largefiles: show progress when checking standin hashes in outgoing changesets

This checking can take a huge amount of time and we should give user a hint
that something is going on.

hgext/largefiles/lfutil.py

@@ -1,583 +1,586 b''
 # Copyright 2009-2010 Gregory P. Ward
 # Copyright 2009-2010 Intelerad Medical Systems Incorporated
 # Copyright 2010-2011 Fog Creek Software
 # Copyright 2010-2011 Unity Technologies
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''largefiles utility code: must not import other modules in this package.'''
 
 import os
 import platform
 import shutil
 import stat
 import copy
 
 from mercurial import dirstate, httpconnection, match as match_, util, scmutil
 from mercurial.i18n import _
 from mercurial import node
 
 shortname = '.hglf'
 shortnameslash = shortname + '/'
 longname = 'largefiles'
 
 
 # -- Private worker functions ------------------------------------------
 
 def getminsize(ui, assumelfiles, opt, default=10):
     lfsize = opt
     if not lfsize and assumelfiles:
         lfsize = ui.config(longname, 'minsize', default=default)
     if lfsize:
         try:
             lfsize = float(lfsize)
         except ValueError:
             raise util.Abort(_('largefiles: size must be number (not %s)\n')
                              % lfsize)
     if lfsize is None:
         raise util.Abort(_('minimum size for largefiles must be specified'))
     return lfsize
 
 def link(src, dest):
     util.makedirs(os.path.dirname(dest))
     try:
         util.oslink(src, dest)
     except OSError:
         # if hardlinks fail, fallback on atomic copy
         dst = util.atomictempfile(dest)
         for chunk in util.filechunkiter(open(src, 'rb')):
             dst.write(chunk)
         dst.close()
         os.chmod(dest, os.stat(src).st_mode)
 
 def usercachepath(ui, hash):
     path = ui.configpath(longname, 'usercache', None)
     if path:
         path = os.path.join(path, hash)
     else:
         if os.name == 'nt':
             appdata = os.getenv('LOCALAPPDATA', os.getenv('APPDATA'))
             if appdata:
                 path = os.path.join(appdata, longname, hash)
         elif platform.system() == 'Darwin':
             home = os.getenv('HOME')
             if home:
                 path = os.path.join(home, 'Library', 'Caches',
                                     longname, hash)
         elif os.name == 'posix':
             path = os.getenv('XDG_CACHE_HOME')
             if path:
                 path = os.path.join(path, longname, hash)
             else:
                 home = os.getenv('HOME')
                 if home:
                     path = os.path.join(home, '.cache', longname, hash)
         else:
             raise util.Abort(_('unknown operating system: %s\n') % os.name)
     return path
 
 def inusercache(ui, hash):
     path = usercachepath(ui, hash)
     return path and os.path.exists(path)
 
 def findfile(repo, hash):
     if instore(repo, hash):
         repo.ui.note(_('found %s in store\n') % hash)
         return storepath(repo, hash)
     elif inusercache(repo.ui, hash):
         repo.ui.note(_('found %s in system cache\n') % hash)
         path = storepath(repo, hash)
         link(usercachepath(repo.ui, hash), path)
         return path
     return None
 
 class largefilesdirstate(dirstate.dirstate):
     def __getitem__(self, key):
         return super(largefilesdirstate, self).__getitem__(unixpath(key))
     def normal(self, f):
         return super(largefilesdirstate, self).normal(unixpath(f))
     def remove(self, f):
         return super(largefilesdirstate, self).remove(unixpath(f))
     def add(self, f):
         return super(largefilesdirstate, self).add(unixpath(f))
     def drop(self, f):
         return super(largefilesdirstate, self).drop(unixpath(f))
     def forget(self, f):
         return super(largefilesdirstate, self).forget(unixpath(f))
     def normallookup(self, f):
         return super(largefilesdirstate, self).normallookup(unixpath(f))
     def _ignore(self, f):
         return False
 
 def openlfdirstate(ui, repo, create=True):
     '''
     Return a dirstate object that tracks largefiles: i.e. its root is
     the repo root, but it is saved in .hg/largefiles/dirstate.
     '''
     lfstoredir = repo.join(longname)
     opener = scmutil.opener(lfstoredir)
     lfdirstate = largefilesdirstate(opener, ui, repo.root,
                                      repo.dirstate._validate)
 
     # If the largefiles dirstate does not exist, populate and create
     # it. This ensures that we create it on the first meaningful
     # largefiles operation in a new clone.
     if create and not os.path.exists(os.path.join(lfstoredir, 'dirstate')):
         matcher = getstandinmatcher(repo)
         standins = repo.dirstate.walk(matcher, [], False, False)
 
         if len(standins) > 0:
             util.makedirs(lfstoredir)
 
         for standin in standins:
             lfile = splitstandin(standin)
             lfdirstate.normallookup(lfile)
     return lfdirstate
 
 def lfdirstatestatus(lfdirstate, repo):
     wctx = repo['.']
     match = match_.always(repo.root, repo.getcwd())
     unsure, s = lfdirstate.status(match, [], False, False, False)
     modified, clean = s.modified, s.clean
     for lfile in unsure:
         try:
             fctx = wctx[standin(lfile)]
         except LookupError:
             fctx = None
         if not fctx or fctx.data().strip() != hashfile(repo.wjoin(lfile)):
             modified.append(lfile)
         else:
             clean.append(lfile)
             lfdirstate.normal(lfile)
     return s
 
 def listlfiles(repo, rev=None, matcher=None):
     '''return a list of largefiles in the working copy or the
     specified changeset'''
 
     if matcher is None:
         matcher = getstandinmatcher(repo)
 
     # ignore unknown files in working directory
     return [splitstandin(f)
             for f in repo[rev].walk(matcher)
             if rev is not None or repo.dirstate[f] != '?']
 
 def instore(repo, hash):
     return os.path.exists(storepath(repo, hash))
 
 def storepath(repo, hash):
     return repo.join(os.path.join(longname, hash))
 
 def copyfromcache(repo, hash, filename):
     '''Copy the specified largefile from the repo or system cache to
     filename in the repository. Return true on success or false if the
     file was not found in either cache (which should not happened:
     this is meant to be called only after ensuring that the needed
     largefile exists in the cache).'''
     path = findfile(repo, hash)
     if path is None:
         return False
     util.makedirs(os.path.dirname(repo.wjoin(filename)))
     # The write may fail before the file is fully written, but we
     # don't use atomic writes in the working copy.
     shutil.copy(path, repo.wjoin(filename))
     return True
 
 def copytostore(repo, rev, file, uploaded=False):
     hash = readstandin(repo, file, rev)
     if instore(repo, hash):
         return
     copytostoreabsolute(repo, repo.wjoin(file), hash)
 
 def copyalltostore(repo, node):
     '''Copy all largefiles in a given revision to the store'''
 
     ctx = repo[node]
     for filename in ctx.files():
         if isstandin(filename) and filename in ctx.manifest():
             realfile = splitstandin(filename)
             copytostore(repo, ctx.node(), realfile)
 
 
 def copytostoreabsolute(repo, file, hash):
     if inusercache(repo.ui, hash):
         link(usercachepath(repo.ui, hash), storepath(repo, hash))
     else:
         util.makedirs(os.path.dirname(storepath(repo, hash)))
         dst = util.atomictempfile(storepath(repo, hash),
                                   createmode=repo.store.createmode)
         for chunk in util.filechunkiter(open(file, 'rb')):
             dst.write(chunk)
         dst.close()
         linktousercache(repo, hash)
 
 def linktousercache(repo, hash):
     path = usercachepath(repo.ui, hash)
     if path:
         link(storepath(repo, hash), path)
 
 def getstandinmatcher(repo, pats=[], opts={}):
     '''Return a match object that applies pats to the standin directory'''
     standindir = repo.wjoin(shortname)
     if pats:
         pats = [os.path.join(standindir, pat) for pat in pats]
     else:
         # no patterns: relative to repo root
         pats = [standindir]
     # no warnings about missing files or directories
     match = scmutil.match(repo[None], pats, opts)
     match.bad = lambda f, msg: None
     return match
 
 def composestandinmatcher(repo, rmatcher):
     '''Return a matcher that accepts standins corresponding to the
     files accepted by rmatcher. Pass the list of files in the matcher
     as the paths specified by the user.'''
     smatcher = getstandinmatcher(repo, rmatcher.files())
     isstandin = smatcher.matchfn
     def composedmatchfn(f):
         return isstandin(f) and rmatcher.matchfn(splitstandin(f))
     smatcher.matchfn = composedmatchfn
 
     return smatcher
 
 def standin(filename):
     '''Return the repo-relative path to the standin for the specified big
     file.'''
     # Notes:
     # 1) Some callers want an absolute path, but for instance addlargefiles
     #    needs it repo-relative so it can be passed to repo[None].add().  So
     #    leave it up to the caller to use repo.wjoin() to get an absolute path.
     # 2) Join with '/' because that's what dirstate always uses, even on
     #    Windows. Change existing separator to '/' first in case we are
     #    passed filenames from an external source (like the command line).
     return shortnameslash + util.pconvert(filename)
 
 def isstandin(filename):
     '''Return true if filename is a big file standin. filename must be
     in Mercurial's internal form (slash-separated).'''
     return filename.startswith(shortnameslash)
 
 def splitstandin(filename):
     # Split on / because that's what dirstate always uses, even on Windows.
     # Change local separator to / first just in case we are passed filenames
     # from an external source (like the command line).
     bits = util.pconvert(filename).split('/', 1)
     if len(bits) == 2 and bits[0] == shortname:
         return bits[1]
     else:
         return None
 
 def updatestandin(repo, standin):
     file = repo.wjoin(splitstandin(standin))
     if os.path.exists(file):
         hash = hashfile(file)
         executable = getexecutable(file)
         writestandin(repo, standin, hash, executable)
 
 def readstandin(repo, filename, node=None):
     '''read hex hash from standin for filename at given node, or working
     directory if no node is given'''
     return repo[node][standin(filename)].data().strip()
 
 def writestandin(repo, standin, hash, executable):
     '''write hash to <repo.root>/<standin>'''
     repo.wwrite(standin, hash + '\n', executable and 'x' or '')
 
 def copyandhash(instream, outfile):
     '''Read bytes from instream (iterable) and write them to outfile,
     computing the SHA-1 hash of the data along the way. Return the hash.'''
     hasher = util.sha1('')
     for data in instream:
         hasher.update(data)
         outfile.write(data)
     return hasher.hexdigest()
 
 def hashrepofile(repo, file):
     return hashfile(repo.wjoin(file))
 
 def hashfile(file):
     if not os.path.exists(file):
         return ''
     hasher = util.sha1('')
     fd = open(file, 'rb')
     for data in util.filechunkiter(fd, 128 * 1024):
         hasher.update(data)
     fd.close()
     return hasher.hexdigest()
 
 def getexecutable(filename):
     mode = os.stat(filename).st_mode
     return ((mode & stat.S_IXUSR) and
             (mode & stat.S_IXGRP) and
             (mode & stat.S_IXOTH))
 
 def urljoin(first, second, *arg):
     def join(left, right):
         if not left.endswith('/'):
             left += '/'
         if right.startswith('/'):
             right = right[1:]
         return left + right
 
     url = join(first, second)
     for a in arg:
         url = join(url, a)
     return url
 
 def hexsha1(data):
     """hexsha1 returns the hex-encoded sha1 sum of the data in the file-like
     object data"""
     h = util.sha1()
     for chunk in util.filechunkiter(data):
         h.update(chunk)
     return h.hexdigest()
 
 def httpsendfile(ui, filename):
     return httpconnection.httpsendfile(ui, filename, 'rb')
 
 def unixpath(path):
     '''Return a version of path normalized for use with the lfdirstate.'''
     return util.pconvert(os.path.normpath(path))
 
 def islfilesrepo(repo):
     if ('largefiles' in repo.requirements and
             util.any(shortnameslash in f[0] for f in repo.store.datafiles())):
         return True
 
     return util.any(openlfdirstate(repo.ui, repo, False))
 
 class storeprotonotcapable(Exception):
     def __init__(self, storetypes):
         self.storetypes = storetypes
 
 def getstandinsstate(repo):
     standins = []
     matcher = getstandinmatcher(repo)
     for standin in repo.dirstate.walk(matcher, [], False, False):
         lfile = splitstandin(standin)
         try:
             hash = readstandin(repo, lfile)
         except IOError:
             hash = None
         standins.append((lfile, hash))
     return standins
 
 def synclfdirstate(repo, lfdirstate, lfile, normallookup):
     lfstandin = standin(lfile)
     if lfstandin in repo.dirstate:
         stat = repo.dirstate._map[lfstandin]
         state, mtime = stat[0], stat[3]
     else:
         state, mtime = '?', -1
     if state == 'n':
         if normallookup or mtime < 0:
             # state 'n' doesn't ensure 'clean' in this case
             lfdirstate.normallookup(lfile)
         else:
             lfdirstate.normal(lfile)
     elif state == 'm':
         lfdirstate.normallookup(lfile)
     elif state == 'r':
         lfdirstate.remove(lfile)
     elif state == 'a':
         lfdirstate.add(lfile)
     elif state == '?':
         lfdirstate.drop(lfile)
 
 def markcommitted(orig, ctx, node):
     repo = ctx._repo
 
     orig(node)
 
     # ATTENTION: "ctx.files()" may differ from "repo[node].files()"
     # because files coming from the 2nd parent are omitted in the latter.
     #
     # The former should be used to get targets of "synclfdirstate",
     # because such files:
     # - are marked as "a" by "patch.patch()" (e.g. via transplant), and
     # - have to be marked as "n" after commit, but
     # - aren't listed in "repo[node].files()"
 
     lfdirstate = openlfdirstate(repo.ui, repo)
     for f in ctx.files():
         if isstandin(f):
             lfile = splitstandin(f)
             synclfdirstate(repo, lfdirstate, lfile, False)
     lfdirstate.write()
 
     # As part of committing, copy all of the largefiles into the cache.
     copyalltostore(repo, node)
 
 def getlfilestoupdate(oldstandins, newstandins):
     changedstandins = set(oldstandins).symmetric_difference(set(newstandins))
     filelist = []
     for f in changedstandins:
         if f[0] not in filelist:
             filelist.append(f[0])
     return filelist
 
 def getlfilestoupload(repo, missing, addfunc):
-    for n in missing:
+    for i, n in enumerate(missing):
+        repo.ui.progress(_('finding outgoing largefiles'), i,
+            unit=_('revision'), total=len(missing))
         parents = [p for p in repo.changelog.parents(n) if p != node.nullid]
 
         oldlfstatus = repo.lfstatus
         repo.lfstatus = False
         try:
             ctx = repo[n]
         finally:
             repo.lfstatus = oldlfstatus
 
         files = set(ctx.files())
         if len(parents) == 2:
             mc = ctx.manifest()
             mp1 = ctx.parents()[0].manifest()
             mp2 = ctx.parents()[1].manifest()
             for f in mp1:
                 if f not in mc:
                     files.add(f)
             for f in mp2:
                 if f not in mc:
                     files.add(f)
             for f in mc:
                 if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                     files.add(f)
         for fn in files:
             if isstandin(fn) and fn in ctx:
                 addfunc(fn, ctx[fn].data().strip())
+    repo.ui.progress(_('finding outgoing largefiles'), None)
 
 def updatestandinsbymatch(repo, match):
     '''Update standins in the working directory according to specified match
 
     This returns (possibly modified) ``match`` object to be used for
     subsequent commit process.
     '''
 
     ui = repo.ui
 
     # Case 1: user calls commit with no specific files or
     # include/exclude patterns: refresh and commit all files that
     # are "dirty".
     if match is None or match.always():
         # Spend a bit of time here to get a list of files we know
         # are modified so we can compare only against those.
         # It can cost a lot of time (several seconds)
         # otherwise to update all standins if the largefiles are
         # large.
         lfdirstate = openlfdirstate(ui, repo)
         dirtymatch = match_.always(repo.root, repo.getcwd())
         unsure, s = lfdirstate.status(dirtymatch, [], False, False,
                                       False)
         modifiedfiles = unsure + s.modified + s.added + s.removed
         lfiles = listlfiles(repo)
         # this only loops through largefiles that exist (not
         # removed/renamed)
         for lfile in lfiles:
             if lfile in modifiedfiles:
                 if os.path.exists(
                         repo.wjoin(standin(lfile))):
                     # this handles the case where a rebase is being
                     # performed and the working copy is not updated
                     # yet.
                     if os.path.exists(repo.wjoin(lfile)):
                         updatestandin(repo,
                             standin(lfile))
 
         return match
 
     lfiles = listlfiles(repo)
     match._files = repo._subdirlfs(match.files(), lfiles)
 
     # Case 2: user calls commit with specified patterns: refresh
     # any matching big files.
     smatcher = composestandinmatcher(repo, match)
     standins = repo.dirstate.walk(smatcher, [], False, False)
 
     # No matching big files: get out of the way and pass control to
     # the usual commit() method.
     if not standins:
         return match
 
     # Refresh all matching big files.  It's possible that the
     # commit will end up failing, in which case the big files will
     # stay refreshed.  No harm done: the user modified them and
     # asked to commit them, so sooner or later we're going to
     # refresh the standins.  Might as well leave them refreshed.
     lfdirstate = openlfdirstate(ui, repo)
     for fstandin in standins:
         lfile = splitstandin(fstandin)
         if lfdirstate[lfile] != 'r':
             updatestandin(repo, fstandin)
 
     # Cook up a new matcher that only matches regular files or
     # standins corresponding to the big files requested by the
     # user.  Have to modify _files to prevent commit() from
     # complaining "not tracked" for big files.
     match = copy.copy(match)
     origmatchfn = match.matchfn
 
     # Check both the list of largefiles and the list of
     # standins because if a largefile was removed, it
     # won't be in the list of largefiles at this point
     match._files += sorted(standins)
 
     actualfiles = []
     for f in match._files:
         fstandin = standin(f)
 
         # ignore known largefiles and standins
         if f in lfiles or fstandin in standins:
             continue
 
         actualfiles.append(f)
     match._files = actualfiles
 
     def matchfn(f):
         if origmatchfn(f):
             return f not in lfiles
         else:
             return f in standins
 
     match.matchfn = matchfn
 
     return match
 
 class automatedcommithook(object):
     '''Stateful hook to update standins at the 1st commit of resuming
 
     For efficiency, updating standins in the working directory should
     be avoided while automated committing (like rebase, transplant and
     so on), because they should be updated before committing.
 
     But the 1st commit of resuming automated committing (e.g. ``rebase
     --continue``) should update them, because largefiles may be
     modified manually.
     '''
     def __init__(self, resuming):
         self.resuming = resuming
 
     def __call__(self, repo, match):
         if self.resuming:
             self.resuming = False # avoids updating at subsequent commits
             return updatestandinsbymatch(repo, match)
         else:
             return match
 
 def getstatuswriter(ui, repo, forcibly=None):
     '''Return the function to write largefiles specific status out
 
     If ``forcibly`` is ``None``, this returns the last element of
     ``repo._lfstatuswriters`` as "default" writer function.
 
     Otherwise, this returns the function to always write out (or
     ignore if ``not forcibly``) status.
     '''
     if forcibly is None:
         return repo._lfstatuswriters[-1]
     else:
         if forcibly:
             return ui.status # forcibly WRITE OUT
         else:
             return lambda *msg, **opts: None # forcibly IGNORE

tests/test-largefiles-misc.t

@@ -1,931 +1,942 b''
 This file contains testcases that tend to be related to special cases or less
 common commands affecting largefile.
 
 Each sections should be independent of each others.
 
   $ USERCACHE="$TESTTMP/cache"; export USERCACHE
   $ mkdir "${USERCACHE}"
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > largefiles=
   > purge=
   > rebase=
   > transplant=
   > [phases]
   > publish=False
   > [largefiles]
   > minsize=2
   > patterns=glob:**.dat
   > usercache=${USERCACHE}
   > [hooks]
   > precommit=sh -c "echo \\"Invoking status precommit hook\\"; hg status"
   > EOF
 
 
 
 Test copies and moves from a directory other than root (issue3516)
 =========================================================================
 
   $ hg init lf_cpmv
   $ cd lf_cpmv
   $ mkdir dira
   $ mkdir dira/dirb
   $ touch dira/dirb/largefile
   $ hg add --large dira/dirb/largefile
   $ hg commit -m "added"
   Invoking status precommit hook
   A dira/dirb/largefile
   $ cd dira
   $ hg cp dirb/largefile foo/largefile
   $ hg ci -m "deep copy"
   Invoking status precommit hook
   A dira/foo/largefile
   $ find . | sort
   .
   ./dirb
   ./dirb/largefile
   ./foo
   ./foo/largefile
   $ hg mv foo/largefile baz/largefile
   $ hg ci -m "moved"
   Invoking status precommit hook
   A dira/baz/largefile
   R dira/foo/largefile
   $ find . | sort
   .
   ./baz
   ./baz/largefile
   ./dirb
   ./dirb/largefile
   $ cd ..
   $ hg mv dira dirc
   moving .hglf/dira/baz/largefile to .hglf/dirc/baz/largefile (glob)
   moving .hglf/dira/dirb/largefile to .hglf/dirc/dirb/largefile (glob)
   $ find * | sort
   dirc
   dirc/baz
   dirc/baz/largefile
   dirc/dirb
   dirc/dirb/largefile
 
   $ hg clone -q . ../fetch
   $ hg --config extensions.fetch= fetch ../fetch
   abort: uncommitted changes
   [255]
   $ hg up -qC
   $ cd ..
 
 Clone a local repository owned by another user
 ===================================================
 
 #if unix-permissions
 
 We have to simulate that here by setting $HOME and removing write permissions
   $ ORIGHOME="$HOME"
   $ mkdir alice
   $ HOME="`pwd`/alice"
   $ cd alice
   $ hg init pubrepo
   $ cd pubrepo
   $ dd if=/dev/zero bs=1k count=11k > a-large-file 2> /dev/null
   $ hg add --large a-large-file
   $ hg commit -m "Add a large file"
   Invoking status precommit hook
   A a-large-file
   $ cd ..
   $ chmod -R a-w pubrepo
   $ cd ..
   $ mkdir bob
   $ HOME="`pwd`/bob"
   $ cd bob
   $ hg clone --pull ../alice/pubrepo pubrepo
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
   updating to branch default
   getting changed largefiles
   1 largefiles updated, 0 removed
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ cd ..
   $ chmod -R u+w alice/pubrepo
   $ HOME="$ORIGHOME"
 
 #endif
 
 
 Symlink to a large largefile should behave the same as a symlink to a normal file
 =====================================================================================
 
 #if symlink
 
   $ hg init largesymlink
   $ cd largesymlink
   $ dd if=/dev/zero bs=1k count=10k of=largefile 2>/dev/null
   $ hg add --large largefile
   $ hg commit -m "commit a large file"
   Invoking status precommit hook
   A largefile
   $ ln -s largefile largelink
   $ hg add largelink
   $ hg commit -m "commit a large symlink"
   Invoking status precommit hook
   A largelink
   $ rm -f largelink
   $ hg up >/dev/null
   $ test -f largelink
   [1]
   $ test -L largelink
   [1]
   $ rm -f largelink # make next part of the test independent of the previous
   $ hg up -C >/dev/null
   $ test -f largelink
   $ test -L largelink
   $ cd ..
 
 #endif
 
 
 test for pattern matching on 'hg status':
 ==============================================
 
 
 to boost performance, largefiles checks whether specified patterns are
 related to largefiles in working directory (NOT to STANDIN) or not.
 
   $ hg init statusmatch
   $ cd statusmatch
 
   $ mkdir -p a/b/c/d
   $ echo normal > a/b/c/d/e.normal.txt
   $ hg add a/b/c/d/e.normal.txt
   $ echo large > a/b/c/d/e.large.txt
   $ hg add --large a/b/c/d/e.large.txt
   $ mkdir -p a/b/c/x
   $ echo normal > a/b/c/x/y.normal.txt
   $ hg add a/b/c/x/y.normal.txt
   $ hg commit -m 'add files'
   Invoking status precommit hook
   A a/b/c/d/e.large.txt
   A a/b/c/d/e.normal.txt
   A a/b/c/x/y.normal.txt
 
 (1) no pattern: no performance boost
   $ hg status -A
   C a/b/c/d/e.large.txt
   C a/b/c/d/e.normal.txt
   C a/b/c/x/y.normal.txt
 
 (2) pattern not related to largefiles: performance boost
   $ hg status -A a/b/c/x
   C a/b/c/x/y.normal.txt
 
 (3) pattern related to largefiles: no performance boost
   $ hg status -A a/b/c/d
   C a/b/c/d/e.large.txt
   C a/b/c/d/e.normal.txt
 
 (4) pattern related to STANDIN (not to largefiles): performance boost
   $ hg status -A .hglf/a
   C .hglf/a/b/c/d/e.large.txt
 
 (5) mixed case: no performance boost
   $ hg status -A a/b/c/x a/b/c/d
   C a/b/c/d/e.large.txt
   C a/b/c/d/e.normal.txt
   C a/b/c/x/y.normal.txt
 
 verify that largefiles doesn't break filesets
 
   $ hg log --rev . --exclude "set:binary()"
   changeset:   0:41bd42f10efa
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     add files
   
 verify that large files in subrepos handled properly
   $ hg init subrepo
   $ echo "subrepo = subrepo" > .hgsub
   $ hg add .hgsub
   $ hg ci -m "add subrepo"
   Invoking status precommit hook
   A .hgsub
   ? .hgsubstate
   $ echo "rev 1" > subrepo/large.txt
   $ hg add --large subrepo/large.txt
   $ hg sum
   parent: 1:8ee150ea2e9c tip
    add subrepo
   branch: default
   commit: 1 subrepos
   update: (current)
   $ hg st
   $ hg st -S
   A subrepo/large.txt
   $ hg ci -S -m "commit top repo"
   committing subrepository subrepo
   Invoking status precommit hook
   A large.txt
   Invoking status precommit hook
   M .hgsubstate
 # No differences
   $ hg st -S
   $ hg sum
   parent: 2:ce4cd0c527a6 tip
    commit top repo
   branch: default
   commit: (clean)
   update: (current)
   $ echo "rev 2" > subrepo/large.txt
   $ hg st -S
   M subrepo/large.txt
   $ hg sum
   parent: 2:ce4cd0c527a6 tip
    commit top repo
   branch: default
   commit: 1 subrepos
   update: (current)
   $ hg ci -m "this commit should fail without -S"
   abort: uncommitted changes in subrepo subrepo
   (use --subrepos for recursive commit)
   [255]
 
 Add a normal file to the subrepo, then test archiving
 
   $ echo 'normal file' > subrepo/normal.txt
   $ touch large.dat
   $ mv subrepo/large.txt subrepo/renamed-large.txt
   $ hg addremove -S --dry-run
   adding large.dat as a largefile
   removing subrepo/large.txt
   adding subrepo/normal.txt
   adding subrepo/renamed-large.txt
   $ hg status -S
   ! subrepo/large.txt
   ? large.dat
   ? subrepo/normal.txt
   ? subrepo/renamed-large.txt
 
   $ hg addremove --dry-run subrepo
   removing subrepo/large.txt (glob)
   adding subrepo/normal.txt (glob)
   adding subrepo/renamed-large.txt (glob)
   $ hg status -S
   ! subrepo/large.txt
   ? large.dat
   ? subrepo/normal.txt
   ? subrepo/renamed-large.txt
   $ cd ..
 
   $ hg -R statusmatch addremove --dry-run statusmatch/subrepo
   removing statusmatch/subrepo/large.txt (glob)
   adding statusmatch/subrepo/normal.txt (glob)
   adding statusmatch/subrepo/renamed-large.txt (glob)
   $ hg -R statusmatch status -S
   ! subrepo/large.txt
   ? large.dat
   ? subrepo/normal.txt
   ? subrepo/renamed-large.txt
 
   $ hg -R statusmatch addremove --dry-run -S
   adding large.dat as a largefile
   removing subrepo/large.txt
   adding subrepo/normal.txt
   adding subrepo/renamed-large.txt
   $ cd statusmatch
 
   $ mv subrepo/renamed-large.txt subrepo/large.txt
   $ hg addremove subrepo
   adding subrepo/normal.txt (glob)
   $ hg forget subrepo/normal.txt
 
   $ hg addremove -S
   adding large.dat as a largefile
   adding subrepo/normal.txt
   $ rm large.dat
 
   $ hg addremove subrepo
   $ hg addremove -S
   removing large.dat
 
 Lock in subrepo, otherwise the change isn't archived
 
   $ hg ci -S -m "add normal file to top level"
   committing subrepository subrepo
   Invoking status precommit hook
   M large.txt
   A normal.txt
   Invoking status precommit hook
   M .hgsubstate
   $ hg archive -S ../lf_subrepo_archive
   $ find ../lf_subrepo_archive | sort
   ../lf_subrepo_archive
   ../lf_subrepo_archive/.hg_archival.txt
   ../lf_subrepo_archive/.hgsub
   ../lf_subrepo_archive/.hgsubstate
   ../lf_subrepo_archive/a
   ../lf_subrepo_archive/a/b
   ../lf_subrepo_archive/a/b/c
   ../lf_subrepo_archive/a/b/c/d
   ../lf_subrepo_archive/a/b/c/d/e.large.txt
   ../lf_subrepo_archive/a/b/c/d/e.normal.txt
   ../lf_subrepo_archive/a/b/c/x
   ../lf_subrepo_archive/a/b/c/x/y.normal.txt
   ../lf_subrepo_archive/subrepo
   ../lf_subrepo_archive/subrepo/large.txt
   ../lf_subrepo_archive/subrepo/normal.txt
 
 Test update with subrepos.
 
   $ hg update 0
   getting changed largefiles
   0 largefiles updated, 1 removed
   0 files updated, 0 files merged, 2 files removed, 0 files unresolved
   $ hg status -S
   $ hg update tip
   getting changed largefiles
   1 largefiles updated, 0 removed
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg status -S
 # modify a large file
   $ echo "modified" > subrepo/large.txt
   $ hg st -S
   M subrepo/large.txt
 # update -C should revert the change.
   $ hg update -C
   getting changed largefiles
   1 largefiles updated, 0 removed
   getting changed largefiles
   0 largefiles updated, 0 removed
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg status -S
 
   $ hg rm -v subrepo/large.txt
   removing subrepo/large.txt (glob)
   $ hg revert -R subrepo subrepo/large.txt
   $ rm subrepo/large.txt
   $ hg addremove -S
   removing subrepo/large.txt
   $ hg st -S
   R subrepo/large.txt
 
 Test archiving a revision that references a subrepo that is not yet
 cloned (see test-subrepo-recursion.t):
 
   $ hg clone -U . ../empty
   $ cd ../empty
   $ hg archive --subrepos -r tip ../archive.tar.gz
   cloning subrepo subrepo from $TESTTMP/statusmatch/subrepo
   $ cd ..
 
 
 
 
 
 
 Test addremove, forget and others
 ==============================================
 
 Test that addremove picks up largefiles prior to the initial commit (issue3541)
 
   $ hg init addrm2
   $ cd addrm2
   $ touch large.dat
   $ touch large2.dat
   $ touch normal
   $ hg add --large large.dat
   $ hg addremove -v
   adding large2.dat as a largefile
   adding normal
 
 Test that forgetting all largefiles reverts to islfilesrepo() == False
 (addremove will add *.dat as normal files now)
   $ hg forget large.dat
   $ hg forget large2.dat
   $ hg addremove -v
   adding large.dat
   adding large2.dat
 
 Test commit's addremove option prior to the first commit
   $ hg forget large.dat
   $ hg forget large2.dat
   $ hg add --large large.dat
   $ hg ci -Am "commit"
   adding large2.dat as a largefile
   Invoking status precommit hook
   A large.dat
   A large2.dat
   A normal
   $ find .hglf | sort
   .hglf
   .hglf/large.dat
   .hglf/large2.dat
 
 Test actions on largefiles using relative paths from subdir
 
   $ mkdir sub
   $ cd sub
   $ echo anotherlarge > anotherlarge
   $ hg add --large anotherlarge
   $ hg st
   A sub/anotherlarge
   $ hg st anotherlarge
   A anotherlarge
   $ hg commit -m anotherlarge anotherlarge
   Invoking status precommit hook
   A sub/anotherlarge
   $ hg log anotherlarge
   changeset:   1:9627a577c5e9
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     anotherlarge
   
   $ hg log -G anotherlarge
   @  changeset:   1:9627a577c5e9
   |  tag:         tip
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     anotherlarge
   |
   $ echo more >> anotherlarge
   $ hg st .
   M anotherlarge
   $ hg cat anotherlarge
   anotherlarge
   $ hg revert anotherlarge
   $ hg st
   ? sub/anotherlarge.orig
   $ cd ..
 
   $ cd ..
 
 Check error message while exchange
 =========================================================
 
 issue3651: summary/outgoing with largefiles shows "no remote repo"
 unexpectedly
 
   $ mkdir issue3651
   $ cd issue3651
 
   $ hg init src
   $ echo a > src/a
   $ hg -R src add --large src/a
   $ hg -R src commit -m '#0'
   Invoking status precommit hook
   A a
 
 check messages when no remote repository is specified:
 "no remote repo" route for "hg outgoing --large" is not tested here,
 because it can't be reproduced easily.
 
   $ hg init clone1
   $ hg -R clone1 -q pull src
   $ hg -R clone1 -q update
   $ hg -R clone1 paths | grep default
   [1]
 
   $ hg -R clone1 summary --large
   parent: 0:fc0bd45326d3 tip
    #0
   branch: default
   commit: (clean)
   update: (current)
   largefiles: (no remote repo)
 
 check messages when there is no files to upload:
 
   $ hg -q clone src clone2
   $ hg -R clone2 paths | grep default
   default = $TESTTMP/issue3651/src (glob)
 
   $ hg -R clone2 summary --large
   parent: 0:fc0bd45326d3 tip
    #0
   branch: default
   commit: (clean)
   update: (current)
   largefiles: (no files to upload)
   $ hg -R clone2 outgoing --large
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   no changes found
   largefiles: no files to upload
   [1]
 
   $ hg -R clone2 outgoing --large --graph --template "{rev}"
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   no changes found
   largefiles: no files to upload
 
 check messages when there are files to upload:
 
   $ echo b > clone2/b
   $ hg -R clone2 add --large clone2/b
   $ hg -R clone2 commit -m '#1'
   Invoking status precommit hook
   A b
   $ hg -R clone2 summary --large
   parent: 1:1acbe71ce432 tip
    #1
   branch: default
   commit: (clean)
   update: (current)
   largefiles: 1 entities for 1 files to upload
   $ hg -R clone2 outgoing --large
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   changeset:   1:1acbe71ce432
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     #1
   
   largefiles to upload (1 entities):
   b
   
   $ hg -R clone2 outgoing --large --graph --template "{rev}"
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   @  1
   
   largefiles to upload (1 entities):
   b
   
 
   $ cp clone2/b clone2/b1
   $ cp clone2/b clone2/b2
   $ hg -R clone2 add --large clone2/b1 clone2/b2
   $ hg -R clone2 commit -m '#2: add largefiles referring same entity'
   Invoking status precommit hook
   A b1
   A b2
   $ hg -R clone2 summary --large
   parent: 2:6095d0695d70 tip
    #2: add largefiles referring same entity
   branch: default
   commit: (clean)
   update: (current)
   largefiles: 1 entities for 3 files to upload
   $ hg -R clone2 outgoing --large -T "{rev}:{node|short}\n"
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   1:1acbe71ce432
   2:6095d0695d70
   largefiles to upload (1 entities):
   b
   b1
   b2
   
   $ hg -R clone2 cat -r 1 clone2/.hglf/b
   89e6c98d92887913cadf06b2adb97f26cde4849b
   $ hg -R clone2 outgoing --large -T "{rev}:{node|short}\n" --debug
   comparing with $TESTTMP/issue3651/src (glob)
   query 1; heads
   searching for changes
   all remote heads known locally
   1:1acbe71ce432
   2:6095d0695d70
+  finding outgoing largefiles: 0/2 revision (0.00%)
+  finding outgoing largefiles: 1/2 revision (50.00%)
   largefiles to upload (1 entities):
   b
       89e6c98d92887913cadf06b2adb97f26cde4849b
   b1
       89e6c98d92887913cadf06b2adb97f26cde4849b
   b2
       89e6c98d92887913cadf06b2adb97f26cde4849b
   
 
   $ echo bbb > clone2/b
   $ hg -R clone2 commit -m '#3: add new largefile entity as existing file'
   Invoking status precommit hook
   M b
   $ echo bbbb > clone2/b
   $ hg -R clone2 commit -m '#4: add new largefile entity as existing file'
   Invoking status precommit hook
   M b
   $ cp clone2/b1 clone2/b
   $ hg -R clone2 commit -m '#5: refer existing largefile entity again'
   Invoking status precommit hook
   M b
   $ hg -R clone2 summary --large
   parent: 5:036794ea641c tip
    #5: refer existing largefile entity again
   branch: default
   commit: (clean)
   update: (current)
   largefiles: 3 entities for 3 files to upload
   $ hg -R clone2 outgoing --large -T "{rev}:{node|short}\n"
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   1:1acbe71ce432
   2:6095d0695d70
   3:7983dce246cc
   4:233f12ada4ae
   5:036794ea641c
   largefiles to upload (3 entities):
   b
   b1
   b2
   
   $ hg -R clone2 cat -r 3 clone2/.hglf/b
   c801c9cfe94400963fcb683246217d5db77f9a9a
   $ hg -R clone2 cat -r 4 clone2/.hglf/b
   13f9ed0898e315bf59dc2973fec52037b6f441a2
   $ hg -R clone2 outgoing --large -T "{rev}:{node|short}\n" --debug
   comparing with $TESTTMP/issue3651/src (glob)
   query 1; heads
   searching for changes
   all remote heads known locally
   1:1acbe71ce432
   2:6095d0695d70
   3:7983dce246cc
   4:233f12ada4ae
   5:036794ea641c
+  finding outgoing largefiles: 0/5 revision (0.00%)
+  finding outgoing largefiles: 1/5 revision (20.00%)
+  finding outgoing largefiles: 2/5 revision (40.00%)
+  finding outgoing largefiles: 3/5 revision (60.00%)
+  finding outgoing largefiles: 4/5 revision (80.00%)
   largefiles to upload (3 entities):
   b
       13f9ed0898e315bf59dc2973fec52037b6f441a2
       89e6c98d92887913cadf06b2adb97f26cde4849b
       c801c9cfe94400963fcb683246217d5db77f9a9a
   b1
       89e6c98d92887913cadf06b2adb97f26cde4849b
   b2
       89e6c98d92887913cadf06b2adb97f26cde4849b
   
 
 Pushing revision #1 causes uploading entity 89e6c98d9288, which is
 shared also by largefiles b1, b2 in revision #2 and b in revision #5.
 
 Then, entity 89e6c98d9288 is not treated as "outgoing entity" at "hg
 summary" and "hg outgoing", even though files in outgoing revision #2
 and #5 refer it.
 
   $ hg -R clone2 push -r 1 -q
   $ hg -R clone2 summary --large
   parent: 5:036794ea641c tip
    #5: refer existing largefile entity again
   branch: default
   commit: (clean)
   update: (current)
   largefiles: 2 entities for 1 files to upload
   $ hg -R clone2 outgoing --large -T "{rev}:{node|short}\n"
   comparing with $TESTTMP/issue3651/src (glob)
   searching for changes
   2:6095d0695d70
   3:7983dce246cc
   4:233f12ada4ae
   5:036794ea641c
   largefiles to upload (2 entities):
   b
   
   $ hg -R clone2 outgoing --large -T "{rev}:{node|short}\n" --debug
   comparing with $TESTTMP/issue3651/src (glob)
   query 1; heads
   searching for changes
   all remote heads known locally
   2:6095d0695d70
   3:7983dce246cc
   4:233f12ada4ae
   5:036794ea641c
+  finding outgoing largefiles: 0/4 revision (0.00%)
+  finding outgoing largefiles: 1/4 revision (25.00%)
+  finding outgoing largefiles: 2/4 revision (50.00%)
+  finding outgoing largefiles: 3/4 revision (75.00%)
   largefiles to upload (2 entities):
   b
       13f9ed0898e315bf59dc2973fec52037b6f441a2
       c801c9cfe94400963fcb683246217d5db77f9a9a
   
 
   $ cd ..
 
 merge action 'd' for 'local renamed directory to d2/g' which has no filename
 ==================================================================================
 
   $ hg init merge-action
   $ cd merge-action
   $ touch l
   $ hg add --large l
   $ mkdir d1
   $ touch d1/f
   $ hg ci -Aqm0
   Invoking status precommit hook
   A d1/f
   A l
   $ echo > d1/f
   $ touch d1/g
   $ hg ci -Aqm1
   Invoking status precommit hook
   M d1/f
   A d1/g
   $ hg up -qr0
   $ hg mv d1 d2
   moving d1/f to d2/f (glob)
   $ hg ci -qm2
   Invoking status precommit hook
   A d2/f
   R d1/f
   $ hg merge
   merging d2/f and d1/f to d2/f
   getting changed largefiles
   0 largefiles updated, 0 removed
   1 files updated, 1 files merged, 0 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
   $ cd ..
 
 
 Merge conflicts:
 =====================
 
   $ hg init merge
   $ cd merge
   $ echo 0 > f-different
   $ echo 0 > f-same
   $ echo 0 > f-unchanged-1
   $ echo 0 > f-unchanged-2
   $ hg add --large *
   $ hg ci -m0
   Invoking status precommit hook
   A f-different
   A f-same
   A f-unchanged-1
   A f-unchanged-2
   $ echo tmp1 > f-unchanged-1
   $ echo tmp1 > f-unchanged-2
   $ echo tmp1 > f-same
   $ hg ci -m1
   Invoking status precommit hook
   M f-same
   M f-unchanged-1
   M f-unchanged-2
   $ echo 2 > f-different
   $ echo 0 > f-unchanged-1
   $ echo 1 > f-unchanged-2
   $ echo 1 > f-same
   $ hg ci -m2
   Invoking status precommit hook
   M f-different
   M f-same
   M f-unchanged-1
   M f-unchanged-2
   $ hg up -qr0
   $ echo tmp2 > f-unchanged-1
   $ echo tmp2 > f-unchanged-2
   $ echo tmp2 > f-same
   $ hg ci -m3
   Invoking status precommit hook
   M f-same
   M f-unchanged-1
   M f-unchanged-2
   created new head
   $ echo 1 > f-different
   $ echo 1 > f-unchanged-1
   $ echo 0 > f-unchanged-2
   $ echo 1 > f-same
   $ hg ci -m4
   Invoking status precommit hook
   M f-different
   M f-same
   M f-unchanged-1
   M f-unchanged-2
   $ hg merge
   largefile f-different has a merge conflict
   ancestor was 09d2af8dd22201dd8d48e5dcfcaed281ff9422c7
   keep (l)ocal e5fa44f2b31c1fb553b6021e7360d07d5d91ff5e or
   take (o)ther 7448d8798a4380162d4b56f9b452e2f6f9e24e7a? l
   getting changed largefiles
   1 largefiles updated, 0 removed
   0 files updated, 4 files merged, 0 files removed, 0 files unresolved
   (branch merge, don't forget to commit)
   $ cat f-different
   1
   $ cat f-same
   1
   $ cat f-unchanged-1
   1
   $ cat f-unchanged-2
   1
   $ cd ..
 
 Test largefile insulation (do not enabled a side effect
 ========================================================
 
 Check whether "largefiles" feature is supported only in repositories
 enabling largefiles extension.
 
   $ mkdir individualenabling
   $ cd individualenabling
 
   $ hg init enabledlocally
   $ echo large > enabledlocally/large
   $ hg -R enabledlocally add --large enabledlocally/large
   $ hg -R enabledlocally commit -m '#0'
   Invoking status precommit hook
   A large
 
   $ hg init notenabledlocally
   $ echo large > notenabledlocally/large
   $ hg -R notenabledlocally add --large notenabledlocally/large
   $ hg -R notenabledlocally commit -m '#0'
   Invoking status precommit hook
   A large
 
   $ cat >> $HGRCPATH <<EOF
   > [extensions]
   > # disable globally
   > largefiles=!
   > EOF
   $ cat >> enabledlocally/.hg/hgrc <<EOF
   > [extensions]
   > # enable locally
   > largefiles=
   > EOF
   $ hg -R enabledlocally root
   $TESTTMP/individualenabling/enabledlocally (glob)
   $ hg -R notenabledlocally root
   abort: repository requires features unknown to this Mercurial: largefiles!
   (see http://mercurial.selenic.com/wiki/MissingRequirement for more information)
   [255]
 
   $ hg init push-dst
   $ hg -R enabledlocally push push-dst
   pushing to push-dst
   abort: required features are not supported in the destination: largefiles
   [255]
 
   $ hg init pull-src
   $ hg -R pull-src pull enabledlocally
   pulling from enabledlocally
   abort: required features are not supported in the destination: largefiles
   [255]
 
   $ hg clone enabledlocally clone-dst
   abort: repository requires features unknown to this Mercurial: largefiles!
   (see http://mercurial.selenic.com/wiki/MissingRequirement for more information)
   [255]
   $ test -d clone-dst
   [1]
   $ hg clone --pull enabledlocally clone-pull-dst
   abort: required features are not supported in the destination: largefiles
   [255]
   $ test -d clone-pull-dst
   [1]
 
 #if serve
 
 Test largefiles specific peer setup, when largefiles is enabled
 locally (issue4109)
 
   $ hg showconfig extensions | grep largefiles
   extensions.largefiles=!
   $ mkdir -p $TESTTMP/individualenabling/usercache
 
   $ hg serve -R enabledlocally -d -p $HGPORT --pid-file hg.pid
   $ cat hg.pid >> $DAEMON_PIDS
 
   $ hg init pull-dst
   $ cat > pull-dst/.hg/hgrc <<EOF
   > [extensions]
   > # enable locally
   > largefiles=
   > [largefiles]
   > # ignore system cache to force largefiles specific wire proto access
   > usercache=$TESTTMP/individualenabling/usercache
   > EOF
   $ hg -R pull-dst -q pull -u http://localhost:$HGPORT
 
   $ "$TESTDIR/killdaemons.py" $DAEMON_PIDS
 #endif
 
   $ cd ..
 
 
 Test "pull --rebase" when rebase is enabled before largefiles (issue3861)
 =========================================================================
 
   $ hg showconfig extensions | grep largefiles
   extensions.largefiles=!
 
   $ mkdir issue3861
   $ cd issue3861
   $ hg init src
   $ hg clone -q src dst
   $ echo a > src/a
   $ hg -R src commit -Aqm "#0"
   Invoking status precommit hook
   A a
 
   $ cat >> dst/.hg/hgrc <<EOF
   > [extensions]
   > largefiles=
   > EOF
   $ hg -R dst pull --rebase
   pulling from $TESTTMP/issue3861/src (glob)
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
   nothing to rebase - working directory parent is already an ancestor of destination bf5e395ced2c
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
   $ cd ..