytes2sys(line[len(TTest.SKIPPED_PREFIX) :]))
             elif line.startswith(TTest.FAILED_PREFIX):
                 line = line.splitlines()[0]
                 failed.append(_bytes2sys(line[len(TTest.FAILED_PREFIX) :]))
 
         return missing, failed
 
     @staticmethod
     def _escapef(m):
         return TTest.ESCAPEMAP[m.group(0)]
 
     @staticmethod
     def _stringescape(s):
         return TTest.ESCAPESUB(TTest._escapef, s)
 
 
 iolock = threading.RLock()
 firstlock = threading.RLock()
 firsterror = False
 
-
-class TestResult(unittest._TextTestResult):
+if PYTHON3:
+    base_class = unittest.TextTestResult
+else:
+    base_class = unittest._TextTestResult
+
+
+class TestResult(base_class):
     """Holds results when executing via unittest."""
 
-    # Don't worry too much about accessing the non-public _TextTestResult.
-    # It is relatively common in Python testing tools.
     def __init__(self, options, *args, **kwargs):
         super(TestResult, self).__init__(*args, **kwargs)
 
         self._options = options
 
         # unittest.TestResult didn't have skipped until 2.7. We need to
         # polyfill it.
         self.skipped = []
 
         # We have a custom "ignored" result that isn't present in any Python
         # unittest implementation. It is very similar to skipped. It may make
         # sense to map it into skip some day.
         self.ignored = []
 
         self.times = []
         self._firststarttime = None
         # Data stored for the benefit of generating xunit reports.
         self.successes = []
         self.faildata = {}
 
         if options.color == 'auto':
             isatty = self.stream.isatty()
             # For some reason, redirecting stdout on Windows disables the ANSI
             # color processing of stderr, which is what is used to print the
             # output.  Therefore, both must be tty on Windows to enable color.
             if WINDOWS:
                 isatty = isatty and sys.stdout.isatty()
             self.color = pygmentspresent and isatty
         elif options.color == 'never':
             self.color = False
         else:  # 'always', for testing purposes
             self.color = pygmentspresent
 
     def onStart(self, test):
         """Can be overriden by custom TestResult"""
 
     def onEnd(self):
         """Can be overriden by custom TestResult"""
 
     def addFailure(self, test, reason):
         self.failures.append((test, reason))
 
         if self._options.first:
             self.stop()
         else:
             with iolock:
                 if reason == "timed out":
                     self.stream.write('t')
                 else:
                     if not self._options.nodiff:
                         self.stream.write('\n')
                         # Exclude the '\n' from highlighting to lex correctly
                         formatted = 'ERROR: %s output changed\n' % test
                         self.stream.write(highlightmsg(formatted, self.color))
                     self.stream.write('!')
 
                 self.stream.flush()
 
     def addSuccess(self, test):
         with iolock:
             super(TestResult, self).addSuccess(test)
         self.successes.append(test)
 
     def addError(self, test, err):
         super(TestResult, self).addError(test, err)
         if self._options.first:
             self.stop()
 
     # Polyfill.
     def addSkip(self, test, reason):
         self.skipped.append((test, reason))
         with iolock:
             if self.showAll:
                 self.stream.writeln('skipped %s' % reason)
             else:
                 self.stream.write('s')
                 self.stream.flush()
 
     def addIgnore(self, test, reason):
         self.ignored.append((test, reason))
         with iolock:
             if self.showAll:
                 self.stream.writeln('ignored %s' % reason)
             else:
                 if reason not in ('not retesting', "doesn't match keyword"):
                     self.stream.write('i')
                 else:
                     self.testsRun += 1
                 self.stream.flush()
 
     def addOutputMismatch(self, test, ret, got, expected):
         """Record a mismatch in test output for a particular test."""
         if self.shouldStop or firsterror:
             # don't print, some other test case already failed and
             # printed, we're just stale and probably failed due to our
             # temp dir getting cleaned up.
             return
 
         accepted = False
         lines = []
 
         with iolock:
             if self._options.nodiff:
                 pass
             elif self._options.view:
                 v = self._options.view
                 subprocess.call(
                     r'"%s" "%s" "%s"'
                     % (v, _bytes2sys(test.refpath), _bytes2sys(test.errpath)),
                     shell=True,
                 )
             else:
                 servefail, lines = getdiff(
                     expected, got, test.refpath, test.errpath
                 )
                 self.stream.write('\n')
                 for line in lines:
                     line = highlightdiff(line, self.color)
                     if PYTHON3:
                         self.stream.flush()
                         self.stream.buffer.write(line)
                         self.stream.buffer.flush()
                     else:
                         self.stream.write(line)
                         self.stream.flush()
 
                 if servefail:
                     raise test.failureException(
                         'server failed to start (HGPORT=%s)' % test._startport
                     )
 
             # handle interactive prompt without releasing iolock
             if self._options.interactive:
                 if test.readrefout() != expected:
                     self.stream.write(
                         'Reference output has changed (run again to prompt '
                         'changes)'
                     )
                 else:
                     self.stream.write('Accept this change? [y/N] ')
                     self.stream.flush()
                     answer = sys.stdin.readline().strip()
                     if answer.lower() in ('y', 'yes'):
                         if test.path.endswith(b'.t'):
                             rename(test.errpath, test.path)
                         else:
                             rename(test.errpath, b'%s.out' % test.path)
                         accepted = True
             if not accepted:
                 self.faildata[test.name] = b''.join(lines)
 
         return accepted
 
     def startTest(self, test):
         super(TestResult, self).startTest(test)
 
         # os.times module computes the user time and system time spent by
         # child's processes along with real elapsed time taken by a process.
         # This module has one limitation. It can only work for Linux user
         # and not for Windows. Hence why we fall back to another function
         # for wall time calculations.
         test.started_times = os.times()
         # TODO use a monotonic clock once support for Python 2.7 is dropped.
         test.started_time = time.time()
         if self._firststarttime is None:  # thread racy but irrelevant
             self._firststarttime = test.started_time
 
     def stopTest(self, test, interrupted=False):
         super(TestResult, self).stopTest(test)
 
         test.stopped_times = os.times()
         stopped_time = time.time()
 
         starttime = test.started_times
         endtime = test.stopped_times
         origin = self._firststarttime
         self.times.append(
             (
                 test.name,
                 endtime[2] - starttime[2],  # user space CPU time
                 endtime[3] - starttime[3],  # sys  space CPU time
                 stopped_time - test.started_time,  # real time
                 test.started_time - origin,  # start date in run context
                 stopped_time - origin,  # end date in run context
             )
         )
 
         if interrupted:
             with iolock:
                 self.stream.writeln(
                     'INTERRUPTED: %s (after %d seconds)'
                     % (test.name, self.times[-1][3])
                 )
 
 
 def getTestResult():
     """
     Returns the relevant test result
     """
     if "CUSTOM_TEST_RESULT" in os.environ:
         testresultmodule = __import__(os.environ["CUSTOM_TEST_RESULT"])
         return testresultmodule.TestResult
     else:
         return TestResult
 
 
 class TestSuite(unittest.TestSuite):
     """Custom unittest TestSuite that knows how to execute Mercurial tests."""
 
     def __init__(
         self,
         testdir,
         jobs=1,
         whitelist=None,
         blacklist=None,
         keywords=None,
         loop=False,
         runs_per_test=1,
         loadtest=None,
         showchannels=False,
         *args,
         **kwargs
     ):
         """Create a new instance that can run tests with a configuration.
 
         testdir specifies the directory where tests are executed from. This
         is typically the ``tests`` directory from Mercurial's source
         repository.
 
         jobs specifies the number of jobs to run concurrently. Each test
         executes on its own thread. Tests actually spawn new processes, so
         state mutation should not be an issue.
 
         If there is only one job, it will use the main thread.
 
         whitelist and blacklist denote tests that have been whitelisted and
         blacklisted, respectively. These arguments don't belong in TestSuite.
         Instead, whitelist and blacklist should be handled by the thing that
         populates the TestSuite with tests. They are present to preserve
         backwards compatible behavior which reports skipped tests as part
         of the results.
 
         keywords denotes key words that will be used to filter which tests
         to execute. This arguably belongs outside of TestSuite.
 
         loop denotes whether to loop over tests forever.
         """
         super(TestSuite, self).__init__(*args, **kwargs)
 
         self._jobs = jobs
         self._whitelist = whitelist
         self._blacklist = blacklist
         self._keywords = keywords
         self._loop = loop
         self._runs_per_test = runs_per_test
         self._loadtest = loadtest
         self._showchannels = showchannels
 
     def run(self, result):
         # We have a number of filters that need to be applied. We do this
         # here instead of inside Test because it makes the running logic for
         # Test simpler.
         tests = []
         num_tests = [0]
         for test in self._tests:
 
             def get():
                 num_tests[0] += 1
                 if getattr(test, 'should_reload', False):
                     return self._loadtest(test, num_tests[0])
                 return test
 
             if not os.path.exists(test.path):
                 result.addSkip(test, "Doesn't exist")
                 continue
 
             is_whitelisted = self._whitelist and (
                 test.relpath in self._whitelist or test.bname in self._whitelist
             )
             if not is_whitelisted:
                 is_blacklisted = self._blacklist and (
                     test.relpath in self._blacklist
                     or test.bname in self._blacklist
                 )
                 if is_blacklisted:
                     result.addSkip(test, 'blacklisted')
                     continue
                 if self._keywords:
                     with open(test.path, 'rb') as f:
                         t = f.read().lower() + test.bname.lower()
                     ignored = False
                     for k in self._keywords.lower().split():
                         if k not in t:
                             result.addIgnore(test, "doesn't match keyword")
                             ignored = True
                             break
 
                     if ignored:
                         continue
             for _ in xrange(self._runs_per_test):
                 tests.append(get())
 
         runtests = list(tests)
         done = queue.Queue()
         running = 0
 
         channels = [""] * self._jobs
 
         def job(test, result):
             for n, v in enumerate(channels):
                 if not v:
                     channel = n
                     break
             else:
                 raise ValueError('Could not find output channel')
             channels[channel] = "=" + test.name[5:].split(".")[0]
             try:
                 test(result)
                 done.put(None)
             except KeyboardInterrupt:
                 pass
             except:  # re-raises
                 done.put(('!', test, 'run-test raised an error, see traceback'))
                 raise
             finally:
                 try:
                     channels[channel] = ''
                 except IndexError:
                     pass
 
         def stat():
             count = 0
             while channels:
                 d = '\n%03s  ' % count
                 for n, v in enumerate(channels):
                     if v:
                         d += v[0]
                         channels[n] = v[1:] or '.'
                     else:
                         d += ' '
                     d += ' '
                 with iolock:
                     sys.stdout.write(d + '  ')
                     sys.stdout.flush()
                 for x in xrange(10):
                     if channels:
                         time.sleep(0.1)
                 count += 1
 
         stoppedearly = False
 
         if self._showchannels:
             statthread = threading.Thread(target=stat, name="stat")
             statthread.start()
 
         try:
             while tests or running:
                 if not done.empty() or running == self._jobs or not tests:
                     try:
                         done.get(True, 1)
                         running -= 1
                         if result and result.shouldStop:
                             stoppedearly = True
                             break
                     except queue.Empty:
                         continue
                 if tests and not running == self._jobs:
                     test = tests.pop(0)
                     if self._loop:
                         if getattr(test, 'should_reload', False):
                             num_tests[0] += 1
                             tests.append(self._loadtest(test, num_tests[0]))
                         else:
                             tests.append(test)
                     if self._jobs == 1:
                         job(test, result)
                     else:
                         t = threading.Thread(
                             target=job, name=test.name, args=(test, result)
                         )
                         t.start()
                     running += 1
 
             # If we stop early we still need to wait on started tests to
             # finish. Otherwise, there is a race between the test completing
             # and the test's cleanup code running. This could result in the
             # test reporting incorrect.
             if stoppedearly:
                 while running:
                     try:
                         done.get(True, 1)
                         running -= 1
                     except queue.Empty:
                         continue
         except KeyboardInterrupt:
             for test in runtests:
                 test.abort()
 
         channels = []
 
         return result
 
 
 # Save the most recent 5 wall-clock runtimes of each test to a
 # human-readable text file named .testtimes. Tests are sorted
 # alphabetically, while times for each test are listed from oldest to
 # newest.
 
 
 def loadtimes(outputdir):
     times = []
     try:
         with open(os.path.join(outputdir, b'.testtimes')) as fp:
             for line in fp:
                 m = re.match('(.*?) ([0-9. ]+)', line)
                 times.append(
                     (m.group(1), [float(t) for t in m.group(2).split()])
                 )
     except IOError as err:
         if err.errno != errno.ENOENT:
             raise
     return times
 
 
 def savetimes(outputdir, result):
     saved = dict(loadtimes(outputdir))
     maxruns = 5
     skipped = {str(t[0]) for t in result.skipped}
     for tdata in result.times:
         test, real = tdata[0], tdata[3]
         if test not in skipped:
             ts = saved.setdefault(test, [])
             ts.append(real)
             ts[:] = ts[-maxruns:]
 
     fd, tmpname = tempfile.mkstemp(
         prefix=b'.testtimes', dir=outputdir, text=True
     )
     with os.fdopen(fd, 'w') as fp:
         for name, ts in sorted(saved.items()):
             fp.write('%s %s\n' % (name, ' '.join(['%.3f' % (t,) for t in ts])))
     timepath = os.path.join(outputdir, b'.testtimes')
     try:
         os.unlink(timepath)
     except OSError:
         pass
     try:
         os.rename(tmpname, timepath)
     except OSError:
         pass
 
 
 class TextTestRunner(unittest.TextTestRunner):
     """Custom unittest test runner that uses appropriate settings."""
 
     def __init__(self, runner, *args, **kwargs):
         super(TextTestRunner, self).__init__(*args, **kwargs)
 
         self._runner = runner
 
         self._result = getTestResult()(
             self._runner.options, self.stream, self.descriptions, self.verbosity
         )
 
     def listtests(self, test):
         test = sorted(test, key=lambda t: t.name)
 
         self._result.onStart(test)
 
         for t in test:
             print(t.name)
             self._result.addSuccess(t)
 
         if self._runner.options.xunit:
             with open(self._runner.options.xunit, "wb") as xuf:
                 self._writexunit(self._result, xuf)
 
         if self._runner.options.json:
             jsonpath = os.path.join(self._runner._outputdir, b'report.json')
             with open(jsonpath, 'w') as fp:
                 self._writejson(self._result, fp)
 
         return self._result
 
     def run(self, test):
         self._result.onStart(test)
         test(self._result)
 
         failed = len(self._result.failures)
         skipped = len(self._result.skipped)
         ignored = len(self._result.ignored)
 
         with iolock:
             self.stream.writeln('')
 
             if not self._runner.options.noskips:
                 for test, msg in sorted(
                     self._result.skipped, key=lambda s: s[0].name
                 ):
                     formatted = 'Skipped %s: %s\n' % (test.name, msg)
                     msg = highlightmsg(formatted, self._result.color)
                     self.stream.write(msg)
             for test, msg in sorted(
                 self._result.failures, key=lambda f: f[0].name
             ):
                 formatted = 'Failed %s: %s\n' % (test.name, msg)
                 self.stream.write(highlightmsg(formatted, self._result.color))
             for test, msg in sorted(
                 self._result.errors, key=lambda e: e[0].name
             ):
                 self.stream.writeln('Errored %s: %s' % (test.name, msg))
 
             if self._runner.options.xunit:
                 with open(self._runner.options.xunit, "wb") as xuf:
                     self._writexunit(self._result, xuf)
 
             if self._runner.options.json:
                 jsonpath = os.path.join(self._runner._outputdir, b'report.json')
                 with open(jsonpath, 'w') as fp:
                     self._writejson(self._result, fp)
 
             self._runner._checkhglib('Tested')
 
             savetimes(self._runner._outputdir, self._result)
 
             if failed and self._runner.options.known_good_rev:
                 self._bisecttests(t for t, m in self._result.failures)
             self.stream.writeln(
                 '# Ran %d tests, %d skipped, %d failed.'
                 % (self._result.testsRun, skipped + ignored, failed)
             )
             if failed:
                 self.stream.writeln(
                     'python hash seed: %s' % os.environ['PYTHONHASHSEED']
                 )
             if self._runner.options.time:
                 self.printtimes(self._result.times)
 
             if self._runner.options.exceptions:
                 exceptions = aggregateexceptions(
                     os.path.join(self._runner._outputdir, b'exceptions')
                 )
 
                 self.stream.writeln('Exceptions Report:')
                 self.stream.writeln(
                     '%d total from %d frames'
                     % (exceptions['total'], len(exceptions['exceptioncounts']))
                 )
                 combined = exceptions['combined']
                 for key in sorted(combined, key=combined.get, reverse=True):
                     frame, line, exc = key
                     totalcount, testcount, leastcount, leasttest = combined[key]
 
                     self.stream.writeln(
                         '%d (%d tests)\t%s: %s (%s - %d total)'
                         % (
                             totalcount,
                             testcount,
                             frame,
                             exc,
                             leasttest,
                             leastcount,
                         )
                     )
 
             self.stream.flush()
 
         return self._result
 
     def _bisecttests(self, tests):
         bisectcmd = ['hg', 'bisect']
         bisectrepo = self._runner.options.bisect_repo
         if bisectrepo:
             bisectcmd.extend(['-R', os.path.abspath(bisectrepo)])
 
         def pread(args):
             env = os.environ.copy()
             env['HGPLAIN'] = '1'
             p = subprocess.Popen(
                 args, stderr=subprocess.STDOUT, stdout=subprocess.PIPE, env=env
             )
             data = p.stdout.read()
             p.wait()
             return data
 
         for test in tests:
             pread(bisectcmd + ['--reset']),
             pread(bisectcmd + ['--bad', '.'])
             pread(bisectcmd + ['--good', self._runner.options.known_good_rev])
             # TODO: we probably need to forward more options
             # that alter hg's behavior inside the tests.
             opts = ''
             withhg = self._runner.options.with_hg
             if withhg:
                 opts += ' --with-hg=%s ' % shellquote(_bytes2sys(withhg))
             rtc = '%s %s %s %s' % (sysexecutable, sys.argv[0], opts, test)
             data = pread(bisectcmd + ['--command', rtc])
             m = re.search(
                 (
                     br'\nThe first (?P<goodbad>bad|good) revision '
                     br'is:\nchangeset: +\d+:(?P<node>[a-f0-9]+)\n.*\n'
                     br'summary: +(?P<summary>[^\n]+)\n'
                 ),
                 data,
                 (re.MULTILINE | re.DOTALL),
             )
             if m is None:
                 self.stream.writeln(
                     'Failed to identify failure point for %s' % test
                 )
                 continue
             dat = m.groupdict()
             verb = 'broken' if dat['goodbad'] == b'bad' else 'fixed'
             self.stream.writeln(
                 '%s %s by %s (%s)'
                 % (
                     test,
                     verb,
                     dat['node'].decode('ascii'),
                     dat['summary'].decode('utf8', 'ignore'),
                 )
             )
 
     def printtimes(self, times):
         # iolock held by run
         self.stream.writeln('# Producing time report')
         times.sort(key=lambda t: (t[3]))
         cols = '%7.3f %7.3f %7.3f %7.3f %7.3f   %s'
         self.stream.writeln(
             '%-7s %-7s %-7s %-7s %-7s   %s'
             % ('start', 'end', 'cuser', 'csys', 'real', 'Test')
         )
         for tdata in times:
             test = tdata[0]
             cuser, csys, real, start, end = tdata[1:6]
             self.stream.writeln(cols % (start, end, cuser, csys, real, test))
 
     @staticmethod
     def _writexunit(result, outf):
         # See http://llg.cubic.org/docs/junit/ for a reference.
         timesd = {t[0]: t[3] for t in result.times}
         doc = minidom.Document()
         s = doc.createElement('testsuite')
         s.setAttribute('errors', "0")  # TODO
         s.setAttribute('failures', str(len(result.failures)))
         s.setAttribute('name', 'run-tests')
         s.setAttribute(
             'skipped', str(len(result.skipped) + len(result.ignored))
         )
         s.setAttribute('tests', str(result.testsRun))
         doc.appendChild(s)
         for tc in result.successes:
             t = doc.createElement('testcase')
             t.setAttribute('name', tc.name)
             tctime = timesd.get(tc.name)
             if tctime is not None:
                 t.setAttribute('time', '%.3f' % tctime)
             s.appendChild(t)
         for tc, err in sorted(result.faildata.items()):
             t = doc.createElement('testcase')
             t.setAttribute('name', tc)
             tctime = timesd.get(tc)
             if tctime is not None:
                 t.setAttribute('time', '%.3f' % tctime)
             # createCDATASection expects a unicode or it will
             # convert using default conversion rules, which will
             # fail if string isn't ASCII.
             err = cdatasafe(err).decode('utf-8', 'replace')
             cd = doc.createCDATASection(err)
             # Use 'failure' here instead of 'error' to match errors = 0,
             # failures = len(result.failures) in the testsuite element.
             failelem = doc.createElement('failure')
             failelem.setAttribute('message', 'output changed')
             failelem.setAttribute('type', 'output-mismatch')
             failelem.appendChild(cd)
             t.appendChild(failelem)
             s.appendChild(t)
         for tc, message in result.skipped:
             # According to the schema, 'skipped' has no attributes. So store
             # the skip message as a text node instead.
             t = doc.createElement('testcase')
             t.setAttribute('name', tc.name)
             binmessage = message.encode('utf-8')
             message = cdatasafe(binmessage).decode('utf-8', 'replace')
             cd = doc.createCDATASection(message)
             skipelem = doc.createElement('skipped')
             skipelem.appendChild(cd)
             t.appendChild(skipelem)
             s.appendChild(t)
         outf.write(doc.toprettyxml(indent='  ', encoding='utf-8'))
 
     @staticmethod
     def _writejson(result, outf):
         timesd = {}
         for tdata in result.times:
             test = tdata[0]
             timesd[test] = tdata[1:]
 
         outcome = {}
         groups = [
             ('success', ((tc, None) for tc in result.successes)),
             ('failure', result.failures),
             ('skip', result.skipped),
         ]
         for res, testcases in groups:
             for tc, __ in testcases:
                 if tc.name in timesd:
                     diff = result.faildata.get(tc.name, b'')
                     try:
                         diff = diff.decode('unicode_escape')
                     except UnicodeDecodeError as e:
                         diff = '%r decoding diff, sorry' % e
                     tres = {
                         'result': res,
                         'time': ('%0.3f' % timesd[tc.name][2]),
                         'cuser': ('%0.3f' % timesd[tc.name][0]),
                         'csys': ('%0.3f' % timesd[tc.name][1]),
                         'start': ('%0.3f' % timesd[tc.name][3]),
                         'end': ('%0.3f' % timesd[tc.name][4]),
                         'diff': diff,
                     }
                 else:
                     # blacklisted test
                     tres = {'result': res}
 
                 outcome[tc.name] = tres
         jsonout = json.dumps(
             outcome, sort_keys=True, indent=4, separators=(',', ': ')
         )
         outf.writelines(("testreport =", jsonout))
 
 
 def sorttests(testdescs, previoustimes, shuffle=False):
     """Do an in-place sort of tests."""
     if shuffle:
         random.shuffle(testdescs)
         return
 
     if previoustimes:
 
         def sortkey(f):
             f = f['path']
             if f in previoustimes:
                 # Use most recent time as estimate
                 return -(previoustimes[f][-1])
             else:
                 # Default to a rather arbitrary value of 1 second for new tests
                 return -1.0
 
     else:
         # keywords for slow tests
         slow = {
             b'svn': 10,
             b'cvs': 10,
             b'hghave': 10,
             b'largefiles-update': 10,
             b'run-tests': 10,
             b'corruption': 10,
             b'race': 10,
             b'i18n': 10,
             b'check': 100,
             b'gendoc': 100,
             b'contrib-perf': 200,
             b'merge-combination': 100,
         }
         perf = {}
 
         def sortkey(f):
             # run largest tests first, as they tend to take the longest
             f = f['path']
             try:
                 return perf[f]
             except KeyError:
                 try:
                     val = -os.stat(f).st_size
                 except OSError as e:
                     if e.errno != errno.ENOENT:
                         raise
                     perf[f] = -1e9  # file does not exist, tell early
                     return -1e9
                 for kw, mul in slow.items():
                     if kw in f:
                         val *= mul
                 if f.endswith(b'.py'):
                     val /= 10.0
                 perf[f] = val / 1000.0
                 return perf[f]
 
     testdescs.sort(key=sortkey)
 
 
 class TestRunner(object):
     """Holds context for executing tests.
 
     Tests rely on a lot of state. This object holds it for them.
     """
 
     # Programs required to run tests.
     REQUIREDTOOLS = [
         b'diff',
         b'grep',
         b'unzip',
         b'gunzip',
         b'bunzip2',
         b'sed',
     ]
 
     # Maps file extensions to test class.
     TESTTYPES = [
         (b'.py', PythonTest),
         (b'.t', TTest),
     ]
 
     def __init__(self):
         self.options = None
         self._hgroot = None
         self._testdir = None
         self._outputdir = None
         self._hgtmp = None
         self._installdir = None
         self._bindir = None
         # a place for run-tests.py to generate executable it needs
         self._custom_bin_dir = None
         self._pythondir = None
         # True if we had to infer the pythondir from --with-hg
         self._pythondir_inferred = False
         self._coveragefile = None
         self._createdfiles = []
         self._hgcommand = None
         self._hgpath = None
         self._portoffset = 0
         self._ports = {}
 
     def run(self, args, parser=None):
         """Run the test suite."""
         oldmask = os.umask(0o22)
         try:
             parser = parser or getparser()
             options = parseargs(args, parser)
             tests = [_sys2bytes(a) for a in options.tests]
             if options.test_list is not None:
                 for listfile in options.test_list:
                     with open(listfile, 'rb') as f:
                         tests.extend(t for t in f.read().splitlines() if t)
             self.options = options
 
             self._checktools()
             testdescs = self.findtests(tests)
             if options.profile_runner:
                 import statprof
 
                 statprof.start()
             result = self._run(testdescs)
             if options.profile_runner:
                 statprof.stop()
                 statprof.display()
             return result
 
         finally:
             os.umask(oldmask)
 
     def _run(self, testdescs):
         testdir = getcwdb()
         # assume all tests in same folder for now
         if testdescs:
             pathname = os.path.dirname(testdescs[0]['path'])
             if pathname:
                 testdir = os.path.join(testdir, pathname)
         self._testdir = osenvironb[b'TESTDIR'] = testdir
         if self.options.outputdir:
             self._outputdir = canonpath(_sys2bytes(self.options.outputdir))
         else:
             self._outputdir = getcwdb()
             if testdescs and pathname:
                 self._outputdir = os.path.join(self._outputdir, pathname)
         previoustimes = {}
         if self.options.order_by_runtime:
             previoustimes = dict(loadtimes(self._outputdir))
         sorttests(testdescs, previoustimes, shuffle=self.options.random)
 
         if 'PYTHONHASHSEED' not in os.environ:
             # use a random python hash seed all the time
             # we do the randomness ourself to know what seed is used
             os.environ['PYTHONHASHSEED'] = str(random.getrandbits(32))
 
         # Rayon (Rust crate for multi-threading) will use all logical CPU cores
         # by default, causing thrashing on high-cpu-count systems.
         # Setting its limit to 3 during tests should still let us uncover
         # multi-threading bugs while keeping the thrashing reasonable.
         os.environ.setdefault("RAYON_NUM_THREADS", "3")
 
         if self.options.tmpdir:
             self.options.keep_tmpdir = True
             tmpdir = _sys2bytes(self.options.tmpdir)
             if os.path.exists(tmpdir):
                 # Meaning of tmpdir has changed since 1.3: we used to create
                 # HGTMP inside tmpdir; now HGTMP is tmpdir.  So fail if
                 # tmpdir already exists.
                 print("error: temp dir %r already exists" % tmpdir)
                 return 1
 
             os.makedirs(tmpdir)
         else:
             d = None
             if WINDOWS:
                 # without this, we get the default temp dir location, but
                 # in all lowercase, which causes troubles with paths (issue3490)
                 d = osenvironb.get(b'TMP', None)
             tmpdir = tempfile.mkdtemp(b'', b'hgtests.', d)
 
         self._hgtmp = osenvironb[b'HGTMP'] = os.path.realpath(tmpdir)
 
         self._custom_bin_dir = os.path.join(self._hgtmp, b'custom-bin')
         os.makedirs(self._custom_bin_dir)
 
         if self.options.with_hg:
             self._installdir = None
             whg = self.options.with_hg
             self._bindir = os.path.dirname(os.path.realpath(whg))
             assert isinstance(self._bindir, bytes)
             self._hgcommand = os.path.basename(whg)
 
             normbin = os.path.normpath(os.path.abspath(whg))
             normbin = normbin.replace(_sys2bytes(os.sep), b'/')
 
             # Other Python scripts in the test harness need to
             # `import mercurial`. If `hg` is a Python script, we assume
             # the Mercurial modules are relative to its path and tell the tests
             # to load Python modules from its directory.
             with open(whg, 'rb') as fh:
                 initial = fh.read(1024)
 
             if re.match(b'#!.*python', initial):
                 self._pythondir = self._bindir
             # If it looks like our in-repo Rust binary, use the source root.
             # This is a bit hacky. But rhg is still not supported outside the
             # source directory. So until it is, do the simple thing.
             elif re.search(b'/rust/target/[^/]+/hg', normbin):
                 self._pythondir = os.path.dirname(self._testdir)
             # Fall back to the legacy behavior.
             else:
                 self._pythondir = self._bindir
             self._pythondir_inferred = True
 
         else:
             self._installdir = os.path.join(self._hgtmp, b"install")
             self._bindir = os.path.join(self._installdir, b"bin")
             self._hgcommand = b'hg'
             self._pythondir = os.path.join(self._installdir, b"lib", b"python")
 
         # Force the use of hg.exe instead of relying on MSYS to recognize hg is
         # a python script and feed it to python.exe.  Legacy stdio is force
         # enabled by hg.exe, and this is a more realistic way to launch hg
         # anyway.
         if WINDOWS and not self._hgcommand.endswith(b'.exe'):
             self._hgcommand += b'.exe'
 
         real_hg = os.path.join(self._bindir, self._hgcommand)
         osenvironb[b'HGTEST_REAL_HG'] = real_hg
         # set CHGHG, then replace "hg" command by "chg"
         chgbindir = self._bindir
         if self.options.chg or self.options.with_chg:
             osenvironb[b'CHG_INSTALLED_AS_HG'] = b'1'
             osenvironb[b'CHGHG'] = real_hg
         else:
             # drop flag for hghave
             osenvironb.pop(b'CHG_INSTALLED_AS_HG', None)
         if self.options.chg:
             self._hgcommand = b'chg'
         elif self.options.with_chg:
             chgbindir = os.path.dirname(os.path.realpath(self.options.with_chg))
             self._hgcommand = os.path.basename(self.options.with_chg)
 
         # configure fallback and replace "hg" command by "rhg"
         rhgbindir = self._bindir
         if self.options.rhg or self.options.with_rhg:
             # Affects hghave.py
             osenvironb[b'RHG_INSTALLED_AS_HG'] = b'1'
             # Affects configuration. Alternatives would be setting configuration through
             # `$HGRCPATH` but some tests override that, or changing `_hgcommand` to include
             # `--config` but that disrupts tests that print command lines and check expected
             # output.
             osenvironb[b'RHG_ON_UNSUPPORTED'] = b'fallback'
             osenvironb[b'RHG_FALLBACK_EXECUTABLE'] = real_hg
         else:
             # drop flag for hghave
             osenvironb.pop(b'RHG_INSTALLED_AS_HG', None)
         if self.options.rhg:
             self._hgcommand = b'rhg'
         elif self.options.with_rhg:
             rhgbindir = os.path.dirname(os.path.realpath(self.options.with_rhg))
             self._hgcommand = os.path.basename(self.options.with_rhg)
 
         if self.options.pyoxidized:
             testdir = os.path.dirname(_sys2bytes(canonpath(sys.argv[0])))
             reporootdir = os.path.dirname(testdir)
             # XXX we should ideally install stuff instead of using the local build
             bin_path = (
                 b'build/pyoxidizer/x86_64-pc-windows-msvc/release/app/hg.exe'
             )
             full_path = os.path.join(reporootdir, bin_path)
             self._hgcommand = full_path
             # Affects hghave.py
             osenvironb[b'PYOXIDIZED_INSTALLED_AS_HG'] = b'1'
         else:
             osenvironb.pop(b'PYOXIDIZED_INSTALLED_AS_HG', None)
 
         osenvironb[b"BINDIR"] = self._bindir
         osenvironb[b"PYTHON"] = PYTHON
 
         fileb = _sys2bytes(__file__)
         runtestdir = os.path.abspath(os.path.dirname(fileb))
         osenvironb[b'RUNTESTDIR'] = runtestdir
         if PYTHON3:
             sepb = _sys2bytes(os.pathsep)
         else:
             sepb = os.pathsep
         path = [self._bindir, runtestdir] + osenvironb[b"PATH"].split(sepb)
         if os.path.islink(__file__):
             # test helper will likely be at the end of the symlink
             realfile = os.path.realpath(fileb)
             realdir = os.path.abspath(os.path.dirname(realfile))
             path.insert(2, realdir)
         if chgbindir != self._bindir:
             path.insert(1, chgbindir)
         if rhgbindir != self._bindir:
             path.insert(1, rhgbindir)
         if self._testdir != runtestdir:
             path = [self._testdir] + path
         path = [self._custom_bin_dir] + path
         osenvironb[b"PATH"] = sepb.join(path)
 
         # Include TESTDIR in PYTHONPATH so that out-of-tree extensions
         # can run .../tests/run-tests.py test-foo where test-foo
         # adds an extension to HGRC. Also include run-test.py directory to
         # import modules like heredoctest.
         pypath = [self._pythondir, self._testdir, runtestdir]
         # We have to augment PYTHONPATH, rather than simply replacing
         # it, in case external libraries are only available via current
         # PYTHONPATH.  (In particular, the Subversion bindings on OS X
         # are in /opt/subversion.)
         oldpypath = osenvironb.get(IMPL_PATH)
         if oldpypath:
             pypath.append(oldpypath)
         osenvironb[IMPL_PATH] = sepb.join(pypath)
 
         if self.options.pure:
             os.environ["HGTEST_RUN_TESTS_PURE"] = "--pure"
             os.environ["HGMODULEPOLICY"] = "py"
         if self.options.rust:
             os.environ["HGMODULEPOLICY"] = "rust+c"
         if self.options.no_rust:
             current_policy = os.environ.get("HGMODULEPOLICY", "")
             if current_policy.startswith("rust+"):
                 os.environ["HGMODULEPOLICY"] = current_policy[len("rust+") :]
             os.environ.pop("HGWITHRUSTEXT", None)
 
         if self.options.allow_slow_tests:
             os.environ["HGTEST_SLOW"] = "slow"
         elif 'HGTEST_SLOW' in os.environ:
             del os.environ['HGTEST_SLOW']
 
         self._coveragefile = os.path.join(self._testdir, b'.coverage')
 
         if self.options.exceptions:
             exceptionsdir = os.path.join(self._outputdir, b'exceptions')
             try:
                 os.makedirs(exceptionsdir)
             except OSError as e:
                 if e.errno != errno.EEXIST:
                     raise
 
             # Remove all existing exception reports.
             for f in os.listdir(exceptionsdir):
                 os.unlink(os.path.join(exceptionsdir, f))
 
             osenvironb[b'HGEXCEPTIONSDIR'] = exceptionsdir
             logexceptions = os.path.join(self._testdir, b'logexceptions.py')
             self.options.extra_config_opt.append(
                 'extensions.logexceptions=%s' % logexceptions.decode('utf-8')
             )
 
         vlog("# Using TESTDIR", _bytes2sys(self._testdir))
         vlog("# Using RUNTESTDIR", _bytes2sys(osenvironb[b'RUNTESTDIR']))
         vlog("# Using HGTMP", _bytes2sys(self._hgtmp))
         vlog("# Using PATH", os.environ["PATH"])
         vlog(
             "# Using",
             _bytes2sys(IMPL_PATH),
             _bytes2sys(osenvironb[IMPL_PATH]),
         )
         vlog("# Writing to directory", _bytes2sys(self._outputdir))
 
         try:
             return self._runtests(testdescs) or 0
         finally:
             time.sleep(0.1)
             self._cleanup()
 
     def findtests(self, args):
         """Finds possible test files from arguments.
 
         If you wish to inject custom tests into the test harness, this would
         be a good function to monkeypatch or override in a derived class.
         """
         if not args:
             if self.options.changed:
                 proc = Popen4(
                     b'hg st --rev "%s" -man0 .'
                     % _sys2bytes(self.options.changed),
                     None,
                     0,
                 )
                 stdout, stderr = proc.communicate()
                 args = stdout.strip(b'\0').split(b'\0')
             else:
                 args = os.listdir(b'.')
 
         expanded_args = []
         for arg in args:
             if os.path.isdir(arg):
                 if not arg.endswith(b'/'):
                     arg += b'/'
                 expanded_args.extend([arg + a for a in os.listdir(arg)])
             else:
                 expanded_args.append(arg)
         args = expanded_args
 
         testcasepattern = re.compile(br'([\w-]+\.t|py)(?:#([a-zA-Z0-9_\-.#]+))')
         tests = []
         for t in args:
             case = []
 
             if not (
                 os.path.basename(t).startswith(b'test-')
                 and (t.endswith(b'.py') or t.endswith(b'.t'))
             ):
 
                 m = testcasepattern.match(os.path.basename(t))
                 if m is not None:
                     t_basename, casestr = m.groups()
                     t = os.path.join(os.path.dirname(t), t_basename)
                     if casestr:
                         case = casestr.split(b'#')
                 else:
                     continue
 
             if t.endswith(b'.t'):
                 # .t file may contain multiple test cases
                 casedimensions = parsettestcases(t)
                 if casedimensions:
                     cases = []
 
                     def addcases(case, casedimensions):
                         if not casedimensions:
                             cases.append(case)
                         else:
                             for c in casedimensions[0]:
                                 addcases(case + [c], casedimensions[1:])
 
                     addcases([], casedimensions)
                     if case and case in cases:
                         cases = [case]
                     elif case:
                         # Ignore invalid cases
                         cases = []
                     else:
                         pass
                     tests += [{'path': t, 'case': c} for c in sorted(cases)]
                 else:
                     tests.append({'path': t})
             else:
                 tests.append({'path': t})
 
         if self.options.retest:
             retest_args = []
             for test in tests:
                 errpath = self._geterrpath(test)
                 if os.path.exists(errpath):
                     retest_args.append(test)
             tests = retest_args
         return tests
 
     def _runtests(self, testdescs):
         def _reloadtest(test, i):
             # convert a test back to its description dict
             desc = {'path': test.path}
             case = getattr(test, '_case', [])
             if case:
                 desc['case'] = case
             return self._gettest(desc, i)
 
         try:
             if self.options.restart:
                 orig = list(testdescs)
                 while testdescs:
                     desc = testdescs[0]
                     errpath = self._geterrpath(desc)
                     if os.path.exists(errpath):
                         break
                     testdescs.pop(0)
                 if not testdescs:
                     print("running all tests")
                     testdescs = orig
 
             tests = [self._gettest(d, i) for i, d in enumerate(testdescs)]
             num_tests = len(tests) * self.options.runs_per_test
 
             jobs = min(num_tests, self.options.jobs)
 
             failed = False
             kws = self.options.keywords
             if kws is not None and PYTHON3:
                 kws = kws.encode('utf-8')
 
             suite = TestSuite(
                 self._testdir,
                 jobs=jobs,
                 whitelist=self.options.whitelisted,
                 blacklist=self.options.blacklist,
                 keywords=kws,
                 loop=self.options.loop,
                 runs_per_test=self.options.runs_per_test,
                 showchannels=self.options.showchannels,
                 tests=tests,
                 loadtest=_reloadtest,
             )
             verbosity = 1
             if self.options.list_tests:
                 verbosity = 0
             elif self.options.verbose:
                 verbosity = 2
             runner = TextTestRunner(self, verbosity=verbosity)
 
             if self.options.list_tests:
                 result = runner.listtests(suite)
             else:
                 self._usecorrectpython()
                 if self._installdir:
                     self._installhg()
                     self._checkhglib("Testing")
                 if self.options.chg:
                     assert self._installdir
                     self._installchg()
                 if self.options.rhg:
                     assert self._installdir
                     self._installrhg()
                 elif self.options.pyoxidized:
                     self._build_pyoxidized()
                 self._use_correct_mercurial()
 
                 log(
                     'running %d tests using %d parallel processes'
                     % (num_tests, jobs)
                 )
 
                 result = runner.run(suite)
 
             if result.failures or result.errors:
                 failed = True
 
             result.onEnd()
 
             if self.options.anycoverage:
                 self._outputcoverage()
         except KeyboardInterrupt:
             failed = True
             print("\ninterrupted!")
 
         if failed:
             return 1
 
     def _geterrpath(self, test):
         # test['path'] is a relative path
         if 'case' in test:
             # for multiple dimensions test cases
             casestr = b'#'.join(test['case'])
             errpath = b'%s#%s.err' % (test['path'], casestr)
         else:
             errpath = b'%s.err' % test['path']
         if self.options.outputdir:
             self._outputdir = canonpath(_sys2bytes(self.options.outputdir))
             errpath = os.path.join(self._outputdir, errpath)
         return errpath
 
     def _getport(self, count):
         port = self._ports.get(count)  # do we have a cached entry?
         if port is None:
             portneeded = 3
             # above 100 tries we just give up and let test reports failure
             for tries in xrange(100):
                 allfree = True
                 port = self.options.port + self._portoffset
                 for idx in xrange(portneeded):
                     if not checkportisavailable(port + idx):
                         allfree = False
                         break
                 self._portoffset += portneeded
                 if allfree:
                     break
             self._ports[count] = port
         return port
 
     def _gettest(self, testdesc, count):
         """Obtain a Test by looking at its filename.
 
         Returns a Test instance. The Test may not be runnable if it doesn't
         map to a known type.
         """
         path = testdesc['path']
         lctest = path.lower()
         testcls = Test
 
         for ext, cls in self.TESTTYPES:
             if lctest.endswith(ext):
                 testcls = cls
                 break
 
         refpath = os.path.join(getcwdb(), path)
         tmpdir = os.path.join(self._hgtmp, b'child%d' % count)
 
         # extra keyword parameters. 'case' is used by .t tests
         kwds = {k: testdesc[k] for k in ['case'] if k in testdesc}
 
         t = testcls(
             refpath,
             self._outputdir,
             tmpdir,
             keeptmpdir=self.options.keep_tmpdir,
             debug=self.options.debug,
             first=self.options.first,
             timeout=self.options.timeout,
             startport=self._getport(count),
             extraconfigopts=self.options.extra_config_opt,
             shell=self.options.shell,
             hgcommand=self._hgcommand,
             usechg=bool(self.options.with_chg or self.options.chg),
             chgdebug=self.options.chg_debug,
             useipv6=useipv6,
             **kwds
         )
         t.should_reload = True
         return t
 
     def _cleanup(self):
         """Clean up state from this test invocation."""
         if self.options.keep_tmpdir:
             return
 
         vlog("# Cleaning up HGTMP", _bytes2sys(self._hgtmp))
         shutil.rmtree(self._hgtmp, True)
         for f in self._createdfiles:
             try:
                 os.remove(f)
             except OSError:
                 pass
 
     def _usecorrectpython(self):
         """Configure the environment to use the appropriate Python in tests."""
         # Tests must use the same interpreter as us or bad things will happen.
         if WINDOWS and PYTHON3:
             pyexe_names = [b'python', b'python3', b'python.exe']
         elif WINDOWS:
             pyexe_names = [b'python', b'python.exe']
         elif PYTHON3:
             pyexe_names = [b'python', b'python3']
         else:
             pyexe_names = [b'python', b'python2']
 
         # os.symlink() is a thing with py3 on Windows, but it requires
         # Administrator rights.
         if not WINDOWS and getattr(os, 'symlink', None):
             msg = "# Making python executable in test path a symlink to '%s'"
             msg %= sysexecutable
             vlog(msg)
             for pyexename in pyexe_names:
                 mypython = os.path.join(self._custom_bin_dir, pyexename)
                 try:
                     if os.readlink(mypython) == sysexecutable:
                         continue
                     os.unlink(mypython)
                 except OSError as err:
                     if err.errno != errno.ENOENT:
                         raise
                 if self._findprogram(pyexename) != sysexecutable:
                     try:
                         os.symlink(sysexecutable, mypython)
                         self._createdfiles.append(mypython)
                     except OSError as err:
                         # child processes may race, which is harmless
                         if err.errno != errno.EEXIST:
                             raise
         elif WINDOWS and not os.getenv('MSYSTEM'):
             raise AssertionError('cannot run test on Windows without MSYSTEM')
         else:
             # Generate explicit file instead of symlink
             #
             # This is especially important as Windows doesn't have
             # `python3.exe`, and MSYS cannot understand the reparse point with
             # that name provided by Microsoft.  Create a simple script on PATH
             # with that name that delegates to the py3 launcher so the shebang
             # lines work.
             esc_executable = _sys2bytes(shellquote(sysexecutable))
             for pyexename in pyexe_names:
                 stub_exec_path = os.path.join(self._custom_bin_dir, pyexename)
                 with open(stub_exec_path, 'wb') as f:
                     f.write(b'#!/bin/sh\n')
                     f.write(b'%s "$@"\n' % esc_executable)
 
             if WINDOWS:
                 if not PYTHON3:
                     # lets try to build a valid python3 executable for the
                     # scrip that requires it.
                     py3exe_name = os.path.join(self._custom_bin_dir, b'python3')
                     with open(py3exe_name, 'wb') as f:
                         f.write(b'#!/bin/sh\n')
                         f.write(b'py -3 "$@"\n')
 
                 # adjust the path to make sur the main python finds it own dll
                 path = os.environ['PATH'].split(os.pathsep)
                 main_exec_dir = os.path.dirname(sysexecutable)
                 extra_paths = [_bytes2sys(self._custom_bin_dir), main_exec_dir]
 
                 # Binaries installed by pip into the user area like pylint.exe may
                 # not be in PATH by default.
                 appdata = os.environ.get('APPDATA')
                 vi = sys.version_info
                 if appdata is not None:
                     python_dir = 'Python%d%d' % (vi[0], vi[1])
                     scripts_path = [appdata, 'Python', python_dir, 'Scripts']
                     if not PYTHON3:
                         scripts_path = [appdata, 'Python', 'Scripts']
                     scripts_dir = os.path.join(*scripts_path)
                     extra_paths.append(scripts_dir)
 
                 os.environ['PATH'] = os.pathsep.join(extra_paths + path)
 
     def _use_correct_mercurial(self):
         target_exec = os.path.join(self._custom_bin_dir, b'hg')
         if self._hgcommand != b'hg':
             # shutil.which only accept bytes from 3.8
             real_exec = which(self._hgcommand)
             if real_exec is None:
                 raise ValueError('could not find exec path for "%s"', real_exec)
             if real_exec == target_exec:
                 # do not overwrite something with itself
                 return
             if WINDOWS:
                 with open(target_exec, 'wb') as f:
                     f.write(b'#!/bin/sh\n')
                     escaped_exec = shellquote(_bytes2sys(real_exec))
                     f.write(b'%s "$@"\n' % _sys2bytes(escaped_exec))
             else:
                 os.symlink(real_exec, target_exec)
             self._createdfiles.append(target_exec)
 
     def _installhg(self):
         """Install hg into the test environment.
 
         This will also configure hg with the appropriate testing settings.
         """
         vlog("# Performing temporary installation of HG")
         installerrs = os.path.join(self._hgtmp, b"install.err")
         compiler = ''
         if self.options.compiler:
             compiler = '--compiler ' + self.options.compiler
         setup_opts = b""
         if self.options.pure:
             setup_opts = b"--pure"
         elif self.options.rust:
             setup_opts = b"--rust"
         elif self.options.no_rust:
             setup_opts = b"--no-rust"
 
         # Run installer in hg root
         script = os.path.realpath(sys.argv[0])
         exe = sysexecutable
         if PYTHON3:
             compiler = _sys2bytes(compiler)
             script = _sys2bytes(script)
             exe = _sys2bytes(exe)
         hgroot = os.path.dirname(os.path.dirname(script))
         self._hgroot = hgroot
         os.chdir(hgroot)
         nohome = b'--home=""'
         if WINDOWS:
             # The --home="" trick works only on OS where os.sep == '/'
             # because of a distutils convert_path() fast-path. Avoid it at
             # least on Windows for now, deal with .pydistutils.cfg bugs
             # when they happen.
             nohome = b''
         cmd = (
             b'"%(exe)s" setup.py %(setup_opts)s clean --all'
             b' build %(compiler)s --build-base="%(base)s"'
             b' install --force --prefix="%(prefix)s"'
             b' --install-lib="%(libdir)s"'
             b' --install-scripts="%(bindir)s" %(nohome)s >%(logfile)s 2>&1'
             % {
                 b'exe': exe,
                 b'setup_opts': setup_opts,
                 b'compiler': compiler,
                 b'base': os.path.join(self._hgtmp, b"build"),
                 b'prefix': self._installdir,
                 b'libdir': self._pythondir,
                 b'bindir': self._bindir,
                 b'nohome': nohome,
                 b'logfile': installerrs,
             }
         )
 
         # setuptools requires install directories to exist.
         def makedirs(p):
             try:
                 os.makedirs(p)
             except OSError as e:
                 if e.errno != errno.EEXIST:
                     raise
 
         makedirs(self._pythondir)
         makedirs(self._bindir)
 
         vlog("# Running", cmd.decode("utf-8"))
         if subprocess.call(_bytes2sys(cmd), shell=True) == 0:
             if not self.options.verbose:
                 try:
                     os.remove(installerrs)
                 except OSError as e:
                     if e.errno != errno.ENOENT:
                         raise
         else:
             with open(installerrs, 'rb') as f:
                 for line in f:
                     if PYTHON3:
                         sys.stdout.buffer.write(line)
                     else:
                         sys.stdout.write(line)
             sys.exit(1)
         os.chdir(self._testdir)
 
         hgbat = os.path.join(self._bindir, b'hg.bat')
         if os.path.isfile(hgbat):
             # hg.bat expects to be put in bin/scripts while run-tests.py
             # installation layout put it in bin/ directly. Fix it
             with open(hgbat, 'rb') as f:
                 data = f.read()
             if br'"%~dp0..\python" "%~dp0hg" %*' in data:
                 data = data.replace(
                     br'"%~dp0..\python" "%~dp0hg" %*',
                     b'"%~dp0python" "%~dp0hg" %*',
                 )
                 with open(hgbat, 'wb') as f:
                     f.write(data)
             else:
                 print('WARNING: cannot fix hg.bat reference to python.exe')
 
         if self.options.anycoverage:
             custom = os.path.join(
                 osenvironb[b'RUNTESTDIR'], b'sitecustomize.py'
             )
             target = os.path.join(self._pythondir, b'sitecustomize.py')
             vlog('# Installing coverage trigger to %s' % target)
             shutil.copyfile(custom, target)
             rc = os.path.join(self._testdir, b'.coveragerc')
             vlog('# Installing coverage rc to %s' % rc)
             osenvironb[b'COVERAGE_PROCESS_START'] = rc
             covdir = os.path.join(self._installdir, b'..', b'coverage')
             try:
                 os.mkdir(covdir)
             except OSError as e:
                 if e.errno != errno.EEXIST:
                     raise
 
             osenvironb[b'COVERAGE_DIR'] = covdir
 
     def _checkhglib(self, verb):
         """Ensure that the 'mercurial' package imported by python is
         the one we expect it to be.  If not, print a warning to stderr."""
         if self._pythondir_inferred:
             # The pythondir has been inferred from --with-hg flag.
             # We cannot expect anything sensible here.
             return
         expecthg = os.path.join(self._pythondir, b'mercurial')
         actualhg = self._gethgpath()
         if os.path.abspath(actualhg) != os.path.abspath(expecthg):
             sys.stderr.write(
                 'warning: %s with unexpected mercurial lib: %s\n'
                 '         (expected %s)\n' % (verb, actualhg, expecthg)
             )
 
     def _gethgpath(self):
         """Return the path to the mercurial package that is actually found by
         the current Python interpreter."""
         if self._hgpath is not None:
             return self._hgpath
 
         cmd = b'"%s" -c "import mercurial; print (mercurial.__path__[0])"'
         cmd = cmd % PYTHON
         if PYTHON3:
             cmd = _bytes2sys(cmd)
 
         p = subprocess.Popen(cmd, stdout=subprocess.PIPE, shell=True)
         out, err = p.communicate()
 
         self._hgpath = out.strip()
 
         return self._hgpath
 
     def _installchg(self):
         """Install chg into the test environment"""
         vlog('# Performing temporary installation of CHG')
         assert os.path.dirname(self._bindir) == self._installdir
         assert self._hgroot, 'must be called after _installhg()'
         cmd = b'"%(make)s" clean install PREFIX="%(prefix)s"' % {
             b'make': b'make',  # TODO: switch by option or environment?
             b'prefix': self._installdir,
         }
         cwd = os.path.join(self._hgroot, b'contrib', b'chg')
         vlog("# Running", cmd)
         proc = subprocess.Popen(
             cmd,
             shell=True,
             cwd=cwd,
             stdin=subprocess.PIPE,
             stdout=subprocess.PIPE,
             stderr=subprocess.STDOUT,
         )
         out, _err = proc.communicate()
         if proc.returncode != 0:
             if PYTHON3:
                 sys.stdout.buffer.write(out)
             else:
                 sys.stdout.write(out)
             sys.exit(1)
 
     def _installrhg(self):
         """Install rhg into the test environment"""
         vlog('# Performing temporary installation of rhg')
         assert os.path.dirname(self._bindir) == self._installdir
         assert self._hgroot, 'must be called after _installhg()'
         cmd = b'"%(make)s" install-rhg PREFIX="%(prefix)s"' % {
             b'make': b'make',  # TODO: switch by option or environment?
             b'prefix': self._installdir,
         }
         cwd = self._hgroot
         vlog("# Running", cmd)
         proc = subprocess.Popen(
             cmd,
             shell=True,
             cwd=cwd,
             stdin=subprocess.PIPE,
             stdout=subprocess.PIPE,
             stderr=subprocess.STDOUT,
         )
         out, _err = proc.communicate()
         if proc.returncode != 0:
             if PYTHON3:
                 sys.stdout.buffer.write(out)
             else:
                 sys.stdout.write(out)
             sys.exit(1)
 
     def _build_pyoxidized(self):
         """build a pyoxidized version of mercurial into the test environment
 
         Ideally this function would be `install_pyoxidier` and would both build
         and install pyoxidier. However we are starting small to get pyoxidizer
         build binary to testing quickly.
         """
         vlog('# build a pyoxidized version of Mercurial')
         assert os.path.dirname(self._bindir) == self._installdir
         assert self._hgroot, 'must be called after _installhg()'
         cmd = b'"%(make)s" pyoxidizer-windows-tests' % {
             b'make': b'make',
         }
         cwd = self._hgroot
         vlog("# Running", cmd)
         proc = subprocess.Popen(
             _bytes2sys(cmd),
             shell=True,
             cwd=_bytes2sys(cwd),
             stdin=subprocess.PIPE,
             stdout=subprocess.PIPE,
             stderr=subprocess.STDOUT,
         )
         out, _err = proc.communicate()
         if proc.returncode != 0:
             if PYTHON3:
                 sys.stdout.buffer.write(out)
             else:
                 sys.stdout.write(out)
             sys.exit(1)
 
     def _outputcoverage(self):
         """Produce code coverage output."""
         import coverage
 
         coverage = coverage.coverage
 
         vlog('# Producing coverage report')
         # chdir is the easiest way to get short, relative paths in the
         # output.
         os.chdir(self._hgroot)
         covdir = os.path.join(_bytes2sys(self._installdir), '..', 'coverage')
         cov = coverage(data_file=os.path.join(covdir, 'cov'))
 
         # Map install directory paths back to source directory.
         cov.config.paths['srcdir'] = ['.', _bytes2sys(self._pythondir)]
 
         cov.combine()
 
         omit = [
             _bytes2sys(os.path.join(x, b'*'))
             for x in [self._bindir, self._testdir]
         ]
         cov.report(ignore_errors=True, omit=omit)
 
         if self.options.htmlcov:
             htmldir = os.path.join(_bytes2sys(self._outputdir), 'htmlcov')
             cov.html_report(directory=htmldir, omit=omit)
         if self.options.annotate:
             adir = os.path.join(_bytes2sys(self._outputdir), 'annotated')
             if not os.path.isdir(adir):
                 os.mkdir(adir)
             cov.annotate(directory=adir, omit=omit)
 
     def _findprogram(self, program):
         """Search PATH for a executable program"""
         dpb = _sys2bytes(os.defpath)
         sepb = _sys2bytes(os.pathsep)
         for p in osenvironb.get(b'PATH', dpb).split(sepb):
             name = os.path.join(p, program)
             if WINDOWS or os.access(name, os.X_OK):
                 return _bytes2sys(name)
         return None
 
     def _checktools(self):
         """Ensure tools required to run tests are present."""
         for p in self.REQUIREDTOOLS:
             if WINDOWS and not p.endswith(b'.exe'):
                 p += b'.exe'
             found = self._findprogram(p)
             p = p.decode("utf-8")
             if found:
                 vlog("# Found prerequisite", p, "at", found)
             else:
                 print("WARNING: Did not find prerequisite tool: %s " % p)
 
 
 def aggregateexceptions(path):
     exceptioncounts = collections.Counter()
     testsbyfailure = collections.defaultdict(set)
     failuresbytest = collections.defaultdict(set)
 
     for f in os.listdir(path):
         with open(os.path.join(path, f), 'rb') as fh:
             data = fh.read().split(b'\0')
             if len(data) != 5:
                 continue
 
             exc, mainframe, hgframe, hgline, testname = data
             exc = exc.decode('utf-8')
             mainframe = mainframe.decode('utf-8')
             hgframe = hgframe.decode('utf-8')
             hgline = hgline.decode('utf-8')
             testname = testname.decode('utf-8')
 
             key = (hgframe, hgline, exc)
             exceptioncounts[key] += 1
             testsbyfailure[key].add(testname)
             failuresbytest[testname].add(key)
 
     # Find test having fewest failures for each failure.
     leastfailing = {}
     for key, tests in testsbyfailure.items():
         fewesttest = None
         fewestcount = 99999999
         for test in sorted(tests):
             if len(failuresbytest[test]) < fewestcount:
                 fewesttest = test
                 fewestcount = len(failuresbytest[test])
 
         leastfailing[key] = (fewestcount, fewesttest)
 
     # Create a combined counter so we can sort by total occurrences and
     # impacted tests.
     combined = {}
     for key in exceptioncounts:
         combined[key] = (
             exceptioncounts[key],
             len(testsbyfailure[key]),
             leastfailing[key][0],
             leastfailing[key][1],
         )
 
     return {
         'exceptioncounts': exceptioncounts,
         'total': sum(exceptioncounts.values()),
         'combined': combined,
         'leastfailing': leastfailing,
         'byfailure': testsbyfailure,
         'bytest': failuresbytest,
     }
 
 
 if __name__ == '__main__':
     if WINDOWS and not os.getenv('MSYSTEM'):
         print('cannot run test on Windows without MSYSTEM', file=sys.stderr)
         print(
             '(if you need to do so contact the mercurial devs: '
             '[e-mail])',
             file=sys.stderr,
         )
         sys.exit(255)
 
     runner = TestRunner()
 
     try:
         import msvcrt
 
         msvcrt.setmode(sys.stdin.fileno(), os.O_BINARY)
         msvcrt.setmode(sys.stdout.fileno(), os.O_BINARY)
         msvcrt.setmode(sys.stderr.fileno(), os.O_BINARY)
     except ImportError:
         pass
 
     sys.exit(runner.run(sys.argv[1:]))

tests/test-chg.t

@@ -1,539 +1,539 b''
 #require chg
 
   $ mkdir log
   $ cp $HGRCPATH $HGRCPATH.unconfigured
   $ cat <<'EOF' >> $HGRCPATH
   > [cmdserver]
   > log = $TESTTMP/log/server.log
   > max-log-files = 1
   > max-log-size = 10 kB
   > EOF
   $ cp $HGRCPATH $HGRCPATH.orig
 
   $ filterlog () {
   >   sed -e 's!^[0-9/]* [0-9:]* ([0-9]*)>!YYYY/MM/DD HH:MM:SS (PID)>!' \
   >       -e 's!\(setprocname\|received fds\|setenv\): .*!\1: ...!' \
   >       -e 's!\(confighash\|mtimehash\) = [0-9a-f]*!\1 = ...!g' \
   >       -e 's!\(in \)[0-9.]*s\b!\1 ...s!g' \
   >       -e 's!\(pid\)=[0-9]*!\1=...!g' \
   >       -e 's!\(/server-\)[0-9a-f]*!\1...!g'
   > }
 
 init repo
 
   $ chg init foo
   $ cd foo
 
 ill-formed config
 
   $ chg status
   $ echo '=brokenconfig' >> $HGRCPATH
   $ chg status
   config error at * =brokenconfig (glob)
   [30]
 
   $ cp $HGRCPATH.orig $HGRCPATH
 
 long socket path
 
   $ sockpath=$TESTTMP/this/path/should/be/longer/than/one-hundred-and-seven/characters/where/107/is/the/typical/size/limit/of/unix-domain-socket
   $ mkdir -p $sockpath
   $ bakchgsockname=$CHGSOCKNAME
   $ CHGSOCKNAME=$sockpath/server
   $ export CHGSOCKNAME
   $ chg root
   $TESTTMP/foo
   $ rm -rf $sockpath
   $ CHGSOCKNAME=$bakchgsockname
   $ export CHGSOCKNAME
 
   $ cd ..
 
 editor
 ------
 
   $ cat >> pushbuffer.py <<EOF
   > def reposetup(ui, repo):
   >     repo.ui.pushbuffer(subproc=True)
   > EOF
 
   $ chg init editor
   $ cd editor
 
 by default, system() should be redirected to the client:
 
   $ touch foo
   $ CHGDEBUG= HGEDITOR=cat chg ci -Am channeled --edit 2>&1 \
   > | egrep "HG:|run 'cat"
   chg: debug: * run 'cat "*"' at '$TESTTMP/editor' (glob)
   HG: Enter commit message.  Lines beginning with 'HG:' are removed.
   HG: Leave message empty to abort commit.
   HG: --
   HG: user: test
   HG: branch 'default'
   HG: added foo
 
 but no redirection should be made if output is captured:
 
   $ touch bar
   $ CHGDEBUG= HGEDITOR=cat chg ci -Am bufferred --edit \
   > --config extensions.pushbuffer="$TESTTMP/pushbuffer.py" 2>&1 \
   > | egrep "HG:|run 'cat"
   [1]
 
 check that commit commands succeeded:
 
   $ hg log -T '{rev}:{desc}\n'
   1:bufferred
   0:channeled
 
   $ cd ..
 
 pager
 -----
 
   $ cat >> fakepager.py <<EOF
   > import sys
   > for line in sys.stdin:
   >     sys.stdout.write('paged! %r\n' % line)
   > EOF
 
 enable pager extension globally, but spawns the master server with no tty:
 
   $ chg init pager
   $ cd pager
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > pager =
   > [pager]
   > pager = "$PYTHON" $TESTTMP/fakepager.py
   > EOF
   $ chg version > /dev/null
   $ touch foo
   $ chg ci -qAm foo
 
 pager should be enabled if the attached client has a tty:
 
   $ chg log -l1 -q --config ui.formatted=True
   paged! '0:1f7b0de80e11\n'
   $ chg log -l1 -q --config ui.formatted=False
   0:1f7b0de80e11
 
 chg waits for pager if runcommand raises
 
   $ cat > $TESTTMP/crash.py <<EOF
   > from mercurial import registrar
   > cmdtable = {}
   > command = registrar.command(cmdtable)
   > @command(b'crash')
   > def pagercrash(ui, repo, *pats, **opts):
   >     ui.write(b'going to crash\n')
   >     raise Exception('.')
   > EOF
 
   $ cat > $TESTTMP/fakepager.py <<EOF
   > from __future__ import absolute_import
   > import sys
   > import time
   > for line in iter(sys.stdin.readline, ''):
   >     if 'crash' in line: # only interested in lines containing 'crash'
   >         # if chg exits when pager is sleeping (incorrectly), the output
   >         # will be captured by the next test case
   >         time.sleep(1)
   >         sys.stdout.write('crash-pager: %s' % line)
   > EOF
 
   $ cat >> .hg/hgrc <<EOF
   > [extensions]
   > crash = $TESTTMP/crash.py
   > EOF
 
   $ chg crash --pager=on --config ui.formatted=True 2>/dev/null
   crash-pager: going to crash
   [255]
 
 no stdout data should be printed after pager quits, and the buffered data
 should never persist (issue6207)
 
 "killed!" may be printed if terminated by SIGPIPE, which isn't important
 in this test.
 
   $ cat > $TESTTMP/bulkwrite.py <<'EOF'
   > import time
   > from mercurial import error, registrar
   > cmdtable = {}
   > command = registrar.command(cmdtable)
   > @command(b'bulkwrite')
   > def bulkwrite(ui, repo, *pats, **opts):
   >     ui.write(b'going to write massive data\n')
   >     ui.flush()
   >     t = time.time()
   >     while time.time() - t < 2:
   >         ui.write(b'x' * 1023 + b'\n')  # will be interrupted by SIGPIPE
   >     raise error.Abort(b"write() doesn't block")
   > EOF
 
   $ cat > $TESTTMP/fakepager.py <<'EOF'
   > import sys
   > import time
   > sys.stdout.write('paged! %r\n' % sys.stdin.readline())
   > time.sleep(1)  # new data will be written
   > EOF
 
   $ cat >> .hg/hgrc <<EOF
   > [extensions]
   > bulkwrite = $TESTTMP/bulkwrite.py
   > EOF
 
   $ chg bulkwrite --pager=on --color no --config ui.formatted=True
   paged! 'going to write massive data\n'
   killed! (?)
   [255]
 
   $ chg bulkwrite --pager=on --color no --config ui.formatted=True
   paged! 'going to write massive data\n'
   killed! (?)
   [255]
 
   $ cd ..
 
 missing stdio
 -------------
 
   $ CHGDEBUG=1 chg version -q 0<&-
   chg: debug: * stdio fds are missing (glob)
   chg: debug: * execute original hg (glob)
   Mercurial Distributed SCM * (glob)
 
 server lifecycle
 ----------------
 
 chg server should be restarted on code change, and old server will shut down
 automatically. In this test, we use the following time parameters:
 
  - "sleep 1" to make mtime different
  - "sleep 2" to notice mtime change (polling interval is 1 sec)
 
 set up repository with an extension:
 
   $ chg init extreload
   $ cd extreload
   $ touch dummyext.py
   $ cat <<EOF >> .hg/hgrc
   > [extensions]
   > dummyext = dummyext.py
   > EOF
 
 isolate socket directory for stable result:
 
   $ OLDCHGSOCKNAME=$CHGSOCKNAME
   $ mkdir chgsock
   $ CHGSOCKNAME=`pwd`/chgsock/server
 
 warm up server:
 
   $ CHGDEBUG= chg log 2>&1 | egrep 'instruction|start'
   chg: debug: * start cmdserver at $TESTTMP/extreload/chgsock/server.* (glob)
 
 new server should be started if extension modified:
 
   $ sleep 1
   $ touch dummyext.py
   $ CHGDEBUG= chg log 2>&1 | egrep 'instruction|start'
   chg: debug: * instruction: unlink $TESTTMP/extreload/chgsock/server-* (glob)
   chg: debug: * instruction: reconnect (glob)
   chg: debug: * start cmdserver at $TESTTMP/extreload/chgsock/server.* (glob)
 
 old server will shut down, while new server should still be reachable:
 
   $ sleep 2
   $ CHGDEBUG= chg log 2>&1 | (egrep 'instruction|start' || true)
 
 socket file should never be unlinked by old server:
 (simulates unowned socket by updating mtime, which makes sure server exits
 at polling cycle)
 
   $ ls chgsock/server-*
   chgsock/server-* (glob)
   $ touch chgsock/server-*
   $ sleep 2
   $ ls chgsock/server-*
   chgsock/server-* (glob)
 
 since no server is reachable from socket file, new server should be started:
 (this test makes sure that old server shut down automatically)
 
   $ CHGDEBUG= chg log 2>&1 | egrep 'instruction|start'
   chg: debug: * start cmdserver at $TESTTMP/extreload/chgsock/server.* (glob)
 
 shut down servers and restore environment:
 
   $ rm -R chgsock
   $ sleep 2
   $ CHGSOCKNAME=$OLDCHGSOCKNAME
   $ cd ..
 
 check that server events are recorded:
 
   $ ls log
   server.log
   server.log.1
 
 print only the last 10 lines, since we aren't sure how many records are
 preserved (since setprocname isn't available on py3 and pure version,
 the 10th-most-recent line is different when using py3):
 
   $ cat log/server.log.1 log/server.log | tail -10 | filterlog
   YYYY/MM/DD HH:MM:SS (PID)> confighash = ... mtimehash = ... (no-setprocname !)
   YYYY/MM/DD HH:MM:SS (PID)> forked worker process (pid=...)
   YYYY/MM/DD HH:MM:SS (PID)> setprocname: ... (setprocname !)
   YYYY/MM/DD HH:MM:SS (PID)> received fds: ...
   YYYY/MM/DD HH:MM:SS (PID)> chdir to '$TESTTMP/extreload'
   YYYY/MM/DD HH:MM:SS (PID)> setumask 18
   YYYY/MM/DD HH:MM:SS (PID)> setenv: ...
   YYYY/MM/DD HH:MM:SS (PID)> confighash = ... mtimehash = ...
   YYYY/MM/DD HH:MM:SS (PID)> validate: []
   YYYY/MM/DD HH:MM:SS (PID)> worker process exited (pid=...)
   YYYY/MM/DD HH:MM:SS (PID)> $TESTTMP/extreload/chgsock/server-... is not owned, exiting.
 
 global data mutated by schems
 -----------------------------
 
   $ hg init schemes
   $ cd schemes
 
 initial state
 
   $ cat > .hg/hgrc <<'EOF'
   > [extensions]
   > schemes =
   > [schemes]
   > foo = https://foo.example.org/
   > EOF
   $ hg debugexpandscheme foo://expanded
   https://foo.example.org/expanded
   $ hg debugexpandscheme bar://unexpanded
   bar://unexpanded
 
 add bar
 
   $ cat > .hg/hgrc <<'EOF'
   > [extensions]
   > schemes =
   > [schemes]
   > foo = https://foo.example.org/
   > bar = https://bar.example.org/
   > EOF
   $ hg debugexpandscheme foo://expanded
   https://foo.example.org/expanded
   $ hg debugexpandscheme bar://expanded
   https://bar.example.org/expanded
 
 remove foo
 
   $ cat > .hg/hgrc <<'EOF'
   > [extensions]
   > schemes =
   > [schemes]
   > bar = https://bar.example.org/
   > EOF
   $ hg debugexpandscheme foo://unexpanded
   foo://unexpanded
   $ hg debugexpandscheme bar://expanded
   https://bar.example.org/expanded
 
   $ cd ..
 
 repository cache
 ----------------
 
   $ rm log/server.log*
   $ cp $HGRCPATH.unconfigured $HGRCPATH
   $ cat <<'EOF' >> $HGRCPATH
   > [cmdserver]
   > log = $TESTTMP/log/server.log
   > max-repo-cache = 1
   > track-log = command, repocache
   > EOF
 
 isolate socket directory for stable result:
 
   $ OLDCHGSOCKNAME=$CHGSOCKNAME
   $ mkdir chgsock
   $ CHGSOCKNAME=`pwd`/chgsock/server
 
 create empty repo and cache it:
 
   $ hg init cached
   $ hg id -R cached
   000000000000 tip
   $ sleep 1
 
 modify repo (and cache will be invalidated):
 
   $ touch cached/a
   $ hg ci -R cached -Am 'add a'
   adding a
   $ sleep 1
 
 read cached repo:
 
   $ hg log -R cached
   changeset:   0:ac82d8b1f7c4
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     add a
   
   $ sleep 1
 
 discard cached from LRU cache:
 
   $ hg clone cached cached2
   updating to branch default
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -R cached2
   ac82d8b1f7c4 tip
   $ sleep 1
 
 read uncached repo:
 
   $ hg log -R cached
   changeset:   0:ac82d8b1f7c4
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     add a
   
   $ sleep 1
 
 shut down servers and restore environment:
 
   $ rm -R chgsock
   $ sleep 2
   $ CHGSOCKNAME=$OLDCHGSOCKNAME
 
 check server log:
 
   $ cat log/server.log | filterlog
   YYYY/MM/DD HH:MM:SS (PID)> worker process exited (pid=...)
   YYYY/MM/DD HH:MM:SS (PID)> worker process exited (pid=...) (?)
   YYYY/MM/DD HH:MM:SS (PID)> init cached
   YYYY/MM/DD HH:MM:SS (PID)> id -R cached
   YYYY/MM/DD HH:MM:SS (PID)> loaded repo into cache: $TESTTMP/cached (in  ...s)
   YYYY/MM/DD HH:MM:SS (PID)> repo from cache: $TESTTMP/cached
   YYYY/MM/DD HH:MM:SS (PID)> ci -R cached -Am 'add a'
   YYYY/MM/DD HH:MM:SS (PID)> loaded repo into cache: $TESTTMP/cached (in  ...s)
   YYYY/MM/DD HH:MM:SS (PID)> repo from cache: $TESTTMP/cached
   YYYY/MM/DD HH:MM:SS (PID)> log -R cached
   YYYY/MM/DD HH:MM:SS (PID)> loaded repo into cache: $TESTTMP/cached (in  ...s)
   YYYY/MM/DD HH:MM:SS (PID)> clone cached cached2
   YYYY/MM/DD HH:MM:SS (PID)> id -R cached2
   YYYY/MM/DD HH:MM:SS (PID)> loaded repo into cache: $TESTTMP/cached2 (in  ...s)
   YYYY/MM/DD HH:MM:SS (PID)> log -R cached
   YYYY/MM/DD HH:MM:SS (PID)> loaded repo into cache: $TESTTMP/cached (in  ...s)
 
 Test that chg works (sets to the user's actual LC_CTYPE) even when python
 "coerces" the locale (py3.7+)
 
   $ cat > $TESTTMP/debugenv.py <<EOF
   > from mercurial import encoding
   > from mercurial import registrar
   > cmdtable = {}
   > command = registrar.command(cmdtable)
   > @command(b'debugenv', [], b'', norepo=True)
   > def debugenv(ui):
   >     for k in [b'LC_ALL', b'LC_CTYPE', b'LANG']:
   >         v = encoding.environ.get(k)
   >         if v is not None:
   >             ui.write(b'%s=%s\n' % (k, encoding.environ[k]))
   > EOF
 (hg keeps python's modified LC_CTYPE, chg doesn't)
   $ (unset LC_ALL; unset LANG; LC_CTYPE= "$CHGHG" \
   >    --config extensions.debugenv=$TESTTMP/debugenv.py debugenv)
   LC_CTYPE=C.UTF-8 (py37 !)
   LC_CTYPE= (no-py37 !)
   $ (unset LC_ALL; unset LANG; LC_CTYPE= chg \
   >    --config extensions.debugenv=$TESTTMP/debugenv.py debugenv)
   LC_CTYPE=
   $ (unset LC_ALL; unset LANG; LC_CTYPE=unsupported_value chg \
   >    --config extensions.debugenv=$TESTTMP/debugenv.py debugenv)
   *cannot change locale* (glob) (?)
   LC_CTYPE=unsupported_value
   $ (unset LC_ALL; unset LANG; LC_CTYPE= chg \
   >    --config extensions.debugenv=$TESTTMP/debugenv.py debugenv)
   LC_CTYPE=
   $ LANG= LC_ALL= LC_CTYPE= chg \
   >    --config extensions.debugenv=$TESTTMP/debugenv.py debugenv
   LC_ALL=
   LC_CTYPE=
   LANG=
 
 Profiling isn't permanently enabled or carried over between chg invocations that
 share the same server
   $ cp $HGRCPATH.orig $HGRCPATH
   $ hg init $TESTTMP/profiling
   $ cd $TESTTMP/profiling
   $ filteredchg() {
-  >   CHGDEBUG=1 chg "$@" 2>&1 | egrep 'Sample count|start cmdserver' || true
+  >   CHGDEBUG=1 chg "$@" 2>&1 | sed -rn 's_^No samples recorded.*$_Sample count: 0_; /Sample count/p; /start cmdserver/p'
   > }
   $ newchg() {
   >   chg --kill-chg-daemon
   >   filteredchg "$@" | egrep -v 'start cmdserver' || true
   > }
 (--profile isn't permanently on just because it was specified when chg was
 started)
   $ newchg log -r . --profile
   Sample count: * (glob)
   $ filteredchg log -r .
 (enabling profiling via config works, even on the first chg command that starts
 a cmdserver)
   $ cat >> $HGRCPATH <<EOF
   > [profiling]
   > type=stat
   > enabled=1
   > EOF
   $ newchg log -r .
   Sample count: * (glob)
   $ filteredchg log -r .
   Sample count: * (glob)
 (test that we aren't accumulating more and more samples each run)
   $ cat > $TESTTMP/debugsleep.py <<EOF
   > import time
   > from mercurial import registrar
   > cmdtable = {}
   > command = registrar.command(cmdtable)
   > @command(b'debugsleep', [], b'', norepo=True)
   > def debugsleep(ui):
   >   start = time.time()
   >   x = 0
   >   while time.time() < start + 0.5:
   >     time.sleep(.1)
   >     x += 1
   >   ui.status(b'%d debugsleep iterations in %.03fs\n' % (x, time.time() - start))
   > EOF
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > debugsleep = $TESTTMP/debugsleep.py
   > EOF
   $ newchg debugsleep > run_1
   $ filteredchg debugsleep > run_2
   $ filteredchg debugsleep > run_3
   $ filteredchg debugsleep > run_4
 FIXME: Run 4 should not be >3x Run 1's number of samples.
   $ "$PYTHON" <<EOF
   > r1 = int(open("run_1", "r").read().split()[-1])
   > r4 = int(open("run_4", "r").read().split()[-1])
   > print("Run 1: %d samples\nRun 4: %d samples\nRun 4 > 3 * Run 1: %s" %
   >       (r1, r4, r4 > (r1 * 3)))
   > EOF
   Run 1: * samples (glob)
   Run 4: * samples (glob)
   Run 4 > 3 * Run 1: False
 (Disabling with --no-profile on the commandline still works, but isn't permanent)
   $ newchg log -r . --no-profile
   $ filteredchg log -r .
   Sample count: * (glob)
   $ filteredchg log -r . --no-profile
   $ filteredchg log -r .
   Sample count: * (glob)

tests/test-merge-halt.t

@@ -1,164 +1,220 b''
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > rebase=
   > [phases]
   > publish=False
   > [merge]
   > EOF
 
   $ hg init repo
   $ cd repo
   $ echo a > a
   $ echo b > b
   $ hg commit -qAm ab
   $ echo c >> a
   $ echo c >> b
   $ hg commit -qAm c
   $ hg up -q ".^"
   $ echo d >> a
   $ echo d >> b
   $ hg commit -qAm d
 
 Testing on-failure=continue
   $ echo on-failure=continue >> $HGRCPATH
   $ hg rebase -s 1 -d 2 --tool false
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
   merging a failed!
   merging b failed!
   unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
   [240]
 
   $ hg resolve --list
   U a
   U b
 
   $ hg rebase --abort
   rebase aborted
 
 Testing on-failure=halt
   $ echo on-failure=halt >> $HGRCPATH
   $ hg rebase -s 1 -d 2 --tool false
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
   merging a failed!
-  merge halted after failed merge (see hg resolve)
+  unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
   [240]
 
   $ hg resolve --list
   U a
   U b
 
   $ hg rebase --abort
   rebase aborted
 
 Testing on-failure=prompt
   $ cat <<EOS >> $HGRCPATH
   > [merge]
   > on-failure=prompt
   > [ui]
   > interactive=1
   > EOS
   $ cat <<EOS | hg rebase -s 1 -d 2 --tool false
   > y
   > n
   > EOS
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
   merging a failed!
   continue merge operation (yn)? y
   merging b failed!
   continue merge operation (yn)? n
-  merge halted after failed merge (see hg resolve)
+  unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
   [240]
 
   $ hg resolve --list
   U a
   U b
 
   $ hg rebase --abort
   rebase aborted
 
 Check that successful tool with failed post-check halts the merge
   $ cat <<EOS >> $HGRCPATH
   > [merge-tools]
   > true.check=changed
   > EOS
   $ cat <<EOS | hg rebase -s 1 -d 2 --tool true
   > y
   > n
   > n
   > EOS
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
    output file a appears unchanged
   was merge successful (yn)? y
    output file b appears unchanged
   was merge successful (yn)? n
   merging b failed!
   continue merge operation (yn)? n
-  merge halted after failed merge (see hg resolve)
+  unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
   [240]
 
   $ hg resolve --list
   R a
   U b
 
   $ hg rebase --abort
   rebase aborted
 
 Check that conflicts with conflict check also halts the merge
   $ cat <<EOS >> $HGRCPATH
   > [merge-tools]
   > true.check=conflicts
   > true.premerge=keep
   > [merge]
   > on-failure=halt
   > EOS
   $ hg rebase -s 1 -d 2 --tool true
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
   merging a failed!
-  merge halted after failed merge (see hg resolve)
+  unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
   [240]
 
   $ hg resolve --list
   U a
   U b
 
   $ hg rebase --abort
   rebase aborted
 
 Check that always-prompt also can halt the merge
   $ cat <<EOS | hg rebase -s 1 -d 2 --tool true --config merge-tools.true.check=prompt
   > y
   > n
   > EOS
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
   was merge of 'a' successful (yn)? y
   was merge of 'b' successful (yn)? n
   merging b failed!
-  merge halted after failed merge (see hg resolve)
+  unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
   [240]
 
   $ hg resolve --list
   R a
   U b
 
   $ hg rebase --abort
   rebase aborted
 
 Check that successful tool otherwise allows the merge to continue
   $ hg rebase -s 1 -d 2 --tool echo --keep --config merge-tools.echo.premerge=keep
   rebasing 1:1f28a51c3c9b "c"
   merging a
   merging b
   $TESTTMP/repo/a *a~base* *a~other* (glob)
   $TESTTMP/repo/b *b~base* *b~other* (glob)
+
+Check that unshelve isn't broken by halting the merge
+  $ cat <<EOS >> $HGRCPATH
+  > [extensions]
+  > shelve =
+  > [merge-tools]
+  > false.check=conflicts
+  > false.premerge=false
+  > EOS
+  $ echo foo > shelve_file1
+  $ echo foo > shelve_file2
+  $ hg ci -qAm foo
+  $ echo bar >> shelve_file1
+  $ echo bar >> shelve_file2
+  $ hg shelve --list
+  $ hg shelve
+  shelved as default
+  2 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ echo baz >> shelve_file1
+  $ echo baz >> shelve_file2
+  $ hg ci -m baz
+  $ hg unshelve --tool false --config merge-tools.false.premerge=keep
+  unshelving change 'default'
+  rebasing shelved changes
+  merging shelve_file1
+  merging shelve_file2
+  merging shelve_file1 failed!
+  unresolved conflicts (see 'hg resolve', then 'hg unshelve --continue')
+  [240]
+  $ hg status --config commands.status.verbose=True
+  M shelve_file1
+  M shelve_file2
+  ? shelve_file1.orig
+  ? shelve_file2.orig
+  # The repository is in an unfinished *unshelve* state.
+  
+  # Unresolved merge conflicts:
+  # 
+  #     shelve_file1
+  #     shelve_file2
+  # 
+  # To mark files as resolved:  hg resolve --mark FILE
+  
+  # To continue:    hg unshelve --continue
+  # To abort:       hg unshelve --abort
+  
+  $ hg resolve --tool false --all --re-merge
+  merging shelve_file1
+  merging shelve_file2
+  merging shelve_file1 failed!
+  merge halted after failed merge (see hg resolve)
+  [240]
+  $ hg shelve --list
+  default         (* ago)    changes to: foo (glob)
+  $ hg unshelve --abort
+  unshelve of 'default' aborted

tests/test-persistent-nodemap.t

@@ -1,1295 +1,1255 b''
 ===================================
 Test the persistent on-disk nodemap
 ===================================
 
 
   $ cat << EOF >> $HGRCPATH
   > [format]
   > use-share-safe=yes
   > [extensions]
   > share=
   > EOF
 
 #if no-rust
 
   $ cat << EOF >> $HGRCPATH
   > [format]
   > use-persistent-nodemap=yes
   > [devel]
   > persistent-nodemap=yes
   > EOF
 
 #endif
 
   $ hg init test-repo --config storage.revlog.persistent-nodemap.slow-path=allow
   $ cd test-repo
 
 Check handling of the default slow-path value
 
 #if no-pure no-rust
 
   $ hg id
   abort: accessing `persistent-nodemap` repository without associated fast implementation.
   (check `hg help config.format.use-persistent-nodemap` for details)
   [255]
 
 Unlock further check (we are here to test the feature)
 
   $ cat << EOF >> $HGRCPATH
   > [storage]
   > # to avoid spamming the test
   > revlog.persistent-nodemap.slow-path=allow
   > EOF
 
 #endif
 
 #if rust
 
 Regression test for a previous bug in Rust/C FFI for the `Revlog_CAPI` capsule:
 in places where `mercurial/cext/revlog.c` function signatures use `Py_ssize_t`
 (64 bits on Linux x86_64), corresponding declarations in `rust/hg-cpython/src/cindex.rs`
 incorrectly used `libc::c_int` (32 bits).
 As a result, -1 passed from Rust for the null revision became 4294967295 in C.
 
   $ hg log -r 00000000
   changeset:   -1:000000000000
   tag:         tip
   user:        
   date:        Thu Jan 01 00:00:00 1970 +0000
   
 
 #endif
 
 
   $ hg debugformat
   format-variant     repo
   fncache:            yes
   dirstate-v2:         no
   dotencode:          yes
   generaldelta:       yes
   share-safe:         yes
   sparserevlog:       yes
   persistent-nodemap: yes
   copies-sdc:          no
   revlog-v2:           no
   changelog-v2:        no
   plain-cl-delta:     yes
   compression:        zlib (no-zstd !)
   compression:        zstd (zstd !)
   compression-level:  default
   $ hg debugbuilddag .+5000 --new-file
 
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5000
   tip-node: 6b02b8c7b96654c25e86ba69eda198d7e6ad8b3c
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
   $ f --size .hg/store/00changelog.n
   .hg/store/00changelog.n: size=62
 
 Simple lookup works
 
   $ ANYNODE=`hg log --template '{node|short}\n' --rev tip`
   $ hg log -r "$ANYNODE" --template '{rev}\n'
   5000
 
 
 #if rust
 
   $ f --sha256 .hg/store/00changelog-*.nd
   .hg/store/00changelog-????????.nd: sha256=2e029d3200bd1a986b32784fc2ef1a3bd60dc331f025718bcf5ff44d93f026fd (glob)
 
   $ f --sha256 .hg/store/00manifest-*.nd
   .hg/store/00manifest-????????.nd: sha256=97117b1c064ea2f86664a124589e47db0e254e8d34739b5c5cc5bf31c9da2b51 (glob)
   $ hg debugnodemap --dump-new | f --sha256 --size
   size=121088, sha256=2e029d3200bd1a986b32784fc2ef1a3bd60dc331f025718bcf5ff44d93f026fd
   $ hg debugnodemap --dump-disk | f --sha256 --bytes=256 --hexdump --size
   size=121088, sha256=2e029d3200bd1a986b32784fc2ef1a3bd60dc331f025718bcf5ff44d93f026fd
   0000: 00 00 00 91 00 00 00 20 00 00 00 bb 00 00 00 e7 |....... ........|
   0010: 00 00 00 66 00 00 00 a1 00 00 01 13 00 00 01 22 |...f..........."|
   0020: 00 00 00 23 00 00 00 fc 00 00 00 ba 00 00 00 5e |...#...........^|
   0030: 00 00 00 df 00 00 01 4e 00 00 01 65 00 00 00 ab |.......N...e....|
   0040: 00 00 00 a9 00 00 00 95 00 00 00 73 00 00 00 38 |...........s...8|
   0050: 00 00 00 cc 00 00 00 92 00 00 00 90 00 00 00 69 |...............i|
   0060: 00 00 00 ec 00 00 00 8d 00 00 01 4f 00 00 00 12 |...........O....|
   0070: 00 00 02 0c 00 00 00 77 00 00 00 9c 00 00 00 8f |.......w........|
   0080: 00 00 00 d5 00 00 00 6b 00 00 00 48 00 00 00 b3 |.......k...H....|
   0090: 00 00 00 e5 00 00 00 b5 00 00 00 8e 00 00 00 ad |................|
   00a0: 00 00 00 7b 00 00 00 7c 00 00 00 0b 00 00 00 2b |...{...|.......+|
   00b0: 00 00 00 c6 00 00 00 1e 00 00 01 08 00 00 00 11 |................|
   00c0: 00 00 01 30 00 00 00 26 00 00 01 9c 00 00 00 35 |...0...&.......5|
   00d0: 00 00 00 b8 00 00 01 31 00 00 00 2c 00 00 00 55 |.......1...,...U|
   00e0: 00 00 00 8a 00 00 00 9a 00 00 00 0c 00 00 01 1e |................|
   00f0: 00 00 00 a4 00 00 00 83 00 00 00 c9 00 00 00 8c |................|
 
 
 #else
 
   $ f --sha256 .hg/store/00changelog-*.nd
   .hg/store/00changelog-????????.nd: sha256=f544f5462ff46097432caf6d764091f6d8c46d6121be315ead8576d548c9dd79 (glob)
   $ hg debugnodemap --dump-new | f --sha256 --size
   size=121088, sha256=f544f5462ff46097432caf6d764091f6d8c46d6121be315ead8576d548c9dd79
   $ hg debugnodemap --dump-disk | f --sha256 --bytes=256 --hexdump --size
   size=121088, sha256=f544f5462ff46097432caf6d764091f6d8c46d6121be315ead8576d548c9dd79
   0000: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   0010: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   0020: ff ff ff ff ff ff f5 06 ff ff ff ff ff ff f3 e7 |................|
   0030: ff ff ef ca ff ff ff ff ff ff ff ff ff ff ff ff |................|
   0040: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   0050: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ed 08 |................|
   0060: ff ff ed 66 ff ff ff ff ff ff ff ff ff ff ff ff |...f............|
   0070: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   0080: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   0090: ff ff ff ff ff ff ff ff ff ff ff ff ff ff f6 ed |................|
   00a0: ff ff ff ff ff ff fe 61 ff ff ff ff ff ff ff ff |.......a........|
   00b0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   00c0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   00d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
   00e0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff f1 02 |................|
   00f0: ff ff ff ff ff ff ed 1b ff ff ff ff ff ff ff ff |................|
 
 #endif
 
   $ hg debugnodemap --check
   revision in index:   5001
   revision in nodemap: 5001
 
 add a new commit
 
   $ hg up
   5001 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ echo foo > foo
   $ hg add foo
 
 
 Check slow-path config value handling
 -------------------------------------
 
 #if no-pure no-rust
 
   $ hg id --config "storage.revlog.persistent-nodemap.slow-path=invalid-value"
   unknown value for config "storage.revlog.persistent-nodemap.slow-path": "invalid-value"
   falling back to default value: abort
   abort: accessing `persistent-nodemap` repository without associated fast implementation.
   (check `hg help config.format.use-persistent-nodemap` for details)
   [255]
 
   $ hg log -r . --config "storage.revlog.persistent-nodemap.slow-path=warn"
   warning: accessing `persistent-nodemap` repository without associated fast implementation.
   (check `hg help config.format.use-persistent-nodemap` for details)
   changeset:   5000:6b02b8c7b966
   tag:         tip
   user:        debugbuilddag
   date:        Thu Jan 01 01:23:20 1970 +0000
   summary:     r5000
   
   $ hg ci -m 'foo' --config "storage.revlog.persistent-nodemap.slow-path=abort"
   abort: accessing `persistent-nodemap` repository without associated fast implementation.
   (check `hg help config.format.use-persistent-nodemap` for details)
   [255]
 
 #else
 
   $ hg id --config "storage.revlog.persistent-nodemap.slow-path=invalid-value"
   unknown value for config "storage.revlog.persistent-nodemap.slow-path": "invalid-value"
   falling back to default value: abort
   6b02b8c7b966+ tip
 
 #endif
 
   $ hg ci -m 'foo'
 
 #if no-pure no-rust
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5001
   tip-node: 16395c3cf7e231394735e6b1717823ada303fb0c
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 #else
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5001
   tip-node: 16395c3cf7e231394735e6b1717823ada303fb0c
   data-length: 121344
   data-unused: 256
   data-unused: 0.211%
 #endif
 
   $ f --size .hg/store/00changelog.n
   .hg/store/00changelog.n: size=62
 
 (The pure code use the debug code that perform incremental update, the C code reencode from scratch)
 
 #if pure
   $ f --sha256 .hg/store/00changelog-*.nd --size
   .hg/store/00changelog-????????.nd: size=121344, sha256=cce54c5da5bde3ad72a4938673ed4064c86231b9c64376b082b163fdb20f8f66 (glob)
 #endif
 
 #if rust
   $ f --sha256 .hg/store/00changelog-*.nd --size
   .hg/store/00changelog-????????.nd: size=121344, sha256=952b042fcf614ceb37b542b1b723e04f18f83efe99bee4e0f5ccd232ef470e58 (glob)
 #endif
 
 #if no-pure no-rust
   $ f --sha256 .hg/store/00changelog-*.nd --size
   .hg/store/00changelog-????????.nd: size=121088, sha256=df7c06a035b96cb28c7287d349d603baef43240be7736fe34eea419a49702e17 (glob)
 #endif
 
   $ hg debugnodemap --check
   revision in index:   5002
   revision in nodemap: 5002
 
 Test code path without mmap
 ---------------------------
 
   $ echo bar > bar
   $ hg add bar
   $ hg ci -m 'bar' --config storage.revlog.persistent-nodemap.mmap=no
 
   $ hg debugnodemap --check --config storage.revlog.persistent-nodemap.mmap=yes
   revision in index:   5003
   revision in nodemap: 5003
   $ hg debugnodemap --check --config storage.revlog.persistent-nodemap.mmap=no
   revision in index:   5003
   revision in nodemap: 5003
 
 
 #if pure
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121600
   data-unused: 512
   data-unused: 0.421%
   $ f --sha256 .hg/store/00changelog-*.nd --size
   .hg/store/00changelog-????????.nd: size=121600, sha256=def52503d049ccb823974af313a98a935319ba61f40f3aa06a8be4d35c215054 (glob)
 #endif
 #if rust
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121600
   data-unused: 512
   data-unused: 0.421%
   $ f --sha256 .hg/store/00changelog-*.nd --size
   .hg/store/00changelog-????????.nd: size=121600, sha256=dacf5b5f1d4585fee7527d0e67cad5b1ba0930e6a0928f650f779aefb04ce3fb (glob)
 #endif
 #if no-pure no-rust
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
   $ f --sha256 .hg/store/00changelog-*.nd --size
   .hg/store/00changelog-????????.nd: size=121088, sha256=59fcede3e3cc587755916ceed29e3c33748cd1aa7d2f91828ac83e7979d935e8 (glob)
 #endif
 
 Test force warming the cache
 
   $ rm .hg/store/00changelog.n
   $ hg debugnodemap --metadata
   $ hg debugupdatecache
 #if pure
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 #else
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 #endif
 
 Check out of sync nodemap
 =========================
 
 First copy old data on the side.
 
   $ mkdir ../tmp-copies
   $ cp .hg/store/00changelog-????????.nd .hg/store/00changelog.n ../tmp-copies
 
 Nodemap lagging behind
 ----------------------
 
 make a new commit
 
   $ echo bar2 > bar
   $ hg ci -m 'bar2'
   $ NODE=`hg log -r tip -T '{node}\n'`
   $ hg log -r "$NODE" -T '{rev}\n'
   5003
 
 If the nodemap is lagging behind, it can catch up fine
 
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5003
   tip-node: c9329770f979ade2d16912267c38ba5f82fd37b3
   data-length: 121344 (pure !)
   data-length: 121344 (rust !)
   data-length: 121152 (no-rust no-pure !)
   data-unused: 192 (pure !)
   data-unused: 192 (rust !)
   data-unused: 0 (no-rust no-pure !)
   data-unused: 0.158% (pure !)
   data-unused: 0.158% (rust !)
   data-unused: 0.000% (no-rust no-pure !)
   $ cp -f ../tmp-copies/* .hg/store/
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
   $ hg log -r "$NODE" -T '{rev}\n'
   5003
 
 changelog altered
 -----------------
 
 If the nodemap is not gated behind a requirements, an unaware client can alter
 the repository so the revlog used to generate the nodemap is not longer
 compatible with the persistent nodemap. We need to detect that.
 
   $ hg up "$NODE~5"
   0 files updated, 0 files merged, 4 files removed, 0 files unresolved
   $ echo bar > babar
   $ hg add babar
   $ hg ci -m 'babar'
   created new head
   $ OTHERNODE=`hg log -r tip -T '{node}\n'`
   $ hg log -r "$OTHERNODE" -T '{rev}\n'
   5004
 
   $ hg --config extensions.strip= strip --rev "$NODE~1" --no-backup
 
 the nodemap should detect the changelog have been tampered with and recover.
 
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: b355ef8adce0949b8bdf6afc72ca853740d65944
   data-length: 121536 (pure !)
   data-length: 121088 (rust !)
   data-length: 121088 (no-pure no-rust !)
   data-unused: 448 (pure !)
   data-unused: 0 (rust !)
   data-unused: 0 (no-pure no-rust !)
   data-unused: 0.000% (rust !)
   data-unused: 0.369% (pure !)
   data-unused: 0.000% (no-pure no-rust !)
 
   $ cp -f ../tmp-copies/* .hg/store/
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5002
   tip-node: 880b18d239dfa9f632413a2071bfdbcc4806a4fd
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
   $ hg log -r "$OTHERNODE" -T '{rev}\n'
   5002
 
 missing data file
 -----------------
 
   $ UUID=`hg debugnodemap --metadata| grep 'uid:' | \
   > sed 's/uid: //'`
   $ FILE=.hg/store/00changelog-"${UUID}".nd
   $ mv $FILE ../tmp-data-file
   $ cp .hg/store/00changelog.n ../tmp-docket
 
 mercurial don't crash
 
   $ hg log -r .
   changeset:   5002:b355ef8adce0
   tag:         tip
   parent:      4998:d918ad6d18d3
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     babar
   
   $ hg debugnodemap --metadata
 
   $ hg debugupdatecache
   $ hg debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5002
   tip-node: b355ef8adce0949b8bdf6afc72ca853740d65944
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
-
-Sub-case: fallback for corrupted data file
-------------------------------------------
-
-Sabotaging the data file so that nodemap resolutions fail, triggering fallback to
-(non-persistent) C implementation.
-
-
-  $ UUID=`hg debugnodemap --metadata| grep 'uid:' | \
-  > sed 's/uid: //'`
-  $ FILE=.hg/store/00changelog-"${UUID}".nd
-  $ python -c "fobj = open('$FILE', 'r+b'); fobj.write(b'\xff' * 121088); fobj.close()"
-
-The nodemap data file is still considered in sync with the docket. This
-would fail without the fallback to the (non-persistent) C implementation:
-
-  $ hg log -r b355ef8adce0949b8bdf6afc72ca853740d65944 -T '{rev}\n' --traceback
-  5002
-
-The nodemap data file hasn't been fixed, more tests can be inserted:
-
-  $ hg debugnodemap --dump-disk | f --bytes=256 --hexdump --size
-  size=121088
-  0000: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0010: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0020: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0030: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0040: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0050: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0060: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0070: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0080: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  0090: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  00a0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  00b0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  00c0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  00d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  00e0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-  00f0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................|
-
   $ mv ../tmp-data-file $FILE
   $ mv ../tmp-docket .hg/store/00changelog.n
 
 Check transaction related property
 ==================================
 
 An up to date nodemap should be available to shell hooks,
 
   $ echo dsljfl > a
   $ hg add a
   $ hg ci -m a
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5003
   tip-node: a52c5079765b5865d97b993b303a18740113bbb2
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
   $ echo babar2 > babar
   $ hg ci -m 'babar2' --config "hooks.pretxnclose.nodemap-test=hg debugnodemap --metadata"
   uid: ???????? (glob)
   tip-rev: 5004
   tip-node: 2f5fb1c06a16834c5679d672e90da7c5f3b1a984
   data-length: 121280 (pure !)
   data-length: 121280 (rust !)
   data-length: 121088 (no-pure no-rust !)
   data-unused: 192 (pure !)
   data-unused: 192 (rust !)
   data-unused: 0 (no-pure no-rust !)
   data-unused: 0.158% (pure !)
   data-unused: 0.158% (rust !)
   data-unused: 0.000% (no-pure no-rust !)
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5004
   tip-node: 2f5fb1c06a16834c5679d672e90da7c5f3b1a984
   data-length: 121280 (pure !)
   data-length: 121280 (rust !)
   data-length: 121088 (no-pure no-rust !)
   data-unused: 192 (pure !)
   data-unused: 192 (rust !)
   data-unused: 0 (no-pure no-rust !)
   data-unused: 0.158% (pure !)
   data-unused: 0.158% (rust !)
   data-unused: 0.000% (no-pure no-rust !)
 
 Another process does not see the pending nodemap content during run.
 
   $ echo qpoasp > a
   $ hg ci -m a2 \
   > --config "hooks.pretxnclose=sh \"$RUNTESTDIR/testlib/wait-on-file\" 20 sync-repo-read sync-txn-pending" \
   > --config "hooks.txnclose=touch sync-txn-close" > output.txt 2>&1 &
 
 (read the repository while the commit transaction is pending)
 
   $ sh "$RUNTESTDIR/testlib/wait-on-file" 20 sync-txn-pending && \
   > hg debugnodemap --metadata && \
   > sh "$RUNTESTDIR/testlib/wait-on-file" 20 sync-txn-close sync-repo-read
   uid: ???????? (glob)
   tip-rev: 5004
   tip-node: 2f5fb1c06a16834c5679d672e90da7c5f3b1a984
   data-length: 121280 (pure !)
   data-length: 121280 (rust !)
   data-length: 121088 (no-pure no-rust !)
   data-unused: 192 (pure !)
   data-unused: 192 (rust !)
   data-unused: 0 (no-pure no-rust !)
   data-unused: 0.158% (pure !)
   data-unused: 0.158% (rust !)
   data-unused: 0.000% (no-pure no-rust !)
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121536 (pure !)
   data-length: 121536 (rust !)
   data-length: 121088 (no-pure no-rust !)
   data-unused: 448 (pure !)
   data-unused: 448 (rust !)
   data-unused: 0 (no-pure no-rust !)
   data-unused: 0.369% (pure !)
   data-unused: 0.369% (rust !)
   data-unused: 0.000% (no-pure no-rust !)
 
   $ cat output.txt
 
 Check that a failing transaction will properly revert the data
 
   $ echo plakfe > a
   $ f --size --sha256 .hg/store/00changelog-*.nd
   .hg/store/00changelog-????????.nd: size=121536, sha256=bb414468d225cf52d69132e1237afba34d4346ee2eb81b505027e6197b107f03 (glob) (pure !)
   .hg/store/00changelog-????????.nd: size=121536, sha256=909ac727bc4d1c0fda5f7bff3c620c98bd4a2967c143405a1503439e33b377da (glob) (rust !)
   .hg/store/00changelog-????????.nd: size=121088, sha256=342d36d30d86dde67d3cb6c002606c4a75bcad665595d941493845066d9c8ee0 (glob) (no-pure no-rust !)
   $ hg ci -m a3 --config "extensions.abort=$RUNTESTDIR/testlib/crash_transaction_late.py"
   transaction abort!
   rollback completed
   abort: This is a late abort
   [255]
   $ hg debugnodemap --metadata
   uid: ???????? (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121536 (pure !)
   data-length: 121536 (rust !)
   data-length: 121088 (no-pure no-rust !)
   data-unused: 448 (pure !)
   data-unused: 448 (rust !)
   data-unused: 0 (no-pure no-rust !)
   data-unused: 0.369% (pure !)
   data-unused: 0.369% (rust !)
   data-unused: 0.000% (no-pure no-rust !)
   $ f --size --sha256 .hg/store/00changelog-*.nd
   .hg/store/00changelog-????????.nd: size=121536, sha256=bb414468d225cf52d69132e1237afba34d4346ee2eb81b505027e6197b107f03 (glob) (pure !)
   .hg/store/00changelog-????????.nd: size=121536, sha256=909ac727bc4d1c0fda5f7bff3c620c98bd4a2967c143405a1503439e33b377da (glob) (rust !)
   .hg/store/00changelog-????????.nd: size=121088, sha256=342d36d30d86dde67d3cb6c002606c4a75bcad665595d941493845066d9c8ee0 (glob) (no-pure no-rust !)
 
 Check that removing content does not confuse the nodemap
 --------------------------------------------------------
 
 removing data with rollback
 
   $ echo aso > a
   $ hg ci -m a4
   $ hg rollback
   repository tip rolled back to revision 5005 (undo commit)
   working directory now based on revision 5005
   $ hg id -r .
   90d5d3ba2fc4 tip
 
 removing data with strip
 
   $ echo aso > a
   $ hg ci -m a4
   $ hg --config extensions.strip= strip -r . --no-backup
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg id -r . --traceback
   90d5d3ba2fc4 tip
 
 (be a good citizen and regenerate the nodemap)
   $ hg debugupdatecaches
   $ hg debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 Check race condition when multiple process write new data to the repository
 ---------------------------------------------------------------------------
 
 In this test, we check that two writers touching the repositories will not
 overwrite each other data. This test is prompted by the existent of issue6554.
 Where a writer ended up using and outdated docket to update the repository. See
 the dedicated extension for details on the race windows and read/write schedule
 necessary to end up in this situation: testlib/persistent-nodemap-race-ext.py
 
 The issue was initially observed on a server with a high push trafic, but it
 can be reproduced using a share and two commiting process which seems simpler.
 
 The test is Rust only as the other implementation does not use the same
 read/write patterns.
 
   $ cd ..
 
 #if rust
 
   $ cp -R test-repo race-repo
   $ hg share race-repo ./other-wc --config format.use-share-safe=yes
   updating working directory
   5001 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg debugformat -R ./race-repo | egrep 'share-safe|persistent-nodemap'
   share-safe:         yes
   persistent-nodemap: yes
   $ hg debugformat -R ./other-wc/ | egrep 'share-safe|persistent-nodemap'
   share-safe:         yes
   persistent-nodemap: yes
   $ hg -R ./other-wc update 'min(head())'
   3 files updated, 0 files merged, 2 files removed, 0 files unresolved
   $ hg -R ./race-repo debugnodemap --metadata
   uid: 43c37dde
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
   $ hg -R ./race-repo log -G -r 'head()'
   @  changeset:   5005:90d5d3ba2fc4
   |  tag:         tip
   ~  user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     a2
   
   o  changeset:   5001:16395c3cf7e2
   |  user:        test
   ~  date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     foo
   
   $ hg -R ./other-wc log -G -r 'head()'
   o  changeset:   5005:90d5d3ba2fc4
   |  tag:         tip
   ~  user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     a2
   
   @  changeset:   5001:16395c3cf7e2
   |  user:        test
   ~  date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     foo
   
   $ echo left-side-race > race-repo/left-side-race
   $ hg -R ./race-repo/ add race-repo/left-side-race
 
   $ echo right-side-race > ./other-wc/right-side-race
   $ hg -R ./other-wc/ add ./other-wc/right-side-race
 
   $ mkdir sync-files
   $ mkdir outputs
   $ (
   >    hg -R ./race-repo/ commit -m left-side-commit \
   >    --config "extensions.race=${RUNTESTDIR}/testlib/persistent-nodemap-race-ext.py" \
   >    --config 'devel.nodemap-race.role=left';
   >    touch sync-files/left-done
   > ) > outputs/left.txt 2>&1 &
   $ (
   >    hg -R ./other-wc/ commit -m right-side-commit \
   >    --config "extensions.race=${RUNTESTDIR}/testlib/persistent-nodemap-race-ext.py" \
   >    --config 'devel.nodemap-race.role=right';
   >    touch sync-files/right-done
   > ) > outputs/right.txt 2>&1 &
   $ (
   >    hg -R ./race-repo/ check-nodemap-race \
   >    --config "extensions.race=${RUNTESTDIR}/testlib/persistent-nodemap-race-ext.py" \
   >    --config 'devel.nodemap-race.role=reader';
   >    touch sync-files/reader-done
   > ) > outputs/reader.txt 2>&1 &
   $ sh "$RUNTESTDIR"/testlib/wait-on-file 10 sync-files/left-done
   $ cat outputs/left.txt
   docket-details:
     uid:         43c37dde
     actual-tip:  5005
     tip-rev:     5005
     data-length: 121088
   nodemap-race: left side locked and ready to commit
   docket-details:
     uid:         43c37dde
     actual-tip:  5005
     tip-rev:     5005
     data-length: 121088
   finalized changelog write
   persisting changelog nodemap
     new data start at 121088
   persisted changelog nodemap
   docket-details:
     uid:         43c37dde
     actual-tip:  5006
     tip-rev:     5006
     data-length: 121280
   $ sh "$RUNTESTDIR"/testlib/wait-on-file 10 sync-files/right-done
   $ cat outputs/right.txt
   nodemap-race: right side start of the locking sequence
   nodemap-race: right side reading changelog
   nodemap-race: right side reading of changelog is done
   docket-details:
     uid:         43c37dde
     actual-tip:  5006
     tip-rev:     5005
     data-length: 121088
   nodemap-race: right side ready to wait for the lock
   nodemap-race: right side locked and ready to commit
   docket-details:
     uid:         43c37dde
     actual-tip:  5006
     tip-rev:     5006
     data-length: 121280
   right ready to write, waiting for reader
   right proceeding with writing its changelog index and nodemap
   finalized changelog write
   persisting changelog nodemap
     new data start at 121280
   persisted changelog nodemap
   docket-details:
     uid:         43c37dde
     actual-tip:  5007
     tip-rev:     5007
     data-length: 121536
   $ sh "$RUNTESTDIR"/testlib/wait-on-file 10 sync-files/reader-done
   $ cat outputs/reader.txt
   reader: reading changelog
   reader ready to read the changelog, waiting for right
   reader: nodemap docket read
   record-data-length: 121280
   actual-data-length: 121280
   file-actual-length: 121536
   reader: changelog read
   docket-details:
     uid:         43c37dde
     actual-tip:  5006
     tip-rev:     5006
     data-length: 121280
   tip-rev:  5006
   tip-node: 492901161367
   node-rev: 5006
 
   $ hg -R ./race-repo log -G -r 'head()'
   o  changeset:   5007:ac4a2abde241
   |  tag:         tip
   ~  parent:      5001:16395c3cf7e2
      user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     right-side-commit
   
   @  changeset:   5006:492901161367
   |  user:        test
   ~  date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     left-side-commit
   
   $ hg -R ./other-wc log -G -r 'head()'
   @  changeset:   5007:ac4a2abde241
   |  tag:         tip
   ~  parent:      5001:16395c3cf7e2
      user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     right-side-commit
   
   o  changeset:   5006:492901161367
   |  user:        test
   ~  date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     left-side-commit
   
 #endif
 
 Test upgrade / downgrade
 ========================
 
   $ cd ./test-repo/
 
 downgrading
 
   $ cat << EOF >> .hg/hgrc
   > [format]
   > use-persistent-nodemap=no
   > EOF
   $ hg debugformat -v
   format-variant     repo config default
   fncache:            yes    yes     yes
   dirstate-v2:         no     no      no
   dotencode:          yes    yes     yes
   generaldelta:       yes    yes     yes
   share-safe:         yes    yes      no
   sparserevlog:       yes    yes     yes
   persistent-nodemap: yes     no      no
   copies-sdc:          no     no      no
   revlog-v2:           no     no      no
   changelog-v2:        no     no      no
   plain-cl-delta:     yes    yes     yes
   compression:        zlib   zlib    zlib (no-zstd !)
   compression:        zstd   zstd    zstd (zstd !)
   compression-level:  default default default
   $ hg debugupgraderepo --run --no-backup --quiet
   upgrade will perform the following actions:
   
   requirements
      preserved: dotencode, fncache, generaldelta, revlogv1, share-safe, sparserevlog, store (no-zstd no-dirstate-v2 !)
      preserved: dotencode, fncache, generaldelta, revlog-compression-zstd, revlogv1, share-safe, sparserevlog, store (zstd no-dirstate-v2 !)
      preserved: dotencode, exp-dirstate-v2, fncache, generaldelta, revlog-compression-zstd, revlogv1, share-safe, sparserevlog, store (zstd dirstate-v2 !)
      removed: persistent-nodemap
   
   processed revlogs:
     - all-filelogs
     - changelog
     - manifest
   
   $ ls -1 .hg/store/ | egrep '00(changelog|manifest)(\.n|-.*\.nd)'
   00changelog-*.nd (glob)
   00manifest-*.nd (glob)
   undo.backup.00changelog.n
   undo.backup.00manifest.n
   $ hg debugnodemap --metadata
 
 
 upgrading
 
   $ cat << EOF >> .hg/hgrc
   > [format]
   > use-persistent-nodemap=yes
   > EOF
   $ hg debugformat -v
   format-variant     repo config default
   fncache:            yes    yes     yes
   dirstate-v2:         no     no      no
   dotencode:          yes    yes     yes
   generaldelta:       yes    yes     yes
   share-safe:         yes    yes      no
   sparserevlog:       yes    yes     yes
   persistent-nodemap:  no    yes      no
   copies-sdc:          no     no      no
   revlog-v2:           no     no      no
   changelog-v2:        no     no      no
   plain-cl-delta:     yes    yes     yes
   compression:        zlib   zlib    zlib (no-zstd !)
   compression:        zstd   zstd    zstd (zstd !)
   compression-level:  default default default
   $ hg debugupgraderepo --run --no-backup --quiet
   upgrade will perform the following actions:
   
   requirements
      preserved: dotencode, fncache, generaldelta, revlogv1, share-safe, sparserevlog, store (no-zstd no-dirstate-v2 !)
      preserved: dotencode, fncache, generaldelta, revlog-compression-zstd, revlogv1, share-safe, sparserevlog, store (zstd no-dirstate-v2 !)
      preserved: dotencode, exp-dirstate-v2, fncache, generaldelta, revlog-compression-zstd, revlogv1, share-safe, sparserevlog, store (zstd dirstate-v2 !)
      added: persistent-nodemap
   
   processed revlogs:
     - all-filelogs
     - changelog
     - manifest
   
   $ ls -1 .hg/store/ | egrep '00(changelog|manifest)(\.n|-.*\.nd)'
   00changelog-*.nd (glob)
   00changelog.n
   00manifest-*.nd (glob)
   00manifest.n
   undo.backup.00changelog.n
   undo.backup.00manifest.n
 
   $ hg debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 Running unrelated upgrade
 
   $ hg debugupgraderepo --run --no-backup --quiet --optimize re-delta-all
   upgrade will perform the following actions:
   
   requirements
      preserved: dotencode, fncache, generaldelta, persistent-nodemap, revlogv1, share-safe, sparserevlog, store (no-zstd no-dirstate-v2 !)
      preserved: dotencode, fncache, generaldelta, persistent-nodemap, revlog-compression-zstd, revlogv1, share-safe, sparserevlog, store (zstd no-dirstate-v2 !)
      preserved: dotencode, exp-dirstate-v2, fncache, generaldelta, persistent-nodemap, revlog-compression-zstd, revlogv1, share-safe, sparserevlog, store (zstd dirstate-v2 !)
   
   optimisations: re-delta-all
   
   processed revlogs:
     - all-filelogs
     - changelog
     - manifest
   
   $ ls -1 .hg/store/ | egrep '00(changelog|manifest)(\.n|-.*\.nd)'
   00changelog-*.nd (glob)
   00changelog.n
   00manifest-*.nd (glob)
   00manifest.n
 
   $ hg debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 Persistent nodemap and local/streaming clone
 ============================================
 
   $ cd ..
 
 standard clone
 --------------
 
 The persistent nodemap should exist after a streaming clone
 
   $ hg clone --pull --quiet -U test-repo standard-clone
   $ ls -1 standard-clone/.hg/store/ | egrep '00(changelog|manifest)(\.n|-.*\.nd)'
   00changelog-*.nd (glob)
   00changelog.n
   00manifest-*.nd (glob)
   00manifest.n
   $ hg -R standard-clone debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 
 local clone
 ------------
 
 The persistent nodemap should exist after a streaming clone
 
   $ hg clone -U test-repo local-clone
   $ ls -1 local-clone/.hg/store/ | egrep '00(changelog|manifest)(\.n|-.*\.nd)'
   00changelog-*.nd (glob)
   00changelog.n
   00manifest-*.nd (glob)
   00manifest.n
   $ hg -R local-clone debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 Test various corruption case
 ============================
 
 Missing datafile
 ----------------
 
 Test behavior with a missing datafile
 
   $ hg clone --quiet --pull test-repo corruption-test-repo
   $ ls -1 corruption-test-repo/.hg/store/00changelog*
   corruption-test-repo/.hg/store/00changelog-*.nd (glob)
   corruption-test-repo/.hg/store/00changelog.d
   corruption-test-repo/.hg/store/00changelog.i
   corruption-test-repo/.hg/store/00changelog.n
   $ rm corruption-test-repo/.hg/store/00changelog*.nd
   $ hg log -R corruption-test-repo -r .
   changeset:   5005:90d5d3ba2fc4
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     a2
   
   $ ls -1 corruption-test-repo/.hg/store/00changelog*
   corruption-test-repo/.hg/store/00changelog.d
   corruption-test-repo/.hg/store/00changelog.i
   corruption-test-repo/.hg/store/00changelog.n
 
 Truncated data file
 -------------------
 
 Test behavior with a too short datafile
 
 rebuild the missing data
   $ hg -R corruption-test-repo debugupdatecache
   $ ls -1 corruption-test-repo/.hg/store/00changelog*
   corruption-test-repo/.hg/store/00changelog-*.nd (glob)
   corruption-test-repo/.hg/store/00changelog.d
   corruption-test-repo/.hg/store/00changelog.i
   corruption-test-repo/.hg/store/00changelog.n
 
 truncate the file
 
   $ datafilepath=`ls corruption-test-repo/.hg/store/00changelog*.nd`
   $ f -s $datafilepath
   corruption-test-repo/.hg/store/00changelog-*.nd: size=121088 (glob)
   $ dd if=$datafilepath bs=1000 count=10 of=$datafilepath-tmp status=noxfer
   10+0 records in
   10+0 records out
   $ mv $datafilepath-tmp $datafilepath
   $ f -s $datafilepath
   corruption-test-repo/.hg/store/00changelog-*.nd: size=10000 (glob)
 
 Check that Mercurial reaction to this event
 
   $ hg -R corruption-test-repo log -r . --traceback
   changeset:   5005:90d5d3ba2fc4
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     a2
   
 
 
 stream clone
 ============
 
 The persistent nodemap should exist after a streaming clone
 
 Simple case
 -----------
 
 No race condition
 
   $ hg clone -U --stream ssh://user@dummy/test-repo stream-clone --debug | egrep '00(changelog|manifest)'
   adding [s] 00manifest.n (62 bytes)
   adding [s] 00manifest-*.nd (118 KB) (glob)
   adding [s] 00changelog.n (62 bytes)
   adding [s] 00changelog-*.nd (118 KB) (glob)
   adding [s] 00manifest.d (452 KB) (no-zstd !)
   adding [s] 00manifest.d (491 KB) (zstd !)
   adding [s] 00changelog.d (360 KB) (no-zstd !)
   adding [s] 00changelog.d (368 KB) (zstd !)
   adding [s] 00manifest.i (313 KB)
   adding [s] 00changelog.i (313 KB)
   $ ls -1 stream-clone/.hg/store/ | egrep '00(changelog|manifest)(\.n|-.*\.nd)'
   00changelog-*.nd (glob)
   00changelog.n
   00manifest-*.nd (glob)
   00manifest.n
   $ hg -R stream-clone debugnodemap --metadata
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 new data appened
 -----------------
 
 Other commit happening on the server during the stream clone
 
 setup the step-by-step stream cloning
 
   $ HG_TEST_STREAM_WALKED_FILE_1="$TESTTMP/sync_file_walked_1"
   $ export HG_TEST_STREAM_WALKED_FILE_1
   $ HG_TEST_STREAM_WALKED_FILE_2="$TESTTMP/sync_file_walked_2"
   $ export HG_TEST_STREAM_WALKED_FILE_2
   $ HG_TEST_STREAM_WALKED_FILE_3="$TESTTMP/sync_file_walked_3"
   $ export HG_TEST_STREAM_WALKED_FILE_3
   $ cat << EOF >> test-repo/.hg/hgrc
   > [extensions]
   > steps=$RUNTESTDIR/testlib/ext-stream-clone-steps.py
   > EOF
 
 Check and record file state beforehand
 
   $ f --size test-repo/.hg/store/00changelog*
   test-repo/.hg/store/00changelog-*.nd: size=121088 (glob)
   test-repo/.hg/store/00changelog.d: size=376891 (zstd !)
   test-repo/.hg/store/00changelog.d: size=368890 (no-zstd !)
   test-repo/.hg/store/00changelog.i: size=320384
   test-repo/.hg/store/00changelog.n: size=62
   $ hg -R test-repo debugnodemap --metadata | tee server-metadata.txt
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 Prepare a commit
 
   $ echo foo >> test-repo/foo
   $ hg -R test-repo/ add test-repo/foo
 
 Do a mix of clone and commit at the same time so that the file listed on disk differ at actual transfer time.
 
   $ (hg clone -U --stream ssh://user@dummy/test-repo stream-clone-race-1 --debug 2>> clone-output | egrep '00(changelog|manifest)' >> clone-output; touch $HG_TEST_STREAM_WALKED_FILE_3) &
   $ $RUNTESTDIR/testlib/wait-on-file 10 $HG_TEST_STREAM_WALKED_FILE_1
   $ hg -R test-repo/ commit -m foo
   $ touch $HG_TEST_STREAM_WALKED_FILE_2
   $ $RUNTESTDIR/testlib/wait-on-file 10 $HG_TEST_STREAM_WALKED_FILE_3
   $ cat clone-output
   adding [s] 00manifest.n (62 bytes)
   adding [s] 00manifest-*.nd (118 KB) (glob)
   adding [s] 00changelog.n (62 bytes)
   adding [s] 00changelog-*.nd (118 KB) (glob)
   adding [s] 00manifest.d (452 KB) (no-zstd !)
   adding [s] 00manifest.d (491 KB) (zstd !)
   adding [s] 00changelog.d (360 KB) (no-zstd !)
   adding [s] 00changelog.d (368 KB) (zstd !)
   adding [s] 00manifest.i (313 KB)
   adding [s] 00changelog.i (313 KB)
 
 Check the result state
 
   $ f --size stream-clone-race-1/.hg/store/00changelog*
   stream-clone-race-1/.hg/store/00changelog-*.nd: size=121088 (glob)
   stream-clone-race-1/.hg/store/00changelog.d: size=368890 (no-zstd !)
   stream-clone-race-1/.hg/store/00changelog.d: size=376891 (zstd !)
   stream-clone-race-1/.hg/store/00changelog.i: size=320384
   stream-clone-race-1/.hg/store/00changelog.n: size=62
 
   $ hg -R stream-clone-race-1 debugnodemap --metadata | tee client-metadata.txt
   uid: * (glob)
   tip-rev: 5005
   tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
   data-length: 121088
   data-unused: 0
   data-unused: 0.000%
 
 We get a usable nodemap, so no rewrite would be needed and the metadata should be identical
 (ie: the following diff should be empty)
 
 This isn't the case for the `no-rust` `no-pure` implementation as it use a very minimal nodemap implementation that unconditionnaly rewrite the nodemap "all the time".
 
 #if no-rust no-pure
   $ diff -u server-metadata.txt client-metadata.txt
   --- server-metadata.txt	* (glob)
   +++ client-metadata.txt	* (glob)
   @@ -1,4 +1,4 @@
   -uid: * (glob)
   +uid: * (glob)
    tip-rev: 5005
    tip-node: 90d5d3ba2fc47db50f712570487cb261a68c8ffe
    data-length: 121088
   [1]
 #else
   $ diff -u server-metadata.txt client-metadata.txt
 #endif
 
 
 Clean up after the test.
 
   $ rm -f "$HG_TEST_STREAM_WALKED_FILE_1"
   $ rm -f "$HG_TEST_STREAM_WALKED_FILE_2"
   $ rm -f "$HG_TEST_STREAM_WALKED_FILE_3"
 
 full regeneration
 -----------------
 
 A full nodemap is generated
 
 (ideally this test would append enough data to make sure the nodemap data file
 get changed, however to make thing simpler we will force the regeneration for
 this test.
 
 Check the initial state
 
   $ f --size test-repo/.hg/store/00changelog*
   test-repo/.hg/store/00changelog-*.nd: size=121344 (glob) (rust !)
   test-repo/.hg/store/00changelog-*.nd: size=121344 (glob) (pure !)
   test-repo/.hg/store/00changelog-*.nd: size=121152 (glob) (no-rust no-pure !)
   test-repo/.hg/store/00changelog.d: size=376950 (zstd !)
   test-repo/.hg/store/00changelog.d: size=368949 (no-zstd !)
   test-repo/.hg/store/00changelog.i: size=320448
   test-repo/.hg/store/00changelog.n: size=62
   $ hg -R test-repo debugnodemap --metadata | tee server-metadata-2.txt
   uid: * (glob)
   tip-rev: 5006
   tip-node: ed2ec1eef9aa2a0ec5057c51483bc148d03e810b
   data-length: 121344 (rust !)
   data-length: 121344 (pure !)
   data-length: 121152 (no-rust no-pure !)
   data-unused: 192 (rust !)
   data-unused: 192 (pure !)
   data-unused: 0 (no-rust no-pure !)
   data-unused: 0.158% (rust !)
   data-unused: 0.158% (pure !)
   data-unused: 0.000% (no-rust no-pure !)
 
 Performe the mix of clone and full refresh of the nodemap, so that the files
 (and filenames) are different between listing time and actual transfer time.
 
   $ (hg clone -U --stream ssh://user@dummy/test-repo stream-clone-race-2 --debug 2>> clone-output-2 | egrep '00(changelog|manifest)' >> clone-output-2; touch $HG_TEST_STREAM_WALKED_FILE_3) &
   $ $RUNTESTDIR/testlib/wait-on-file 10 $HG_TEST_STREAM_WALKED_FILE_1
   $ rm test-repo/.hg/store/00changelog.n
   $ rm test-repo/.hg/store/00changelog-*.nd
   $ hg -R test-repo/ debugupdatecache
   $ touch $HG_TEST_STREAM_WALKED_FILE_2
   $ $RUNTESTDIR/testlib/wait-on-file 10 $HG_TEST_STREAM_WALKED_FILE_3
 
 (note: the stream clone code wronly pick the `undo.` files)
 
   $ cat clone-output-2
   adding [s] undo.backup.00manifest.n (62 bytes) (known-bad-output !)
   adding [s] undo.backup.00changelog.n (62 bytes) (known-bad-output !)
   adding [s] 00manifest.n (62 bytes)
   adding [s] 00manifest-*.nd (118 KB) (glob)
   adding [s] 00changelog.n (62 bytes)
   adding [s] 00changelog-*.nd (118 KB) (glob)
   adding [s] 00manifest.d (492 KB) (zstd !)
   adding [s] 00manifest.d (452 KB) (no-zstd !)
   adding [s] 00changelog.d (360 KB) (no-zstd !)
   adding [s] 00changelog.d (368 KB) (zstd !)
   adding [s] 00manifest.i (313 KB)
   adding [s] 00changelog.i (313 KB)
 
 Check the result.
 
   $ f --size stream-clone-race-2/.hg/store/00changelog*
   stream-clone-race-2/.hg/store/00changelog-*.nd: size=121344 (glob) (rust !)
   stream-clone-race-2/.hg/store/00changelog-*.nd: size=121344 (glob) (pure !)
   stream-clone-race-2/.hg/store/00changelog-*.nd: size=121152 (glob) (no-rust no-pure !)
   stream-clone-race-2/.hg/store/00changelog.d: size=376950 (zstd !)
   stream-clone-race-2/.hg/store/00changelog.d: size=368949 (no-zstd !)
   stream-clone-race-2/.hg/store/00changelog.i: size=320448
   stream-clone-race-2/.hg/store/00changelog.n: size=62
 
   $ hg -R stream-clone-race-2 debugnodemap --metadata | tee client-metadata-2.txt
   uid: * (glob)
   tip-rev: 5006
   tip-node: ed2ec1eef9aa2a0ec5057c51483bc148d03e810b
   data-length: 121344 (rust !)
   data-unused: 192 (rust !)
   data-unused: 0.158% (rust !)
   data-length: 121152 (no-rust no-pure !)
   data-unused: 0 (no-rust no-pure !)
   data-unused: 0.000% (no-rust no-pure !)
   data-length: 121344 (pure !)
   data-unused: 192 (pure !)
   data-unused: 0.158% (pure !)
 
 We get a usable nodemap, so no rewrite would be needed and the metadata should be identical
 (ie: the following diff should be empty)
 
 This isn't the case for the `no-rust` `no-pure` implementation as it use a very minimal nodemap implementation that unconditionnaly rewrite the nodemap "all the time".
 
 #if no-rust no-pure
   $ diff -u server-metadata-2.txt client-metadata-2.txt
   --- server-metadata-2.txt	* (glob)
   +++ client-metadata-2.txt	* (glob)
   @@ -1,4 +1,4 @@
   -uid: * (glob)
   +uid: * (glob)
    tip-rev: 5006
    tip-node: ed2ec1eef9aa2a0ec5057c51483bc148d03e810b
    data-length: 121152
   [1]
 #else
   $ diff -u server-metadata-2.txt client-metadata-2.txt
 #endif
 
 Clean up after the test
 
   $ rm -f $HG_TEST_STREAM_WALKED_FILE_1
   $ rm -f $HG_TEST_STREAM_WALKED_FILE_2
   $ rm -f $HG_TEST_STREAM_WALKED_FILE_3