py3: use `x.hex()` instead of `pycompat.sysstr(node.hex(x))`

hgext/git/gitutil.py

@@ -1,50 +1,48
 """utilities to assist in working with pygit2"""
 
-from mercurial.node import bin, hex, sha1nodeconstants
-
-from mercurial import pycompat
+from mercurial.node import bin, sha1nodeconstants
 
 pygit2_module = None
 
 
 def get_pygit2():
     global pygit2_module
     if pygit2_module is None:
         try:
             import pygit2 as pygit2_module
 
             pygit2_module.InvalidSpecError
         except (ImportError, AttributeError):
             pass
     return pygit2_module
 
 
 def pygit2_version():
     mod = get_pygit2()
     v = "N/A"
 
     if mod:
         try:
             v = mod.__version__
         except AttributeError:
             pass
 
     return b"(pygit2 %s)" % v.encode("utf-8")
 
 
 def togitnode(n):
     """Wrapper to convert a Mercurial binary node to a unicode hexlified node.
 
     pygit2 and sqlite both need nodes as strings, not bytes.
     """
     assert len(n) == 20
-    return pycompat.sysstr(hex(n))
+    return n.hex()
 
 
 def fromgitnode(n):
     """Opposite of togitnode."""
     assert len(n) == 40
     return bin(n)
 
 
 nullgit = togitnode(sha1nodeconstants.nullid)

mercurial/branchmap.py

@@ -1,880 +1,880
 # branchmap.py - logic to computes, maintain and stores branchmap for local repo
 #
 # Copyright 2005-2007 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.
 
 
 import struct
 
 from .node import (
     bin,
     hex,
     nullrev,
 )
 from . import (
     encoding,
     error,
     obsolete,
     pycompat,
     scmutil,
     util,
 )
 from .utils import (
     repoviewutil,
     stringutil,
 )
 
 if pycompat.TYPE_CHECKING:
     from typing import (
         Any,
         Callable,
         Dict,
         Iterable,
         List,
         Optional,
         Set,
         Tuple,
         Union,
     )
     from . import localrepo
 
     assert any(
         (
             Any,
             Callable,
             Dict,
             Iterable,
             List,
             Optional,
             Set,
             Tuple,
             Union,
             localrepo,
         )
     )
 
 subsettable = repoviewutil.subsettable
 
 calcsize = struct.calcsize
 pack_into = struct.pack_into
 unpack_from = struct.unpack_from
 
 
 class BranchMapCache:
     """mapping of filtered views of repo with their branchcache"""
 
     def __init__(self):
         self._per_filter = {}
 
     def __getitem__(self, repo):
         self.updatecache(repo)
         return self._per_filter[repo.filtername]
 
     def updatecache(self, repo):
         """Update the cache for the given filtered view on a repository"""
         # This can trigger updates for the caches for subsets of the filtered
         # view, e.g. when there is no cache for this filtered view or the cache
         # is stale.
 
         cl = repo.changelog
         filtername = repo.filtername
         bcache = self._per_filter.get(filtername)
         if bcache is None or not bcache.validfor(repo):
             # cache object missing or cache object stale? Read from disk
             bcache = branchcache.fromfile(repo)
 
         revs = []
         if bcache is None:
             # no (fresh) cache available anymore, perhaps we can re-use
             # the cache for a subset, then extend that to add info on missing
             # revisions.
             subsetname = subsettable.get(filtername)
             if subsetname is not None:
                 subset = repo.filtered(subsetname)
                 bcache = self[subset].copy()
                 extrarevs = subset.changelog.filteredrevs - cl.filteredrevs
                 revs.extend(r for r in extrarevs if r <= bcache.tiprev)
             else:
                 # nothing to fall back on, start empty.
                 bcache = branchcache(repo)
 
         revs.extend(cl.revs(start=bcache.tiprev + 1))
         if revs:
             bcache.update(repo, revs)
 
         assert bcache.validfor(repo), filtername
         self._per_filter[repo.filtername] = bcache
 
     def replace(self, repo, remotebranchmap):
         """Replace the branchmap cache for a repo with a branch mapping.
 
         This is likely only called during clone with a branch map from a
         remote.
 
         """
         cl = repo.changelog
         clrev = cl.rev
         clbranchinfo = cl.branchinfo
         rbheads = []
         closed = set()
         for bheads in remotebranchmap.values():
             rbheads += bheads
             for h in bheads:
                 r = clrev(h)
                 b, c = clbranchinfo(r)
                 if c:
                     closed.add(h)
 
         if rbheads:
             rtiprev = max((int(clrev(node)) for node in rbheads))
             cache = branchcache(
                 repo,
                 remotebranchmap,
                 repo[rtiprev].node(),
                 rtiprev,
                 closednodes=closed,
             )
 
             # Try to stick it as low as possible
             # filter above served are unlikely to be fetch from a clone
             for candidate in (b'base', b'immutable', b'served'):
                 rview = repo.filtered(candidate)
                 if cache.validfor(rview):
                     self._per_filter[candidate] = cache
                     cache.write(rview)
                     return
 
     def clear(self):
         self._per_filter.clear()
 
     def write_delayed(self, repo):
         unfi = repo.unfiltered()
         for filtername, cache in self._per_filter.items():
             if cache._delayed:
                 repo = unfi.filtered(filtername)
                 cache.write(repo)
 
 
 def _unknownnode(node):
     """raises ValueError when branchcache found a node which does not exists"""
-    raise ValueError('node %s does not exist' % pycompat.sysstr(hex(node)))
+    raise ValueError('node %s does not exist' % node.hex())
 
 
 def _branchcachedesc(repo):
     if repo.filtername is not None:
         return b'branch cache (%s)' % repo.filtername
     else:
         return b'branch cache'
 
 
 class branchcache:
     """A dict like object that hold branches heads cache.
 
     This cache is used to avoid costly computations to determine all the
     branch heads of a repo.
 
     The cache is serialized on disk in the following format:
 
     <tip hex node> <tip rev number> [optional filtered repo hex hash]
     <branch head hex node> <open/closed state> <branch name>
     <branch head hex node> <open/closed state> <branch name>
     ...
 
     The first line is used to check if the cache is still valid. If the
     branch cache is for a filtered repo view, an optional third hash is
     included that hashes the hashes of all filtered and obsolete revisions.
 
     The open/closed state is represented by a single letter 'o' or 'c'.
     This field can be used to avoid changelog reads when determining if a
     branch head closes a branch or not.
     """
 
     def __init__(
         self,
         repo,
         entries=(),
         tipnode=None,
         tiprev=nullrev,
         filteredhash=None,
         closednodes=None,
         hasnode=None,
     ):
         # type: (localrepo.localrepository, Union[Dict[bytes, List[bytes]], Iterable[Tuple[bytes, List[bytes]]]], bytes,  int, Optional[bytes], Optional[Set[bytes]], Optional[Callable[[bytes], bool]]) -> None
         """hasnode is a function which can be used to verify whether changelog
         has a given node or not. If it's not provided, we assume that every node
         we have exists in changelog"""
         self._repo = repo
         self._delayed = False
         if tipnode is None:
             self.tipnode = repo.nullid
         else:
             self.tipnode = tipnode
         self.tiprev = tiprev
         self.filteredhash = filteredhash
         # closednodes is a set of nodes that close their branch. If the branch
         # cache has been updated, it may contain nodes that are no longer
         # heads.
         if closednodes is None:
             self._closednodes = set()
         else:
             self._closednodes = closednodes
         self._entries = dict(entries)
         # whether closed nodes are verified or not
         self._closedverified = False
         # branches for which nodes are verified
         self._verifiedbranches = set()
         self._hasnode = hasnode
         if self._hasnode is None:
             self._hasnode = lambda x: True
 
     def _verifyclosed(self):
         """verify the closed nodes we have"""
         if self._closedverified:
             return
         for node in self._closednodes:
             if not self._hasnode(node):
                 _unknownnode(node)
 
         self._closedverified = True
 
     def _verifybranch(self, branch):
         """verify head nodes for the given branch."""
         if branch not in self._entries or branch in self._verifiedbranches:
             return
         for n in self._entries[branch]:
             if not self._hasnode(n):
                 _unknownnode(n)
 
         self._verifiedbranches.add(branch)
 
     def _verifyall(self):
         """verifies nodes of all the branches"""
         needverification = set(self._entries.keys()) - self._verifiedbranches
         for b in needverification:
             self._verifybranch(b)
 
     def __iter__(self):
         return iter(self._entries)
 
     def __setitem__(self, key, value):
         self._entries[key] = value
 
     def __getitem__(self, key):
         self._verifybranch(key)
         return self._entries[key]
 
     def __contains__(self, key):
         self._verifybranch(key)
         return key in self._entries
 
     def iteritems(self):
         for k, v in self._entries.items():
             self._verifybranch(k)
             yield k, v
 
     items = iteritems
 
     def hasbranch(self, label):
         """checks whether a branch of this name exists or not"""
         self._verifybranch(label)
         return label in self._entries
 
     @classmethod
     def fromfile(cls, repo):
         f = None
         try:
             f = repo.cachevfs(cls._filename(repo))
             lineiter = iter(f)
             cachekey = next(lineiter).rstrip(b'\n').split(b" ", 2)
             last, lrev = cachekey[:2]
             last, lrev = bin(last), int(lrev)
             filteredhash = None
             hasnode = repo.changelog.hasnode
             if len(cachekey) > 2:
                 filteredhash = bin(cachekey[2])
             bcache = cls(
                 repo,
                 tipnode=last,
                 tiprev=lrev,
                 filteredhash=filteredhash,
                 hasnode=hasnode,
             )
             if not bcache.validfor(repo):
                 # invalidate the cache
                 raise ValueError('tip differs')
             bcache.load(repo, lineiter)
         except (IOError, OSError):
             return None
 
         except Exception as inst:
             if repo.ui.debugflag:
                 msg = b'invalid %s: %s\n'
                 repo.ui.debug(
                     msg
                     % (
                         _branchcachedesc(repo),
                         stringutil.forcebytestr(inst),
                     )
                 )
             bcache = None
 
         finally:
             if f:
                 f.close()
 
         return bcache
 
     def load(self, repo, lineiter):
         """fully loads the branchcache by reading from the file using the line
         iterator passed"""
         for line in lineiter:
             line = line.rstrip(b'\n')
             if not line:
                 continue
             node, state, label = line.split(b" ", 2)
             if state not in b'oc':
                 raise ValueError('invalid branch state')
             label = encoding.tolocal(label.strip())
             node = bin(node)
             self._entries.setdefault(label, []).append(node)
             if state == b'c':
                 self._closednodes.add(node)
 
     @staticmethod
     def _filename(repo):
         """name of a branchcache file for a given repo or repoview"""
         filename = b"branch2"
         if repo.filtername:
             filename = b'%s-%s' % (filename, repo.filtername)
         return filename
 
     def validfor(self, repo):
         """check that cache contents are valid for (a subset of) this repo
 
         - False when the order of changesets changed or if we detect a strip.
         - True when cache is up-to-date for the current repo or its subset."""
         try:
             node = repo.changelog.node(self.tiprev)
         except IndexError:
             # changesets were stripped and now we don't even have enough to
             # find tiprev
             return False
         if self.tipnode != node:
             # tiprev doesn't correspond to tipnode: repo was stripped, or this
             # repo has a different order of changesets
             return False
         tiphash = scmutil.filteredhash(repo, self.tiprev, needobsolete=True)
         # hashes don't match if this repo view has a different set of filtered
         # revisions (e.g. due to phase changes) or obsolete revisions (e.g.
         # history was rewritten)
         return self.filteredhash == tiphash
 
     def _branchtip(self, heads):
         """Return tuple with last open head in heads and false,
         otherwise return last closed head and true."""
         tip = heads[-1]
         closed = True
         for h in reversed(heads):
             if h not in self._closednodes:
                 tip = h
                 closed = False
                 break
         return tip, closed
 
     def branchtip(self, branch):
         """Return the tipmost open head on branch head, otherwise return the
         tipmost closed head on branch.
         Raise KeyError for unknown branch."""
         return self._branchtip(self[branch])[0]
 
     def iteropen(self, nodes):
         return (n for n in nodes if n not in self._closednodes)
 
     def branchheads(self, branch, closed=False):
         self._verifybranch(branch)
         heads = self._entries[branch]
         if not closed:
             heads = list(self.iteropen(heads))
         return heads
 
     def iterbranches(self):
         for bn, heads in self.items():
             yield (bn, heads) + self._branchtip(heads)
 
     def iterheads(self):
         """returns all the heads"""
         self._verifyall()
         return self._entries.values()
 
     def copy(self):
         """return an deep copy of the branchcache object"""
         return type(self)(
             self._repo,
             self._entries,
             self.tipnode,
             self.tiprev,
             self.filteredhash,
             self._closednodes,
         )
 
     def write(self, repo):
         tr = repo.currenttransaction()
         if not getattr(tr, 'finalized', True):
             # Avoid premature writing.
             #
             # (The cache warming setup by localrepo will update the file later.)
             self._delayed = True
             return
         try:
             filename = self._filename(repo)
             with repo.cachevfs(filename, b"w", atomictemp=True) as f:
                 cachekey = [hex(self.tipnode), b'%d' % self.tiprev]
                 if self.filteredhash is not None:
                     cachekey.append(hex(self.filteredhash))
                 f.write(b" ".join(cachekey) + b'\n')
                 nodecount = 0
                 for label, nodes in sorted(self._entries.items()):
                     label = encoding.fromlocal(label)
                     for node in nodes:
                         nodecount += 1
                         if node in self._closednodes:
                             state = b'c'
                         else:
                             state = b'o'
                         f.write(b"%s %s %s\n" % (hex(node), state, label))
             repo.ui.log(
                 b'branchcache',
                 b'wrote %s with %d labels and %d nodes\n',
                 _branchcachedesc(repo),
                 len(self._entries),
                 nodecount,
             )
             self._delayed = False
         except (IOError, OSError, error.Abort) as inst:
             # Abort may be raised by read only opener, so log and continue
             repo.ui.debug(
                 b"couldn't write branch cache: %s\n"
                 % stringutil.forcebytestr(inst)
             )
 
     def update(self, repo, revgen):
         """Given a branchhead cache, self, that may have extra nodes or be
         missing heads, and a generator of nodes that are strictly a superset of
         heads missing, this function updates self to be correct.
         """
         starttime = util.timer()
         cl = repo.changelog
         # collect new branch entries
         newbranches = {}
         getbranchinfo = repo.revbranchcache().branchinfo
         for r in revgen:
             branch, closesbranch = getbranchinfo(r)
             newbranches.setdefault(branch, []).append(r)
             if closesbranch:
                 self._closednodes.add(cl.node(r))
 
         # new tip revision which we found after iterating items from new
         # branches
         ntiprev = self.tiprev
 
         # Delay fetching the topological heads until they are needed.
         # A repository without non-continous branches can skip this part.
         topoheads = None
 
         # If a changeset is visible, its parents must be visible too, so
         # use the faster unfiltered parent accessor.
         parentrevs = repo.unfiltered().changelog.parentrevs
 
         # Faster than using ctx.obsolete()
         obsrevs = obsolete.getrevs(repo, b'obsolete')
 
         for branch, newheadrevs in newbranches.items():
             # For every branch, compute the new branchheads.
             # A branchhead is a revision such that no descendant is on
             # the same branch.
             #
             # The branchheads are computed iteratively in revision order.
             # This ensures topological order, i.e. parents are processed
             # before their children. Ancestors are inclusive here, i.e.
             # any revision is an ancestor of itself.
             #
             # Core observations:
             # - The current revision is always a branchhead for the
             #   repository up to that point.
             # - It is the first revision of the branch if and only if
             #   there was no branchhead before. In that case, it is the
             #   only branchhead as there are no possible ancestors on
             #   the same branch.
             # - If a parent is on the same branch, a branchhead can
             #   only be an ancestor of that parent, if it is parent
             #   itself. Otherwise it would have been removed as ancestor
             #   of that parent before.
             # - Therefore, if all parents are on the same branch, they
             #   can just be removed from the branchhead set.
             # - If one parent is on the same branch and the other is not
             #   and there was exactly one branchhead known, the existing
             #   branchhead can only be an ancestor if it is the parent.
             #   Otherwise it would have been removed as ancestor of
             #   the parent before. The other parent therefore can't have
             #   a branchhead as ancestor.
             # - In all other cases, the parents on different branches
             #   could have a branchhead as ancestor. Those parents are
             #   kept in the "uncertain" set. If all branchheads are also
             #   topological heads, they can't have descendants and further
             #   checks can be skipped. Otherwise, the ancestors of the
             #   "uncertain" set are removed from branchheads.
             #   This computation is heavy and avoided if at all possible.
             bheads = self._entries.get(branch, [])
             bheadset = {cl.rev(node) for node in bheads}
             uncertain = set()
             for newrev in sorted(newheadrevs):
                 if newrev in obsrevs:
                     # We ignore obsolete changesets as they shouldn't be
                     # considered heads.
                     continue
 
                 if not bheadset:
                     bheadset.add(newrev)
                     continue
 
                 parents = [p for p in parentrevs(newrev) if p != nullrev]
                 samebranch = set()
                 otherbranch = set()
                 obsparents = set()
                 for p in parents:
                     if p in obsrevs:
                         # We ignored this obsolete changeset earlier, but now
                         # that it has non-ignored children, we need to make
                         # sure their ancestors are not considered heads. To
                         # achieve that, we will simply treat this obsolete
                         # changeset as a parent from other branch.
                         obsparents.add(p)
                     elif p in bheadset or getbranchinfo(p)[0] == branch:
                         samebranch.add(p)
                     else:
                         otherbranch.add(p)
                 if not (len(bheadset) == len(samebranch) == 1):
                     uncertain.update(otherbranch)
                     uncertain.update(obsparents)
                 bheadset.difference_update(samebranch)
                 bheadset.add(newrev)
 
             if uncertain:
                 if topoheads is None:
                     topoheads = set(cl.headrevs())
                 if bheadset - topoheads:
                     floorrev = min(bheadset)
                     if floorrev <= max(uncertain):
                         ancestors = set(cl.ancestors(uncertain, floorrev))
                         bheadset -= ancestors
             if bheadset:
                 self[branch] = [cl.node(rev) for rev in sorted(bheadset)]
             tiprev = max(newheadrevs)
             if tiprev > ntiprev:
                 ntiprev = tiprev
 
         if ntiprev > self.tiprev:
             self.tiprev = ntiprev
             self.tipnode = cl.node(ntiprev)
 
         if not self.validfor(repo):
             # old cache key is now invalid for the repo, but we've just updated
             # the cache and we assume it's valid, so let's make the cache key
             # valid as well by recomputing it from the cached data
             self.tipnode = repo.nullid
             self.tiprev = nullrev
             for heads in self.iterheads():
                 if not heads:
                     # all revisions on a branch are obsolete
                     continue
                 # note: tiprev is not necessarily the tip revision of repo,
                 # because the tip could be obsolete (i.e. not a head)
                 tiprev = max(cl.rev(node) for node in heads)
                 if tiprev > self.tiprev:
                     self.tipnode = cl.node(tiprev)
                     self.tiprev = tiprev
         self.filteredhash = scmutil.filteredhash(
             repo, self.tiprev, needobsolete=True
         )
 
         duration = util.timer() - starttime
         repo.ui.log(
             b'branchcache',
             b'updated %s in %.4f seconds\n',
             _branchcachedesc(repo),
             duration,
         )
 
         self.write(repo)
 
 
 class remotebranchcache(branchcache):
     """Branchmap info for a remote connection, should not write locally"""
 
     def write(self, repo):
         pass
 
 
 # Revision branch info cache
 
 _rbcversion = b'-v1'
 _rbcnames = b'rbc-names' + _rbcversion
 _rbcrevs = b'rbc-revs' + _rbcversion
 # [4 byte hash prefix][4 byte branch name number with sign bit indicating open]
 _rbcrecfmt = b'>4sI'
 _rbcrecsize = calcsize(_rbcrecfmt)
 _rbcmininc = 64 * _rbcrecsize
 _rbcnodelen = 4
 _rbcbranchidxmask = 0x7FFFFFFF
 _rbccloseflag = 0x80000000
 
 
 class revbranchcache:
     """Persistent cache, mapping from revision number to branch name and close.
     This is a low level cache, independent of filtering.
 
     Branch names are stored in rbc-names in internal encoding separated by 0.
     rbc-names is append-only, and each branch name is only stored once and will
     thus have a unique index.
 
     The branch info for each revision is stored in rbc-revs as constant size
     records. The whole file is read into memory, but it is only 'parsed' on
     demand. The file is usually append-only but will be truncated if repo
     modification is detected.
     The record for each revision contains the first 4 bytes of the
     corresponding node hash, and the record is only used if it still matches.
     Even a completely trashed rbc-revs fill thus still give the right result
     while converging towards full recovery ... assuming no incorrectly matching
     node hashes.
     The record also contains 4 bytes where 31 bits contains the index of the
     branch and the last bit indicate that it is a branch close commit.
     The usage pattern for rbc-revs is thus somewhat similar to 00changelog.i
     and will grow with it but be 1/8th of its size.
     """
 
     def __init__(self, repo, readonly=True):
         assert repo.filtername is None
         self._repo = repo
         self._names = []  # branch names in local encoding with static index
         self._rbcrevs = bytearray()
         self._rbcsnameslen = 0  # length of names read at _rbcsnameslen
         try:
             bndata = repo.cachevfs.read(_rbcnames)
             self._rbcsnameslen = len(bndata)  # for verification before writing
             if bndata:
                 self._names = [
                     encoding.tolocal(bn) for bn in bndata.split(b'\0')
                 ]
         except (IOError, OSError):
             if readonly:
                 # don't try to use cache - fall back to the slow path
                 self.branchinfo = self._branchinfo
 
         if self._names:
             try:
                 data = repo.cachevfs.read(_rbcrevs)
                 self._rbcrevs[:] = data
             except (IOError, OSError) as inst:
                 repo.ui.debug(
                     b"couldn't read revision branch cache: %s\n"
                     % stringutil.forcebytestr(inst)
                 )
         # remember number of good records on disk
         self._rbcrevslen = min(
             len(self._rbcrevs) // _rbcrecsize, len(repo.changelog)
         )
         if self._rbcrevslen == 0:
             self._names = []
         self._rbcnamescount = len(self._names)  # number of names read at
         # _rbcsnameslen
 
     def _clear(self):
         self._rbcsnameslen = 0
         del self._names[:]
         self._rbcnamescount = 0
         self._rbcrevslen = len(self._repo.changelog)
         self._rbcrevs = bytearray(self._rbcrevslen * _rbcrecsize)
         util.clearcachedproperty(self, b'_namesreverse')
 
     @util.propertycache
     def _namesreverse(self):
         return {b: r for r, b in enumerate(self._names)}
 
     def branchinfo(self, rev):
         """Return branch name and close flag for rev, using and updating
         persistent cache."""
         changelog = self._repo.changelog
         rbcrevidx = rev * _rbcrecsize
 
         # avoid negative index, changelog.read(nullrev) is fast without cache
         if rev == nullrev:
             return changelog.branchinfo(rev)
 
         # if requested rev isn't allocated, grow and cache the rev info
         if len(self._rbcrevs) < rbcrevidx + _rbcrecsize:
             return self._branchinfo(rev)
 
         # fast path: extract data from cache, use it if node is matching
         reponode = changelog.node(rev)[:_rbcnodelen]
         cachenode, branchidx = unpack_from(
             _rbcrecfmt, util.buffer(self._rbcrevs), rbcrevidx
         )
         close = bool(branchidx & _rbccloseflag)
         if close:
             branchidx &= _rbcbranchidxmask
         if cachenode == b'\0\0\0\0':
             pass
         elif cachenode == reponode:
             try:
                 return self._names[branchidx], close
             except IndexError:
                 # recover from invalid reference to unknown branch
                 self._repo.ui.debug(
                     b"referenced branch names not found"
                     b" - rebuilding revision branch cache from scratch\n"
                 )
                 self._clear()
         else:
             # rev/node map has changed, invalidate the cache from here up
             self._repo.ui.debug(
                 b"history modification detected - truncating "
                 b"revision branch cache to revision %d\n" % rev
             )
             truncate = rbcrevidx + _rbcrecsize
             del self._rbcrevs[truncate:]
             self._rbcrevslen = min(self._rbcrevslen, truncate)
 
         # fall back to slow path and make sure it will be written to disk
         return self._branchinfo(rev)
 
     def _branchinfo(self, rev):
         """Retrieve branch info from changelog and update _rbcrevs"""
         changelog = self._repo.changelog
         b, close = changelog.branchinfo(rev)
         if b in self._namesreverse:
             branchidx = self._namesreverse[b]
         else:
             branchidx = len(self._names)
             self._names.append(b)
             self._namesreverse[b] = branchidx
         reponode = changelog.node(rev)
         if close:
             branchidx |= _rbccloseflag
         self._setcachedata(rev, reponode, branchidx)
         return b, close
 
     def setdata(self, rev, changelogrevision):
         """add new data information to the cache"""
         branch, close = changelogrevision.branchinfo
 
         if branch in self._namesreverse:
             branchidx = self._namesreverse[branch]
         else:
             branchidx = len(self._names)
             self._names.append(branch)
             self._namesreverse[branch] = branchidx
         if close:
             branchidx |= _rbccloseflag
         self._setcachedata(rev, self._repo.changelog.node(rev), branchidx)
         # If no cache data were readable (non exists, bad permission, etc)
         # the cache was bypassing itself by setting:
         #
         #   self.branchinfo = self._branchinfo
         #
         # Since we now have data in the cache, we need to drop this bypassing.
         if 'branchinfo' in vars(self):
             del self.branchinfo
 
     def _setcachedata(self, rev, node, branchidx):
         """Writes the node's branch data to the in-memory cache data."""
         if rev == nullrev:
             return
         rbcrevidx = rev * _rbcrecsize
         requiredsize = rbcrevidx + _rbcrecsize
         rbccur = len(self._rbcrevs)
         if rbccur < requiredsize:
             # bytearray doesn't allocate extra space at least in Python 3.7.
             # When multiple changesets are added in a row, precise resize would
             # result in quadratic complexity. Overallocate to compensate by
             # use the classic doubling technique for dynamic arrays instead.
             # If there was a gap in the map before, less space will be reserved.
             self._rbcrevs.extend(b'\0' * max(_rbcmininc, requiredsize))
         pack_into(_rbcrecfmt, self._rbcrevs, rbcrevidx, node, branchidx)
         self._rbcrevslen = min(self._rbcrevslen, rev)
 
         tr = self._repo.currenttransaction()
         if tr:
             tr.addfinalize(b'write-revbranchcache', self.write)
 
     def write(self, tr=None):
         """Save branch cache if it is dirty."""
         repo = self._repo
         wlock = None
         step = b''
         try:
             # write the new names
             if self._rbcnamescount < len(self._names):
                 wlock = repo.wlock(wait=False)
                 step = b' names'
                 self._writenames(repo)
 
             # write the new revs
             start = self._rbcrevslen * _rbcrecsize
             if start != len(self._rbcrevs):
                 step = b''
                 if wlock is None:
                     wlock = repo.wlock(wait=False)
                 self._writerevs(repo, start)
 
         except (IOError, OSError, error.Abort, error.LockError) as inst:
             repo.ui.debug(
                 b"couldn't write revision branch cache%s: %s\n"
                 % (step, stringutil.forcebytestr(inst))
             )
         finally:
             if wlock is not None:
                 wlock.release()
 
     def _writenames(self, repo):
         """write the new branch names to revbranchcache"""
         if self._rbcnamescount != 0:
             f = repo.cachevfs.open(_rbcnames, b'ab')
             if f.tell() == self._rbcsnameslen:
                 f.write(b'\0')
             else:
                 f.close()
                 repo.ui.debug(b"%s changed - rewriting it\n" % _rbcnames)
                 self._rbcnamescount = 0
                 self._rbcrevslen = 0
         if self._rbcnamescount == 0:
             # before rewriting names, make sure references are removed
             repo.cachevfs.unlinkpath(_rbcrevs, ignoremissing=True)
             f = repo.cachevfs.open(_rbcnames, b'wb')
         f.write(
             b'\0'.join(
                 encoding.fromlocal(b)
                 for b in self._names[self._rbcnamescount :]
             )
         )
         self._rbcsnameslen = f.tell()
         f.close()
         self._rbcnamescount = len(self._names)
 
     def _writerevs(self, repo, start):
         """write the new revs to revbranchcache"""
         revs = min(len(repo.changelog), len(self._rbcrevs) // _rbcrecsize)
         with repo.cachevfs.open(_rbcrevs, b'ab') as f:
             if f.tell() != start:
                 repo.ui.debug(
                     b"truncating cache/%s to %d\n" % (_rbcrevs, start)
                 )
                 f.seek(start)
                 if f.tell() != start:
                     start = 0
                     f.seek(start)
                 f.truncate()
             end = revs * _rbcrecsize
             f.write(self._rbcrevs[start:end])
         self._rbcrevslen = revs

mercurial/obsolete.py

@@ -1,1149 +1,1147
 # obsolete.py - obsolete markers handling
 #
 # Copyright 2012 Pierre-Yves David <[e-mail]>
 #                Logilab SA        <[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.
 
 """Obsolete marker handling
 
 An obsolete marker maps an old changeset to a list of new
 changesets. If the list of new changesets is empty, the old changeset
 is said to be "killed". Otherwise, the old changeset is being
 "replaced" by the new changesets.
 
 Obsolete markers can be used to record and distribute changeset graph
 transformations performed by history rewrite operations, and help
 building new tools to reconcile conflicting rewrite actions. To
 facilitate conflict resolution, markers include various annotations
 besides old and news changeset identifiers, such as creation date or
 author name.
 
 The old obsoleted changeset is called a "predecessor" and possible
 replacements are called "successors". Markers that used changeset X as
 a predecessor are called "successor markers of X" because they hold
 information about the successors of X. Markers that use changeset Y as
 a successors are call "predecessor markers of Y" because they hold
 information about the predecessors of Y.
 
 Examples:
 
 - When changeset A is replaced by changeset A', one marker is stored:
 
     (A, (A',))
 
 - When changesets A and B are folded into a new changeset C, two markers are
   stored:
 
     (A, (C,)) and (B, (C,))
 
 - When changeset A is simply "pruned" from the graph, a marker is created:
 
     (A, ())
 
 - When changeset A is split into B and C, a single marker is used:
 
     (A, (B, C))
 
   We use a single marker to distinguish the "split" case from the "divergence"
   case. If two independent operations rewrite the same changeset A in to A' and
   A'', we have an error case: divergent rewriting. We can detect it because
   two markers will be created independently:
 
   (A, (B,)) and (A, (C,))
 
 Format
 ------
 
 Markers are stored in an append-only file stored in
 '.hg/store/obsstore'.
 
 The file starts with a version header:
 
 - 1 unsigned byte: version number, starting at zero.
 
 The header is followed by the markers. Marker format depend of the version. See
 comment associated with each format for details.
 
 """
 
 import binascii
 import errno
 import struct
 
 from .i18n import _
 from .pycompat import getattr
 from .node import (
     bin,
     hex,
 )
 from . import (
     encoding,
     error,
     obsutil,
     phases,
     policy,
     pycompat,
     util,
 )
 from .utils import (
     dateutil,
     hashutil,
 )
 
 parsers = policy.importmod('parsers')
 
 _pack = struct.pack
 _unpack = struct.unpack
 _calcsize = struct.calcsize
 propertycache = util.propertycache
 
 # Options for obsolescence
 createmarkersopt = b'createmarkers'
 allowunstableopt = b'allowunstable'
 allowdivergenceopt = b'allowdivergence'
 exchangeopt = b'exchange'
 
 
 def _getoptionvalue(repo, option):
     """Returns True if the given repository has the given obsolete option
     enabled.
     """
     configkey = b'evolution.%s' % option
     newconfig = repo.ui.configbool(b'experimental', configkey)
 
     # Return the value only if defined
     if newconfig is not None:
         return newconfig
 
     # Fallback on generic option
     try:
         return repo.ui.configbool(b'experimental', b'evolution')
     except (error.ConfigError, AttributeError):
         # Fallback on old-fashion config
         # inconsistent config: experimental.evolution
         result = set(repo.ui.configlist(b'experimental', b'evolution'))
 
         if b'all' in result:
             return True
 
         # Temporary hack for next check
         newconfig = repo.ui.config(b'experimental', b'evolution.createmarkers')
         if newconfig:
             result.add(b'createmarkers')
 
         return option in result
 
 
 def getoptions(repo):
     """Returns dicts showing state of obsolescence features."""
 
     createmarkersvalue = _getoptionvalue(repo, createmarkersopt)
     if createmarkersvalue:
         unstablevalue = _getoptionvalue(repo, allowunstableopt)
         divergencevalue = _getoptionvalue(repo, allowdivergenceopt)
         exchangevalue = _getoptionvalue(repo, exchangeopt)
     else:
         # if we cannot create obsolescence markers, we shouldn't exchange them
         # or perform operations that lead to instability or divergence
         unstablevalue = False
         divergencevalue = False
         exchangevalue = False
 
     return {
         createmarkersopt: createmarkersvalue,
         allowunstableopt: unstablevalue,
         allowdivergenceopt: divergencevalue,
         exchangeopt: exchangevalue,
     }
 
 
 def isenabled(repo, option):
     """Returns True if the given repository has the given obsolete option
     enabled.
     """
     return getoptions(repo)[option]
 
 
 # Creating aliases for marker flags because evolve extension looks for
 # bumpedfix in obsolete.py
 bumpedfix = obsutil.bumpedfix
 usingsha256 = obsutil.usingsha256
 
 ## Parsing and writing of version "0"
 #
 # The header is followed by the markers. Each marker is made of:
 #
 # - 1 uint8 : number of new changesets "N", can be zero.
 #
 # - 1 uint32: metadata size "M" in bytes.
 #
 # - 1 byte: a bit field. It is reserved for flags used in common
 #   obsolete marker operations, to avoid repeated decoding of metadata
 #   entries.
 #
 # - 20 bytes: obsoleted changeset identifier.
 #
 # - N*20 bytes: new changesets identifiers.
 #
 # - M bytes: metadata as a sequence of nul-terminated strings. Each
 #   string contains a key and a value, separated by a colon ':', without
 #   additional encoding. Keys cannot contain '\0' or ':' and values
 #   cannot contain '\0'.
 _fm0version = 0
 _fm0fixed = b'>BIB20s'
 _fm0node = b'20s'
 _fm0fsize = _calcsize(_fm0fixed)
 _fm0fnodesize = _calcsize(_fm0node)
 
 
 def _fm0readmarkers(data, off, stop):
     # Loop on markers
     while off < stop:
         # read fixed part
         cur = data[off : off + _fm0fsize]
         off += _fm0fsize
         numsuc, mdsize, flags, pre = _unpack(_fm0fixed, cur)
         # read replacement
         sucs = ()
         if numsuc:
             s = _fm0fnodesize * numsuc
             cur = data[off : off + s]
             sucs = _unpack(_fm0node * numsuc, cur)
             off += s
         # read metadata
         # (metadata will be decoded on demand)
         metadata = data[off : off + mdsize]
         if len(metadata) != mdsize:
             raise error.Abort(
                 _(
                     b'parsing obsolete marker: metadata is too '
                     b'short, %d bytes expected, got %d'
                 )
                 % (mdsize, len(metadata))
             )
         off += mdsize
         metadata = _fm0decodemeta(metadata)
         try:
             when, offset = metadata.pop(b'date', b'0 0').split(b' ')
             date = float(when), int(offset)
         except ValueError:
             date = (0.0, 0)
         parents = None
         if b'p2' in metadata:
             parents = (metadata.pop(b'p1', None), metadata.pop(b'p2', None))
         elif b'p1' in metadata:
             parents = (metadata.pop(b'p1', None),)
         elif b'p0' in metadata:
             parents = ()
         if parents is not None:
             try:
                 parents = tuple(bin(p) for p in parents)
                 # if parent content is not a nodeid, drop the data
                 for p in parents:
                     if len(p) != 20:
                         parents = None
                         break
             except binascii.Error:
                 # if content cannot be translated to nodeid drop the data.
                 parents = None
 
         metadata = tuple(sorted(metadata.items()))
 
         yield (pre, sucs, flags, metadata, date, parents)
 
 
 def _fm0encodeonemarker(marker):
     pre, sucs, flags, metadata, date, parents = marker
     if flags & usingsha256:
         raise error.Abort(_(b'cannot handle sha256 with old obsstore format'))
     metadata = dict(metadata)
     time, tz = date
     metadata[b'date'] = b'%r %i' % (time, tz)
     if parents is not None:
         if not parents:
             # mark that we explicitly recorded no parents
             metadata[b'p0'] = b''
         for i, p in enumerate(parents, 1):
             metadata[b'p%i' % i] = hex(p)
     metadata = _fm0encodemeta(metadata)
     numsuc = len(sucs)
     format = _fm0fixed + (_fm0node * numsuc)
     data = [numsuc, len(metadata), flags, pre]
     data.extend(sucs)
     return _pack(format, *data) + metadata
 
 
 def _fm0encodemeta(meta):
     """Return encoded metadata string to string mapping.
 
     Assume no ':' in key and no '\0' in both key and value."""
     for key, value in meta.items():
         if b':' in key or b'\0' in key:
             raise ValueError(b"':' and '\0' are forbidden in metadata key'")
         if b'\0' in value:
             raise ValueError(b"':' is forbidden in metadata value'")
     return b'\0'.join([b'%s:%s' % (k, meta[k]) for k in sorted(meta)])
 
 
 def _fm0decodemeta(data):
     """Return string to string dictionary from encoded version."""
     d = {}
     for l in data.split(b'\0'):
         if l:
             key, value = l.split(b':', 1)
             d[key] = value
     return d
 
 
 ## Parsing and writing of version "1"
 #
 # The header is followed by the markers. Each marker is made of:
 #
 # - uint32: total size of the marker (including this field)
 #
 # - float64: date in seconds since epoch
 #
 # - int16: timezone offset in minutes
 #
 # - uint16: a bit field. It is reserved for flags used in common
 #   obsolete marker operations, to avoid repeated decoding of metadata
 #   entries.
 #
 # - uint8: number of successors "N", can be zero.
 #
 # - uint8: number of parents "P", can be zero.
 #
 #     0: parents data stored but no parent,
 #     1: one parent stored,
 #     2: two parents stored,
 #     3: no parent data stored
 #
 # - uint8: number of metadata entries M
 #
 # - 20 or 32 bytes: predecessor changeset identifier.
 #
 # - N*(20 or 32) bytes: successors changesets identifiers.
 #
 # - P*(20 or 32) bytes: parents of the predecessors changesets.
 #
 # - M*(uint8, uint8): size of all metadata entries (key and value)
 #
 # - remaining bytes: the metadata, each (key, value) pair after the other.
 _fm1version = 1
 _fm1fixed = b'>IdhHBBB'
 _fm1nodesha1 = b'20s'
 _fm1nodesha256 = b'32s'
 _fm1nodesha1size = _calcsize(_fm1nodesha1)
 _fm1nodesha256size = _calcsize(_fm1nodesha256)
 _fm1fsize = _calcsize(_fm1fixed)
 _fm1parentnone = 3
 _fm1metapair = b'BB'
 _fm1metapairsize = _calcsize(_fm1metapair)
 
 
 def _fm1purereadmarkers(data, off, stop):
     # make some global constants local for performance
     noneflag = _fm1parentnone
     sha2flag = usingsha256
     sha1size = _fm1nodesha1size
     sha2size = _fm1nodesha256size
     sha1fmt = _fm1nodesha1
     sha2fmt = _fm1nodesha256
     metasize = _fm1metapairsize
     metafmt = _fm1metapair
     fsize = _fm1fsize
     unpack = _unpack
 
     # Loop on markers
     ufixed = struct.Struct(_fm1fixed).unpack
 
     while off < stop:
         # read fixed part
         o1 = off + fsize
         t, secs, tz, flags, numsuc, numpar, nummeta = ufixed(data[off:o1])
 
         if flags & sha2flag:
             nodefmt = sha2fmt
             nodesize = sha2size
         else:
             nodefmt = sha1fmt
             nodesize = sha1size
 
         (prec,) = unpack(nodefmt, data[o1 : o1 + nodesize])
         o1 += nodesize
 
         # read 0 or more successors
         if numsuc == 1:
             o2 = o1 + nodesize
             sucs = (data[o1:o2],)
         else:
             o2 = o1 + nodesize * numsuc
             sucs = unpack(nodefmt * numsuc, data[o1:o2])
 
         # read parents
         if numpar == noneflag:
             o3 = o2
             parents = None
         elif numpar == 1:
             o3 = o2 + nodesize
             parents = (data[o2:o3],)
         else:
             o3 = o2 + nodesize * numpar
             parents = unpack(nodefmt * numpar, data[o2:o3])
 
         # read metadata
         off = o3 + metasize * nummeta
         metapairsize = unpack(b'>' + (metafmt * nummeta), data[o3:off])
         metadata = []
         for idx in range(0, len(metapairsize), 2):
             o1 = off + metapairsize[idx]
             o2 = o1 + metapairsize[idx + 1]
             metadata.append((data[off:o1], data[o1:o2]))
             off = o2
 
         yield (prec, sucs, flags, tuple(metadata), (secs, tz * 60), parents)
 
 
 def _fm1encodeonemarker(marker):
     pre, sucs, flags, metadata, date, parents = marker
     # determine node size
     _fm1node = _fm1nodesha1
     if flags & usingsha256:
         _fm1node = _fm1nodesha256
     numsuc = len(sucs)
     numextranodes = 1 + numsuc
     if parents is None:
         numpar = _fm1parentnone
     else:
         numpar = len(parents)
         numextranodes += numpar
     formatnodes = _fm1node * numextranodes
     formatmeta = _fm1metapair * len(metadata)
     format = _fm1fixed + formatnodes + formatmeta
     # tz is stored in minutes so we divide by 60
     tz = date[1] // 60
     data = [None, date[0], tz, flags, numsuc, numpar, len(metadata), pre]
     data.extend(sucs)
     if parents is not None:
         data.extend(parents)
     totalsize = _calcsize(format)
     for key, value in metadata:
         lk = len(key)
         lv = len(value)
         if lk > 255:
             msg = (
                 b'obsstore metadata key cannot be longer than 255 bytes'
                 b' (key "%s" is %u bytes)'
             ) % (key, lk)
             raise error.ProgrammingError(msg)
         if lv > 255:
             msg = (
                 b'obsstore metadata value cannot be longer than 255 bytes'
                 b' (value "%s" for key "%s" is %u bytes)'
             ) % (value, key, lv)
             raise error.ProgrammingError(msg)
         data.append(lk)
         data.append(lv)
         totalsize += lk + lv
     data[0] = totalsize
     data = [_pack(format, *data)]
     for key, value in metadata:
         data.append(key)
         data.append(value)
     return b''.join(data)
 
 
 def _fm1readmarkers(data, off, stop):
     native = getattr(parsers, 'fm1readmarkers', None)
     if not native:
         return _fm1purereadmarkers(data, off, stop)
     return native(data, off, stop)
 
 
 # mapping to read/write various marker formats
 # <version> -> (decoder, encoder)
 formats = {
     _fm0version: (_fm0readmarkers, _fm0encodeonemarker),
     _fm1version: (_fm1readmarkers, _fm1encodeonemarker),
 }
 
 
 def _readmarkerversion(data):
     return _unpack(b'>B', data[0:1])[0]
 
 
 @util.nogc
 def _readmarkers(data, off=None, stop=None):
     """Read and enumerate markers from raw data"""
     diskversion = _readmarkerversion(data)
     if not off:
         off = 1  # skip 1 byte version number
     if stop is None:
         stop = len(data)
     if diskversion not in formats:
         msg = _(b'parsing obsolete marker: unknown version %r') % diskversion
         raise error.UnknownVersion(msg, version=diskversion)
     return diskversion, formats[diskversion][0](data, off, stop)
 
 
 def encodeheader(version=_fm0version):
     return _pack(b'>B', version)
 
 
 def encodemarkers(markers, addheader=False, version=_fm0version):
     # Kept separate from flushmarkers(), it will be reused for
     # markers exchange.
     encodeone = formats[version][1]
     if addheader:
         yield encodeheader(version)
     for marker in markers:
         yield encodeone(marker)
 
 
 @util.nogc
 def _addsuccessors(successors, markers):
     for mark in markers:
         successors.setdefault(mark[0], set()).add(mark)
 
 
 @util.nogc
 def _addpredecessors(predecessors, markers):
     for mark in markers:
         for suc in mark[1]:
             predecessors.setdefault(suc, set()).add(mark)
 
 
 @util.nogc
 def _addchildren(children, markers):
     for mark in markers:
         parents = mark[5]
         if parents is not None:
             for p in parents:
                 children.setdefault(p, set()).add(mark)
 
 
 def _checkinvalidmarkers(repo, markers):
     """search for marker with invalid data and raise error if needed
 
     Exist as a separated function to allow the evolve extension for a more
     subtle handling.
     """
     for mark in markers:
         if repo.nullid in mark[1]:
             raise error.Abort(
                 _(
                     b'bad obsolescence marker detected: '
                     b'invalid successors nullid'
                 )
             )
 
 
 class obsstore:
     """Store obsolete markers
 
     Markers can be accessed with two mappings:
     - predecessors[x] -> set(markers on predecessors edges of x)
     - successors[x] -> set(markers on successors edges of x)
     - children[x]   -> set(markers on predecessors edges of children(x)
     """
 
     fields = (b'prec', b'succs', b'flag', b'meta', b'date', b'parents')
     # prec:    nodeid, predecessors changesets
     # succs:   tuple of nodeid, successor changesets (0-N length)
     # flag:    integer, flag field carrying modifier for the markers (see doc)
     # meta:    binary blob in UTF-8, encoded metadata dictionary
     # date:    (float, int) tuple, date of marker creation
     # parents: (tuple of nodeid) or None, parents of predecessors
     #          None is used when no data has been recorded
 
     def __init__(self, repo, svfs, defaultformat=_fm1version, readonly=False):
         # caches for various obsolescence related cache
         self.caches = {}
         self.svfs = svfs
         self.repo = repo
         self._defaultformat = defaultformat
         self._readonly = readonly
 
     def __iter__(self):
         return iter(self._all)
 
     def __len__(self):
         return len(self._all)
 
     def __nonzero__(self):
         from . import statichttprepo
 
         if isinstance(self.repo, statichttprepo.statichttprepository):
             # If repo is accessed via static HTTP, then we can't use os.stat()
             # to just peek at the file size.
             return len(self._data) > 1
         if not self._cached('_all'):
             try:
                 return self.svfs.stat(b'obsstore').st_size > 1
             except OSError as inst:
                 if inst.errno != errno.ENOENT:
                     raise
                 # just build an empty _all list if no obsstore exists, which
                 # avoids further stat() syscalls
         return bool(self._all)
 
     __bool__ = __nonzero__
 
     @property
     def readonly(self):
         """True if marker creation is disabled
 
         Remove me in the future when obsolete marker is always on."""
         return self._readonly
 
     def create(
         self,
         transaction,
         prec,
         succs=(),
         flag=0,
         parents=None,
         date=None,
         metadata=None,
         ui=None,
     ):
         """obsolete: add a new obsolete marker
 
         * ensuring it is hashable
         * check mandatory metadata
         * encode metadata
 
         If you are a human writing code creating marker you want to use the
         `createmarkers` function in this module instead.
 
         return True if a new marker have been added, False if the markers
         already existed (no op).
         """
         flag = int(flag)
         if metadata is None:
             metadata = {}
         if date is None:
             if b'date' in metadata:
                 # as a courtesy for out-of-tree extensions
                 date = dateutil.parsedate(metadata.pop(b'date'))
             elif ui is not None:
                 date = ui.configdate(b'devel', b'default-date')
                 if date is None:
                     date = dateutil.makedate()
             else:
                 date = dateutil.makedate()
         if flag & usingsha256:
             if len(prec) != 32:
                 raise ValueError(prec)
             for succ in succs:
                 if len(succ) != 32:
                     raise ValueError(succ)
         else:
             if len(prec) != 20:
                 raise ValueError(prec)
             for succ in succs:
                 if len(succ) != 20:
                     raise ValueError(succ)
         if prec in succs:
-            raise ValueError(
-                'in-marker cycle with %s' % pycompat.sysstr(hex(prec))
-            )
+            raise ValueError('in-marker cycle with %s' % prec.hex())
 
         metadata = tuple(sorted(metadata.items()))
         for k, v in metadata:
             try:
                 # might be better to reject non-ASCII keys
                 k.decode('utf-8')
                 v.decode('utf-8')
             except UnicodeDecodeError:
                 raise error.ProgrammingError(
                     b'obsstore metadata must be valid UTF-8 sequence '
                     b'(key = %r, value = %r)'
                     % (pycompat.bytestr(k), pycompat.bytestr(v))
                 )
 
         marker = (bytes(prec), tuple(succs), flag, metadata, date, parents)
         return bool(self.add(transaction, [marker]))
 
     def add(self, transaction, markers):
         """Add new markers to the store
 
         Take care of filtering duplicate.
         Return the number of new marker."""
         if self._readonly:
             raise error.Abort(
                 _(b'creating obsolete markers is not enabled on this repo')
             )
         known = set()
         getsuccessors = self.successors.get
         new = []
         for m in markers:
             if m not in getsuccessors(m[0], ()) and m not in known:
                 known.add(m)
                 new.append(m)
         if new:
             f = self.svfs(b'obsstore', b'ab')
             try:
                 offset = f.tell()
                 transaction.add(b'obsstore', offset)
                 # offset == 0: new file - add the version header
                 data = b''.join(encodemarkers(new, offset == 0, self._version))
                 f.write(data)
             finally:
                 # XXX: f.close() == filecache invalidation == obsstore rebuilt.
                 # call 'filecacheentry.refresh()'  here
                 f.close()
             addedmarkers = transaction.changes.get(b'obsmarkers')
             if addedmarkers is not None:
                 addedmarkers.update(new)
             self._addmarkers(new, data)
             # new marker *may* have changed several set. invalidate the cache.
             self.caches.clear()
         # records the number of new markers for the transaction hooks
         previous = int(transaction.hookargs.get(b'new_obsmarkers', b'0'))
         transaction.hookargs[b'new_obsmarkers'] = b'%d' % (previous + len(new))
         return len(new)
 
     def mergemarkers(self, transaction, data):
         """merge a binary stream of markers inside the obsstore
 
         Returns the number of new markers added."""
         version, markers = _readmarkers(data)
         return self.add(transaction, markers)
 
     @propertycache
     def _data(self):
         return self.svfs.tryread(b'obsstore')
 
     @propertycache
     def _version(self):
         if len(self._data) >= 1:
             return _readmarkerversion(self._data)
         else:
             return self._defaultformat
 
     @propertycache
     def _all(self):
         data = self._data
         if not data:
             return []
         self._version, markers = _readmarkers(data)
         markers = list(markers)
         _checkinvalidmarkers(self.repo, markers)
         return markers
 
     @propertycache
     def successors(self):
         successors = {}
         _addsuccessors(successors, self._all)
         return successors
 
     @propertycache
     def predecessors(self):
         predecessors = {}
         _addpredecessors(predecessors, self._all)
         return predecessors
 
     @propertycache
     def children(self):
         children = {}
         _addchildren(children, self._all)
         return children
 
     def _cached(self, attr):
         return attr in self.__dict__
 
     def _addmarkers(self, markers, rawdata):
         markers = list(markers)  # to allow repeated iteration
         self._data = self._data + rawdata
         self._all.extend(markers)
         if self._cached('successors'):
             _addsuccessors(self.successors, markers)
         if self._cached('predecessors'):
             _addpredecessors(self.predecessors, markers)
         if self._cached('children'):
             _addchildren(self.children, markers)
         _checkinvalidmarkers(self.repo, markers)
 
     def relevantmarkers(self, nodes):
         """return a set of all obsolescence markers relevant to a set of nodes.
 
         "relevant" to a set of nodes mean:
 
         - marker that use this changeset as successor
         - prune marker of direct children on this changeset
         - recursive application of the two rules on predecessors of these
           markers
 
         It is a set so you cannot rely on order."""
 
         pendingnodes = set(nodes)
         seenmarkers = set()
         seennodes = set(pendingnodes)
         precursorsmarkers = self.predecessors
         succsmarkers = self.successors
         children = self.children
         while pendingnodes:
             direct = set()
             for current in pendingnodes:
                 direct.update(precursorsmarkers.get(current, ()))
                 pruned = [m for m in children.get(current, ()) if not m[1]]
                 direct.update(pruned)
                 pruned = [m for m in succsmarkers.get(current, ()) if not m[1]]
                 direct.update(pruned)
             direct -= seenmarkers
             pendingnodes = {m[0] for m in direct}
             seenmarkers |= direct
             pendingnodes -= seennodes
             seennodes |= pendingnodes
         return seenmarkers
 
 
 def makestore(ui, repo):
     """Create an obsstore instance from a repo."""
     # read default format for new obsstore.
     # developer config: format.obsstore-version
     defaultformat = ui.configint(b'format', b'obsstore-version')
     # rely on obsstore class default when possible.
     kwargs = {}
     if defaultformat is not None:
         kwargs['defaultformat'] = defaultformat
     readonly = not isenabled(repo, createmarkersopt)
     store = obsstore(repo, repo.svfs, readonly=readonly, **kwargs)
     if store and readonly:
         ui.warn(
             _(b'obsolete feature not enabled but %i markers found!\n')
             % len(list(store))
         )
     return store
 
 
 def commonversion(versions):
     """Return the newest version listed in both versions and our local formats.
 
     Returns None if no common version exists.
     """
     versions.sort(reverse=True)
     # search for highest version known on both side
     for v in versions:
         if v in formats:
             return v
     return None
 
 
 # arbitrary picked to fit into 8K limit from HTTP server
 # you have to take in account:
 # - the version header
 # - the base85 encoding
 _maxpayload = 5300
 
 
 def _pushkeyescape(markers):
     """encode markers into a dict suitable for pushkey exchange
 
     - binary data is base85 encoded
     - split in chunks smaller than 5300 bytes"""
     keys = {}
     parts = []
     currentlen = _maxpayload * 2  # ensure we create a new part
     for marker in markers:
         nextdata = _fm0encodeonemarker(marker)
         if len(nextdata) + currentlen > _maxpayload:
             currentpart = []
             currentlen = 0
             parts.append(currentpart)
         currentpart.append(nextdata)
         currentlen += len(nextdata)
     for idx, part in enumerate(reversed(parts)):
         data = b''.join([_pack(b'>B', _fm0version)] + part)
         keys[b'dump%i' % idx] = util.b85encode(data)
     return keys
 
 
 def listmarkers(repo):
     """List markers over pushkey"""
     if not repo.obsstore:
         return {}
     return _pushkeyescape(sorted(repo.obsstore))
 
 
 def pushmarker(repo, key, old, new):
     """Push markers over pushkey"""
     if not key.startswith(b'dump'):
         repo.ui.warn(_(b'unknown key: %r') % key)
         return False
     if old:
         repo.ui.warn(_(b'unexpected old value for %r') % key)
         return False
     data = util.b85decode(new)
     with repo.lock(), repo.transaction(b'pushkey: obsolete markers') as tr:
         repo.obsstore.mergemarkers(tr, data)
         repo.invalidatevolatilesets()
         return True
 
 
 # mapping of 'set-name' -> <function to compute this set>
 cachefuncs = {}
 
 
 def cachefor(name):
     """Decorator to register a function as computing the cache for a set"""
 
     def decorator(func):
         if name in cachefuncs:
             msg = b"duplicated registration for volatileset '%s' (existing: %r)"
             raise error.ProgrammingError(msg % (name, cachefuncs[name]))
         cachefuncs[name] = func
         return func
 
     return decorator
 
 
 def getrevs(repo, name):
     """Return the set of revision that belong to the <name> set
 
     Such access may compute the set and cache it for future use"""
     repo = repo.unfiltered()
     with util.timedcm('getrevs %s', name):
         if not repo.obsstore:
             return frozenset()
         if name not in repo.obsstore.caches:
             repo.obsstore.caches[name] = cachefuncs[name](repo)
         return repo.obsstore.caches[name]
 
 
 # To be simple we need to invalidate obsolescence cache when:
 #
 # - new changeset is added:
 # - public phase is changed
 # - obsolescence marker are added
 # - strip is used a repo
 def clearobscaches(repo):
     """Remove all obsolescence related cache from a repo
 
     This remove all cache in obsstore is the obsstore already exist on the
     repo.
 
     (We could be smarter here given the exact event that trigger the cache
     clearing)"""
     # only clear cache is there is obsstore data in this repo
     if b'obsstore' in repo._filecache:
         repo.obsstore.caches.clear()
 
 
 def _mutablerevs(repo):
     """the set of mutable revision in the repository"""
     return repo._phasecache.getrevset(repo, phases.mutablephases)
 
 
 @cachefor(b'obsolete')
 def _computeobsoleteset(repo):
     """the set of obsolete revisions"""
     getnode = repo.changelog.node
     notpublic = _mutablerevs(repo)
     isobs = repo.obsstore.successors.__contains__
     return frozenset(r for r in notpublic if isobs(getnode(r)))
 
 
 @cachefor(b'orphan')
 def _computeorphanset(repo):
     """the set of non obsolete revisions with obsolete parents"""
     pfunc = repo.changelog.parentrevs
     mutable = _mutablerevs(repo)
     obsolete = getrevs(repo, b'obsolete')
     others = mutable - obsolete
     unstable = set()
     for r in sorted(others):
         # A rev is unstable if one of its parent is obsolete or unstable
         # this works since we traverse following growing rev order
         for p in pfunc(r):
             if p in obsolete or p in unstable:
                 unstable.add(r)
                 break
     return frozenset(unstable)
 
 
 @cachefor(b'suspended')
 def _computesuspendedset(repo):
     """the set of obsolete parents with non obsolete descendants"""
     suspended = repo.changelog.ancestors(getrevs(repo, b'orphan'))
     return frozenset(r for r in getrevs(repo, b'obsolete') if r in suspended)
 
 
 @cachefor(b'extinct')
 def _computeextinctset(repo):
     """the set of obsolete parents without non obsolete descendants"""
     return getrevs(repo, b'obsolete') - getrevs(repo, b'suspended')
 
 
 @cachefor(b'phasedivergent')
 def _computephasedivergentset(repo):
     """the set of revs trying to obsolete public revisions"""
     bumped = set()
     # util function (avoid attribute lookup in the loop)
     phase = repo._phasecache.phase  # would be faster to grab the full list
     public = phases.public
     cl = repo.changelog
     torev = cl.index.get_rev
     tonode = cl.node
     obsstore = repo.obsstore
     for rev in repo.revs(b'(not public()) and (not obsolete())'):
         # We only evaluate mutable, non-obsolete revision
         node = tonode(rev)
         # (future) A cache of predecessors may worth if split is very common
         for pnode in obsutil.allpredecessors(
             obsstore, [node], ignoreflags=bumpedfix
         ):
             prev = torev(pnode)  # unfiltered! but so is phasecache
             if (prev is not None) and (phase(repo, prev) <= public):
                 # we have a public predecessor
                 bumped.add(rev)
                 break  # Next draft!
     return frozenset(bumped)
 
 
 @cachefor(b'contentdivergent')
 def _computecontentdivergentset(repo):
     """the set of rev that compete to be the final successors of some revision."""
     divergent = set()
     obsstore = repo.obsstore
     newermap = {}
     tonode = repo.changelog.node
     for rev in repo.revs(b'(not public()) - obsolete()'):
         node = tonode(rev)
         mark = obsstore.predecessors.get(node, ())
         toprocess = set(mark)
         seen = set()
         while toprocess:
             prec = toprocess.pop()[0]
             if prec in seen:
                 continue  # emergency cycle hanging prevention
             seen.add(prec)
             if prec not in newermap:
                 obsutil.successorssets(repo, prec, cache=newermap)
             newer = [n for n in newermap[prec] if n]
             if len(newer) > 1:
                 divergent.add(rev)
                 break
             toprocess.update(obsstore.predecessors.get(prec, ()))
     return frozenset(divergent)
 
 
 def makefoldid(relation, user):
 
     folddigest = hashutil.sha1(user)
     for p in relation[0] + relation[1]:
         folddigest.update(b'%d' % p.rev())
         folddigest.update(p.node())
     # Since fold only has to compete against fold for the same successors, it
     # seems fine to use a small ID. Smaller ID save space.
     return hex(folddigest.digest())[:8]
 
 
 def createmarkers(
     repo, relations, flag=0, date=None, metadata=None, operation=None
 ):
     """Add obsolete markers between changesets in a repo
 
     <relations> must be an iterable of ((<old>,...), (<new>, ...)[,{metadata}])
     tuple. `old` and `news` are changectx. metadata is an optional dictionary
     containing metadata for this marker only. It is merged with the global
     metadata specified through the `metadata` argument of this function.
     Any string values in metadata must be UTF-8 bytes.
 
     Trying to obsolete a public changeset will raise an exception.
 
     Current user and date are used except if specified otherwise in the
     metadata attribute.
 
     This function operates within a transaction of its own, but does
     not take any lock on the repo.
     """
     # prepare metadata
     if metadata is None:
         metadata = {}
     if b'user' not in metadata:
         luser = (
             repo.ui.config(b'devel', b'user.obsmarker') or repo.ui.username()
         )
         metadata[b'user'] = encoding.fromlocal(luser)
 
     # Operation metadata handling
     useoperation = repo.ui.configbool(
         b'experimental', b'evolution.track-operation'
     )
     if useoperation and operation:
         metadata[b'operation'] = operation
 
     # Effect flag metadata handling
     saveeffectflag = repo.ui.configbool(
         b'experimental', b'evolution.effect-flags'
     )
 
     with repo.transaction(b'add-obsolescence-marker') as tr:
         markerargs = []
         for rel in relations:
             predecessors = rel[0]
             if not isinstance(predecessors, tuple):
                 # preserve compat with old API until all caller are migrated
                 predecessors = (predecessors,)
             if len(predecessors) > 1 and len(rel[1]) != 1:
                 msg = b'Fold markers can only have 1 successors, not %d'
                 raise error.ProgrammingError(msg % len(rel[1]))
             foldid = None
             foldsize = len(predecessors)
             if 1 < foldsize:
                 foldid = makefoldid(rel, metadata[b'user'])
             for foldidx, prec in enumerate(predecessors, 1):
                 sucs = rel[1]
                 localmetadata = metadata.copy()
                 if len(rel) > 2:
                     localmetadata.update(rel[2])
                 if foldid is not None:
                     localmetadata[b'fold-id'] = foldid
                     localmetadata[b'fold-idx'] = b'%d' % foldidx
                     localmetadata[b'fold-size'] = b'%d' % foldsize
 
                 if not prec.mutable():
                     raise error.Abort(
                         _(b"cannot obsolete public changeset: %s") % prec,
                         hint=b"see 'hg help phases' for details",
                     )
                 nprec = prec.node()
                 nsucs = tuple(s.node() for s in sucs)
                 npare = None
                 if not nsucs:
                     npare = tuple(p.node() for p in prec.parents())
                 if nprec in nsucs:
                     raise error.Abort(
                         _(b"changeset %s cannot obsolete itself") % prec
                     )
 
                 # Effect flag can be different by relation
                 if saveeffectflag:
                     # The effect flag is saved in a versioned field name for
                     # future evolution
                     effectflag = obsutil.geteffectflag(prec, sucs)
                     localmetadata[obsutil.EFFECTFLAGFIELD] = b"%d" % effectflag
 
                 # Creating the marker causes the hidden cache to become
                 # invalid, which causes recomputation when we ask for
                 # prec.parents() above.  Resulting in n^2 behavior.  So let's
                 # prepare all of the args first, then create the markers.
                 markerargs.append((nprec, nsucs, npare, localmetadata))
 
         for args in markerargs:
             nprec, nsucs, npare, localmetadata = args
             repo.obsstore.create(
                 tr,
                 nprec,
                 nsucs,
                 flag,
                 parents=npare,
                 date=date,
                 metadata=localmetadata,
                 ui=repo.ui,
             )
             repo.filteredrevcache.clear()