largefiles: correctly handle dirstate status when rebasing

When rebasing, we need to trust that the standins are always correct. The
rebase operation updates the standins according to the changeset it is
rebasing. We need to make the largefiles in the working copy match.  If we
don't make them match, then they get accidentally reverted, either during
the rebase or during the next commit after the rebase.

This worked previously only becuase we were relying on the behavior that
largefiles with a changed standin, but unchanged contents, never showed up in
the list of modified largefiles.  Unfortunately, pre-commit hooks can get
an incorrect status this way, and it also results in extra execution of code.

The solution is to simply trust the standins when we are about to commit a
rebased changeset, and politely ask updatelfiles() to pull the new contents
down.  In this case, updatelfiles() will also mark any files it has pulled
down as dirty in the lfdirstate so that pre-commit hooks will get correct
status output.

hgext/largefiles/lfcommands.py

@@ -1,479 +1,485 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.
 
 '''High-level command function for lfconvert, plus the cmdtable.'''
 
 import os
 import shutil
 
 from mercurial import util, match as match_, hg, node, context, error
 from mercurial.i18n import _
 
 import lfutil
 import basestore
 
 # -- Commands ----------------------------------------------------------
 
 def lfconvert(ui, src, dest, *pats, **opts):
     '''convert a normal repository to a largefiles repository
 
     Convert repository SOURCE to a new repository DEST, identical to
     SOURCE except that certain files will be converted as largefiles:
     specifically, any file that matches any PATTERN *or* whose size is
     above the minimum size threshold is converted as a largefile. The
     size used to determine whether or not to track a file as a
     largefile is the size of the first version of the file. The
     minimum size can be specified either with --size or in
     configuration as ``largefiles.size``.
 
     After running this command you will need to make sure that
     largefiles is enabled anywhere you intend to push the new
     repository.
 
     Use --to-normal to convert largefiles back to normal files; after
     this, the DEST repository can be used without largefiles at all.'''
 
     if opts['to_normal']:
         tolfile = False
     else:
         tolfile = True
         size = lfutil.getminsize(ui, True, opts.get('size'), default=None)
 
     if not hg.islocal(src):
         raise util.Abort(_('%s is not a local Mercurial repo') % src)
     if not hg.islocal(dest):
         raise util.Abort(_('%s is not a local Mercurial repo') % dest)
 
     rsrc = hg.repository(ui, src)
     ui.status(_('initializing destination %s\n') % dest)
     rdst = hg.repository(ui, dest, create=True)
 
     success = False
     try:
         # Lock destination to prevent modification while it is converted to.
         # Don't need to lock src because we are just reading from its history
         # which can't change.
         dst_lock = rdst.lock()
 
         # Get a list of all changesets in the source.  The easy way to do this
         # is to simply walk the changelog, using changelog.nodesbewteen().
         # Take a look at mercurial/revlog.py:639 for more details.
         # Use a generator instead of a list to decrease memory usage
         ctxs = (rsrc[ctx] for ctx in rsrc.changelog.nodesbetween(None,
             rsrc.heads())[0])
         revmap = {node.nullid: node.nullid}
         if tolfile:
             lfiles = set()
             normalfiles = set()
             if not pats:
                 pats = ui.configlist(lfutil.longname, 'patterns', default=[])
             if pats:
                 matcher = match_.match(rsrc.root, '', list(pats))
             else:
                 matcher = None
 
             lfiletohash = {}
             for ctx in ctxs:
                 ui.progress(_('converting revisions'), ctx.rev(),
                     unit=_('revision'), total=rsrc['tip'].rev())
                 _lfconvert_addchangeset(rsrc, rdst, ctx, revmap,
                     lfiles, normalfiles, matcher, size, lfiletohash)
             ui.progress(_('converting revisions'), None)
 
             if os.path.exists(rdst.wjoin(lfutil.shortname)):
                 shutil.rmtree(rdst.wjoin(lfutil.shortname))
 
             for f in lfiletohash.keys():
                 if os.path.isfile(rdst.wjoin(f)):
                     os.unlink(rdst.wjoin(f))
                 try:
                     os.removedirs(os.path.dirname(rdst.wjoin(f)))
                 except OSError:
                     pass
 
             # If there were any files converted to largefiles, add largefiles
             # to the destination repository's requirements.
             if lfiles:
                 rdst.requirements.add('largefiles')
                 rdst._writerequirements()
         else:
             for ctx in ctxs:
                 ui.progress(_('converting revisions'), ctx.rev(),
                     unit=_('revision'), total=rsrc['tip'].rev())
                 _addchangeset(ui, rsrc, rdst, ctx, revmap)
 
             ui.progress(_('converting revisions'), None)
         success = True
     finally:
         if not success:
             # we failed, remove the new directory
             shutil.rmtree(rdst.root)
         dst_lock.release()
 
 def _addchangeset(ui, rsrc, rdst, ctx, revmap):
  # Convert src parents to dst parents
     parents = []
     for p in ctx.parents():
         parents.append(revmap[p.node()])
     while len(parents) < 2:
         parents.append(node.nullid)
 
     # Generate list of changed files
     files = set(ctx.files())
     if node.nullid not in parents:
         mc = ctx.manifest()
         mp1 = ctx.parents()[0].manifest()
         mp2 = ctx.parents()[1].manifest()
         files |= (set(mp1) | set(mp2)) - set(mc)
         for f in mc:
             if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                 files.add(f)
 
     def getfilectx(repo, memctx, f):
         if lfutil.standin(f) in files:
             # if the file isn't in the manifest then it was removed
             # or renamed, raise IOError to indicate this
             try:
                 fctx = ctx.filectx(lfutil.standin(f))
             except error.LookupError:
                 raise IOError()
             renamed = fctx.renamed()
             if renamed:
                 renamed = lfutil.splitstandin(renamed[0])
 
             hash = fctx.data().strip()
             path = lfutil.findfile(rsrc, hash)
             ### TODO: What if the file is not cached?
             data = ''
             fd = None
             try:
                 fd = open(path, 'rb')
                 data = fd.read()
             finally:
                 if fd:
                     fd.close()
             return context.memfilectx(f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
         else:
             try:
                 fctx = ctx.filectx(f)
             except error.LookupError:
                 raise IOError()
             renamed = fctx.renamed()
             if renamed:
                 renamed = renamed[0]
             data = fctx.data()
             if f == '.hgtags':
                 newdata = []
                 for line in data.splitlines():
                     id, name = line.split(' ', 1)
                     newdata.append('%s %s\n' % (node.hex(revmap[node.bin(id)]),
                         name))
                 data = ''.join(newdata)
             return context.memfilectx(f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
 
     dstfiles = []
     for file in files:
         if lfutil.isstandin(file):
             dstfiles.append(lfutil.splitstandin(file))
         else:
             dstfiles.append(file)
     # Commit
     mctx = context.memctx(rdst, parents, ctx.description(), dstfiles,
                           getfilectx, ctx.user(), ctx.date(), ctx.extra())
     ret = rdst.commitctx(mctx)
     rdst.dirstate.setparents(ret)
     revmap[ctx.node()] = rdst.changelog.tip()
 
 def _lfconvert_addchangeset(rsrc, rdst, ctx, revmap, lfiles, normalfiles,
         matcher, size, lfiletohash):
     # Convert src parents to dst parents
     parents = []
     for p in ctx.parents():
         parents.append(revmap[p.node()])
     while len(parents) < 2:
         parents.append(node.nullid)
 
     # Generate list of changed files
     files = set(ctx.files())
     if node.nullid not in parents:
         mc = ctx.manifest()
         mp1 = ctx.parents()[0].manifest()
         mp2 = ctx.parents()[1].manifest()
         files |= (set(mp1) | set(mp2)) - set(mc)
         for f in mc:
             if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                 files.add(f)
 
     dstfiles = []
     for f in files:
         if f not in lfiles and f not in normalfiles:
             islfile = _islfile(f, ctx, matcher, size)
             # If this file was renamed or copied then copy
             # the lfileness of its predecessor
             if f in ctx.manifest():
                 fctx = ctx.filectx(f)
                 renamed = fctx.renamed()
                 renamedlfile = renamed and renamed[0] in lfiles
                 islfile |= renamedlfile
                 if 'l' in fctx.flags():
                     if renamedlfile:
                         raise util.Abort(
                             _('renamed/copied largefile %s becomes symlink')
                             % f)
                     islfile = False
             if islfile:
                 lfiles.add(f)
             else:
                 normalfiles.add(f)
 
         if f in lfiles:
             dstfiles.append(lfutil.standin(f))
             # largefile in manifest if it has not been removed/renamed
             if f in ctx.manifest():
                 if 'l' in ctx.filectx(f).flags():
                     if renamed and renamed[0] in lfiles:
                         raise util.Abort(_('largefile %s becomes symlink') % f)
 
                 # largefile was modified, update standins
                 fullpath = rdst.wjoin(f)
                 util.makedirs(os.path.dirname(fullpath))
                 m = util.sha1('')
                 m.update(ctx[f].data())
                 hash = m.hexdigest()
                 if f not in lfiletohash or lfiletohash[f] != hash:
                     try:
                         fd = open(fullpath, 'wb')
                         fd.write(ctx[f].data())
                     finally:
                         if fd:
                             fd.close()
                     executable = 'x' in ctx[f].flags()
                     os.chmod(fullpath, lfutil.getmode(executable))
                     lfutil.writestandin(rdst, lfutil.standin(f), hash,
                         executable)
                     lfiletohash[f] = hash
         else:
             # normal file
             dstfiles.append(f)
 
     def getfilectx(repo, memctx, f):
         if lfutil.isstandin(f):
             # if the file isn't in the manifest then it was removed
             # or renamed, raise IOError to indicate this
             srcfname = lfutil.splitstandin(f)
             try:
                 fctx = ctx.filectx(srcfname)
             except error.LookupError:
                 raise IOError()
             renamed = fctx.renamed()
             if renamed:
                 # standin is always a largefile because largefile-ness
                 # doesn't change after rename or copy
                 renamed = lfutil.standin(renamed[0])
 
             return context.memfilectx(f, lfiletohash[srcfname] + '\n', 'l' in
                 fctx.flags(), 'x' in fctx.flags(), renamed)
         else:
             try:
                 fctx = ctx.filectx(f)
             except error.LookupError:
                 raise IOError()
             renamed = fctx.renamed()
             if renamed:
                 renamed = renamed[0]
 
             data = fctx.data()
             if f == '.hgtags':
                 newdata = []
                 for line in data.splitlines():
                     id, name = line.split(' ', 1)
                     newdata.append('%s %s\n' % (node.hex(revmap[node.bin(id)]),
                         name))
                 data = ''.join(newdata)
             return context.memfilectx(f, data, 'l' in fctx.flags(),
                                       'x' in fctx.flags(), renamed)
 
     # Commit
     mctx = context.memctx(rdst, parents, ctx.description(), dstfiles,
                           getfilectx, ctx.user(), ctx.date(), ctx.extra())
     ret = rdst.commitctx(mctx)
     rdst.dirstate.setparents(ret)
     revmap[ctx.node()] = rdst.changelog.tip()
 
 def _islfile(file, ctx, matcher, size):
     '''Return true if file should be considered a largefile, i.e.
     matcher matches it or it is larger than size.'''
     # never store special .hg* files as largefiles
     if file == '.hgtags' or file == '.hgignore' or file == '.hgsigs':
         return False
     if matcher and matcher(file):
         return True
     try:
         return ctx.filectx(file).size() >= size * 1024 * 1024
     except error.LookupError:
         return False
 
 def uploadlfiles(ui, rsrc, rdst, files):
     '''upload largefiles to the central store'''
 
     if not files:
         return
 
     store = basestore._openstore(rsrc, rdst, put=True)
 
     at = 0
     files = filter(lambda h: not store.exists(h), files)
     for hash in files:
         ui.progress(_('uploading largefiles'), at, unit='largefile',
                     total=len(files))
         source = lfutil.findfile(rsrc, hash)
         if not source:
             raise util.Abort(_('largefile %s missing from store'
                                ' (needs to be uploaded)') % hash)
         # XXX check for errors here
         store.put(source, hash)
         at += 1
     ui.progress(_('uploading largefiles'), None)
 
 def verifylfiles(ui, repo, all=False, contents=False):
     '''Verify that every big file revision in the current changeset
     exists in the central store.  With --contents, also verify that
     the contents of each big file revision are correct (SHA-1 hash
     matches the revision ID).  With --all, check every changeset in
     this repository.'''
     if all:
         # Pass a list to the function rather than an iterator because we know a
         # list will work.
         revs = range(len(repo))
     else:
         revs = ['.']
 
     store = basestore._openstore(repo)
     return store.verify(revs, contents=contents)
 
 def cachelfiles(ui, repo, node):
     '''cachelfiles ensures that all largefiles needed by the specified revision
     are present in the repository's largefile cache.
 
     returns a tuple (cached, missing).  cached is the list of files downloaded
     by this operation; missing is the list of files that were needed but could
     not be found.'''
     lfiles = lfutil.listlfiles(repo, node)
     toget = []
 
     for lfile in lfiles:
         expectedhash = repo[node][lfutil.standin(lfile)].data().strip()
         # if it exists and its hash matches, it might have been locally
         # modified before updating and the user chose 'local'.  in this case,
         # it will not be in any store, so don't look for it.
         if ((not os.path.exists(repo.wjoin(lfile)) or
              expectedhash != lfutil.hashfile(repo.wjoin(lfile))) and
             not lfutil.findfile(repo, expectedhash)):
             toget.append((lfile, expectedhash))
 
     if toget:
         store = basestore._openstore(repo)
         ret = store.get(toget)
         return ret
 
     return ([], [])
 
 def updatelfiles(ui, repo, filelist=None, printmessage=True):
     wlock = repo.wlock()
     try:
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         lfiles = set(lfutil.listlfiles(repo)) | set(lfdirstate)
 
         if filelist is not None:
             lfiles = [f for f in lfiles if f in filelist]
 
         printed = False
         if printmessage and lfiles:
             ui.status(_('getting changed largefiles\n'))
             printed = True
             cachelfiles(ui, repo, '.')
 
         updated, removed = 0, 0
         for i in map(lambda f: _updatelfile(repo, lfdirstate, f), lfiles):
             # increment the appropriate counter according to _updatelfile's
             # return value
             updated += i > 0 and i or 0
             removed -= i < 0 and i or 0
             if printmessage and (removed or updated) and not printed:
                 ui.status(_('getting changed largefiles\n'))
                 printed = True
 
         lfdirstate.write()
         if printed and printmessage:
             ui.status(_('%d largefiles updated, %d removed\n') % (updated,
                 removed))
     finally:
         wlock.release()
 
 def _updatelfile(repo, lfdirstate, lfile):
     '''updates a single largefile and copies the state of its standin from
     the repository's dirstate to its state in the lfdirstate.
 
     returns 1 if the file was modified, -1 if the file was removed, 0 if the
     file was unchanged, and None if the needed largefile was missing from the
     cache.'''
     ret = 0
     abslfile = repo.wjoin(lfile)
     absstandin = repo.wjoin(lfutil.standin(lfile))
     if os.path.exists(absstandin):
         if os.path.exists(absstandin+'.orig'):
             shutil.copyfile(abslfile, abslfile+'.orig')
         expecthash = lfutil.readstandin(repo, lfile)
         if (expecthash != '' and
             (not os.path.exists(abslfile) or
              expecthash != lfutil.hashfile(abslfile))):
             if not lfutil.copyfromcache(repo, expecthash, lfile):
                 # use normallookup() to allocate entry in largefiles dirstate,
                 # because lack of it misleads lfiles_repo.status() into
                 # recognition that such cache missing files are REMOVED.
                 lfdirstate.normallookup(lfile)
                 return None # don't try to set the mode
             ret = 1
         mode = os.stat(absstandin).st_mode
         if mode != os.stat(abslfile).st_mode:
             os.chmod(abslfile, mode)
             ret = 1
     else:
         # Remove lfiles for which the standin is deleted, unless the
         # lfile is added to the repository again. This happens when a
         # largefile is converted back to a normal file: the standin
         # disappears, but a new (normal) file appears as the lfile.
         if os.path.exists(abslfile) and lfile not in repo[None]:
             os.unlink(abslfile)
             ret = -1
     state = repo.dirstate[lfutil.standin(lfile)]
     if state == 'n':
-        lfdirstate.normal(lfile)
+        # When rebasing, we need to synchronize the standin and the largefile,
+        # because otherwise the largefile will get reverted.  But for commit's
+        # sake, we have to mark the file as unclean.
+        if getattr(repo, "_isrebasing", False):
+           lfdirstate.normallookup(lfile)
+        else:
+            lfdirstate.normal(lfile)
     elif state == 'r':
         lfdirstate.remove(lfile)
     elif state == 'a':
         lfdirstate.add(lfile)
     elif state == '?':
         lfdirstate.drop(lfile)
     return ret
 
 # -- hg commands declarations ------------------------------------------------
 
 cmdtable = {
     'lfconvert': (lfconvert,
                   [('s', 'size', '',
                     _('minimum size (MB) for files to be converted '
                       'as largefiles'),
                     'SIZE'),
                   ('', 'to-normal', False,
                    _('convert from a largefiles repo to a normal repo')),
                   ],
                   _('hg lfconvert SOURCE DEST [FILE ...]')),
     }

hgext/largefiles/lfutil.py

@@ -1,460 +1,462 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 errno
 import platform
 import shutil
 import stat
 import tempfile
 
 from mercurial import dirstate, httpconnection, match as match_, util, scmutil
 from mercurial.i18n import _
 
 shortname = '.hglf'
 longname = 'largefiles'
 
 
 # -- Portability wrappers ----------------------------------------------
 
 def dirstate_walk(dirstate, matcher, unknown=False, ignored=False):
     return dirstate.walk(matcher, [], unknown, ignored)
 
 def repo_add(repo, list):
     add = repo[None].add
     return add(list)
 
 def repo_remove(repo, list, unlink=False):
     def remove(list, unlink):
         wlock = repo.wlock()
         try:
             if unlink:
                 for f in list:
                     try:
                         util.unlinkpath(repo.wjoin(f))
                     except OSError, inst:
                         if inst.errno != errno.ENOENT:
                             raise
             repo[None].forget(list)
         finally:
             wlock.release()
     return remove(list, unlink=unlink)
 
 def repo_forget(repo, list):
     forget = repo[None].forget
     return forget(list)
 
 def findoutgoing(repo, remote, force):
     from mercurial import discovery
     common, _anyinc, _heads = discovery.findcommonincoming(repo,
         remote, force=force)
     return repo.changelog.findmissing(common)
 
 # -- 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):
     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)
     elif inusercache(repo.ui, hash):
         repo.ui.note(_('Found %s in system cache\n') % hash)
         path = storepath(repo, hash)
         util.makedirs(os.path.dirname(path))
         link(usercachepath(repo.ui, hash), path)
     else:
         return None
     return storepath(repo, hash)
 
 class largefiles_dirstate(dirstate.dirstate):
     def __getitem__(self, key):
         return super(largefiles_dirstate, self).__getitem__(unixpath(key))
     def normal(self, f):
         return super(largefiles_dirstate, self).normal(unixpath(f))
     def remove(self, f):
         return super(largefiles_dirstate, self).remove(unixpath(f))
     def add(self, f):
         return super(largefiles_dirstate, self).add(unixpath(f))
     def drop(self, f):
         return super(largefiles_dirstate, self).drop(unixpath(f))
     def forget(self, f):
         return super(largefiles_dirstate, self).forget(unixpath(f))
+    def normallookup(self, f):
+        return super(largefiles_dirstate, self).normallookup(unixpath(f))
 
 def openlfdirstate(ui, repo):
     '''
     Return a dirstate object that tracks largefiles: i.e. its root is
     the repo root, but it is saved in .hg/largefiles/dirstate.
     '''
     admin = repo.join(longname)
     opener = scmutil.opener(admin)
     lfdirstate = largefiles_dirstate(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. It also gives us an easy
     # way to forcibly rebuild largefiles state:
     #   rm .hg/largefiles/dirstate && hg status
     # Or even, if things are really messed up:
     #   rm -rf .hg/largefiles && hg status
     if not os.path.exists(os.path.join(admin, 'dirstate')):
         util.makedirs(admin)
         matcher = getstandinmatcher(repo)
         for standin in dirstate_walk(repo.dirstate, matcher):
             lfile = splitstandin(standin)
             hash = readstandin(repo, lfile)
             lfdirstate.normallookup(lfile)
             try:
                 if hash == hashfile(repo.wjoin(lfile)):
                     lfdirstate.normal(lfile)
             except OSError, err:
                 if err.errno != errno.ENOENT:
                     raise
 
         lfdirstate.write()
 
     return lfdirstate
 
 def lfdirstate_status(lfdirstate, repo, rev):
     wlock = repo.wlock()
     try:
         match = match_.always(repo.root, repo.getcwd())
         s = lfdirstate.status(match, [], False, False, False)
         unsure, modified, added, removed, missing, unknown, ignored, clean = s
         for lfile in unsure:
             if repo[rev][standin(lfile)].data().strip() != \
                     hashfile(repo.wjoin(lfile)):
                 modified.append(lfile)
             else:
                 clean.append(lfile)
                 lfdirstate.normal(lfile)
         lfdirstate.write()
     finally:
         wlock.release()
     return (modified, added, removed, missing, unknown, ignored, clean)
 
 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)
     if instore(repo, hash):
         return
     copytostoreabsolute(repo, repo.wjoin(file), hash)
 
 def copytostoreabsolute(repo, file, hash):
     util.makedirs(os.path.dirname(storepath(repo, hash)))
     if inusercache(repo.ui, hash):
         link(usercachepath(repo.ui, hash), storepath(repo, hash))
     else:
         dst = util.atomictempfile(storepath(repo, hash))
         for chunk in util.filechunkiter(open(file, 'rb')):
             dst.write(chunk)
         dst.close()
         util.copymode(file, storepath(repo, hash))
         linktousercache(repo, hash)
 
 def linktousercache(repo, hash):
     path = usercachepath(repo.ui, hash)
     if path:
         util.makedirs(os.path.dirname(path))
         link(storepath(repo, hash), path)
 
 def getstandinmatcher(repo, pats=[], opts={}):
     '''Return a match object that applies pats to the standin directory'''
     standindir = repo.pathto(shortname)
     if pats:
         # patterns supplied: search standin directory relative to current dir
         cwd = repo.getcwd()
         if os.path.isabs(cwd):
             # cwd is an absolute path for hg -R <reponame>
             # work relative to the repository root in this case
             cwd = ''
         pats = [os.path.join(standindir, cwd, pat) for pat in pats]
     elif os.path.isdir(standindir):
         # no patterns: relative to repo root
         pats = [standindir]
     else:
         # no patterns and no standin dir: return matcher that matches nothing
         match = match_.match(repo.root, None, [], exact=True)
         match.matchfn = lambda f: False
         return match
     return getmatcher(repo, pats, opts, showbad=False)
 
 def getmatcher(repo, pats=[], opts={}, showbad=True):
     '''Wrapper around scmutil.match() that adds showbad: if false,
     neuter the match object's bad() method so it does not print any
     warnings about missing files or directories.'''
     match = scmutil.match(repo[None], pats, opts)
 
     if not showbad:
         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 composed_matchfn(f):
         return isstandin(f) and rmatcher.matchfn(splitstandin(f))
     smatcher.matchfn = composed_matchfn
 
     return smatcher
 
 def standin(filename):
     '''Return the repo-relative path to the standin for the specified big
     file.'''
     # Notes:
     # 1) Most callers want an absolute path, but _create_standin() needs
     #    it repo-relative so lfadd() can pass it to repo_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 shortname + '/' + filename.replace(os.sep, '/')
 
 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(shortname + '/')
 
 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 = filename.replace(os.sep, '/').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>'''
     writehash(hash, repo.wjoin(standin), executable)
 
 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.  Close outfile
     when done and return the binary hash.'''
     hasher = util.sha1('')
     for data in instream:
         hasher.update(data)
         outfile.write(data)
 
     # Blecch: closing a file that somebody else opened is rude and
     # wrong. But it's so darn convenient and practical! After all,
     # outfile was opened just to copy and hash.
     outfile.close()
 
     return hasher.digest()
 
 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 blockstream(fd):
         hasher.update(data)
     fd.close()
     return hasher.hexdigest()
 
 class limitreader(object):
     def __init__(self, f, limit):
         self.f = f
         self.limit = limit
 
     def read(self, length):
         if self.limit == 0:
             return ''
         length = length > self.limit and self.limit or length
         self.limit -= length
         return self.f.read(length)
 
     def close(self):
         pass
 
 def blockstream(infile, blocksize=128 * 1024):
     """Generator that yields blocks of data from infile and closes infile."""
     while True:
         data = infile.read(blocksize)
         if not data:
             break
         yield data
     # same blecch as copyandhash() above
     infile.close()
 
 def readhash(filename):
     rfile = open(filename, 'rb')
     hash = rfile.read(40)
     rfile.close()
     if len(hash) < 40:
         raise util.Abort(_('bad hash in \'%s\' (only %d bytes long)')
                          % (filename, len(hash)))
     return hash
 
 def writehash(hash, filename, executable):
     util.makedirs(os.path.dirname(filename))
     util.writefile(filename, hash + '\n')
     os.chmod(filename, getmode(executable))
 
 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 getmode(executable):
     if executable:
         return 0755
     else:
         return 0644
 
 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 os.path.normpath(path).replace(os.sep, '/')
 
 def islfilesrepo(repo):
     return ('largefiles' in repo.requirements and
             util.any(shortname + '/' in f[0] for f in repo.store.datafiles()))
 
 def mkstemp(repo, prefix):
     '''Returns a file descriptor and a filename corresponding to a temporary
     file in the repo's largefiles store.'''
     path = repo.join(longname)
     util.makedirs(path)
     return tempfile.mkstemp(prefix=prefix, dir=path)
 
 class storeprotonotcapable(Exception):
     def __init__(self, storetypes):
         self.storetypes = storetypes

hgext/largefiles/reposetup.py

@@ -1,446 +1,450 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.
 
 '''setup for largefiles repositories: reposetup'''
 import copy
 import types
 import os
 
 from mercurial import context, error, manifest, match as match_, util
 from mercurial import node as node_
 from mercurial.i18n import _
 
 import lfcommands
 import proto
 import lfutil
 
 def reposetup(ui, repo):
     # wire repositories should be given new wireproto functions but not the
     # other largefiles modifications
     if not repo.local():
         return proto.wirereposetup(ui, repo)
 
     for name in ('status', 'commitctx', 'commit', 'push'):
         method = getattr(repo, name)
         if (isinstance(method, types.FunctionType) and
             method.func_name == 'wrap'):
             ui.warn(_('largefiles: repo method %r appears to have already been'
                     ' wrapped by another extension: '
                     'largefiles may behave incorrectly\n')
                     % name)
 
     class lfiles_repo(repo.__class__):
         lfstatus = False
         def status_nolfiles(self, *args, **kwargs):
             return super(lfiles_repo, self).status(*args, **kwargs)
 
         # When lfstatus is set, return a context that gives the names
         # of largefiles instead of their corresponding standins and
         # identifies the largefiles as always binary, regardless of
         # their actual contents.
         def __getitem__(self, changeid):
             ctx = super(lfiles_repo, self).__getitem__(changeid)
             if self.lfstatus:
                 class lfiles_manifestdict(manifest.manifestdict):
                     def __contains__(self, filename):
                         if super(lfiles_manifestdict,
                                 self).__contains__(filename):
                             return True
                         return super(lfiles_manifestdict,
                             self).__contains__(lfutil.standin(filename))
                 class lfiles_ctx(ctx.__class__):
                     def files(self):
                         filenames = super(lfiles_ctx, self).files()
                         return [lfutil.splitstandin(f) or f for f in filenames]
                     def manifest(self):
                         man1 = super(lfiles_ctx, self).manifest()
                         man1.__class__ = lfiles_manifestdict
                         return man1
                     def filectx(self, path, fileid=None, filelog=None):
                         try:
                             result = super(lfiles_ctx, self).filectx(path,
                                 fileid, filelog)
                         except error.LookupError:
                             # Adding a null character will cause Mercurial to
                             # identify this as a binary file.
                             result = super(lfiles_ctx, self).filectx(
                                 lfutil.standin(path), fileid, filelog)
                             olddata = result.data
                             result.data = lambda: olddata() + '\0'
                         return result
                 ctx.__class__ = lfiles_ctx
             return ctx
 
         # Figure out the status of big files and insert them into the
         # appropriate list in the result. Also removes standin files
         # from the listing. Revert to the original status if
         # self.lfstatus is False.
         def status(self, node1='.', node2=None, match=None, ignored=False,
                 clean=False, unknown=False, listsubrepos=False):
             listignored, listclean, listunknown = ignored, clean, unknown
             if not self.lfstatus:
                 return super(lfiles_repo, self).status(node1, node2, match,
                     listignored, listclean, listunknown, listsubrepos)
             else:
                 # some calls in this function rely on the old version of status
                 self.lfstatus = False
                 if isinstance(node1, context.changectx):
                     ctx1 = node1
                 else:
                     ctx1 = repo[node1]
                 if isinstance(node2, context.changectx):
                     ctx2 = node2
                 else:
                     ctx2 = repo[node2]
                 working = ctx2.rev() is None
                 parentworking = working and ctx1 == self['.']
 
                 def inctx(file, ctx):
                     try:
                         if ctx.rev() is None:
                             return file in ctx.manifest()
                         ctx[file]
                         return True
                     except KeyError:
                         return False
 
                 if match is None:
                     match = match_.always(self.root, self.getcwd())
 
                 # First check if there were files specified on the
                 # command line.  If there were, and none of them were
                 # largefiles, we should just bail here and let super
                 # handle it -- thus gaining a big performance boost.
                 lfdirstate = lfutil.openlfdirstate(ui, self)
                 if match.files() and not match.anypats():
                     matchedfiles = [f for f in match.files() if f in lfdirstate]
                     if not matchedfiles:
                         return super(lfiles_repo, self).status(node1, node2,
                                 match, listignored, listclean,
                                 listunknown, listsubrepos)
 
                 # Create a copy of match that matches standins instead
                 # of largefiles.
                 def tostandin(file):
                     if inctx(lfutil.standin(file), ctx2):
                         return lfutil.standin(file)
                     return file
 
                 # Create a function that we can use to override what is
                 # normally the ignore matcher.  We've already checked
                 # for ignored files on the first dirstate walk, and
                 # unecessarily re-checking here causes a huge performance
                 # hit because lfdirstate only knows about largefiles
                 def _ignoreoverride(self):
                     return False
 
                 m = copy.copy(match)
                 m._files = [tostandin(f) for f in m._files]
 
                 # Get ignored files here even if we weren't asked for them; we
                 # must use the result here for filtering later
                 result = super(lfiles_repo, self).status(node1, node2, m,
                     True, clean, unknown, listsubrepos)
                 if working:
                     # hold the wlock while we read largefiles and
                     # update the lfdirstate
                     wlock = repo.wlock()
                     try:
                         # Any non-largefiles that were explicitly listed must be
                         # taken out or lfdirstate.status will report an error.
                         # The status of these files was already computed using
                         # super's status.
                         # Override lfdirstate's ignore matcher to not do
                         # anything
                         orig_ignore = lfdirstate._ignore
                         lfdirstate._ignore = _ignoreoverride
 
                         match._files = [f for f in match._files if f in
                             lfdirstate]
                         # Don't waste time getting the ignored and unknown
                         # files again; we already have them
                         s = lfdirstate.status(match, [], False,
                                 listclean, False)
                         (unsure, modified, added, removed, missing, unknown,
                                 ignored, clean) = s
                         # Replace the list of ignored and unknown files with
                         # the previously caclulated lists, and strip out the
                         # largefiles
                         lfiles = set(lfdirstate._map)
                         ignored = set(result[5]).difference(lfiles)
                         unknown = set(result[4]).difference(lfiles)
                         if parentworking:
                             for lfile in unsure:
                                 standin = lfutil.standin(lfile)
                                 if standin not in ctx1:
                                     # from second parent
                                     modified.append(lfile)
                                 elif ctx1[standin].data().strip() \
                                         != lfutil.hashfile(self.wjoin(lfile)):
                                     modified.append(lfile)
                                 else:
                                     clean.append(lfile)
                                     lfdirstate.normal(lfile)
                             lfdirstate.write()
                         else:
                             tocheck = unsure + modified + added + clean
                             modified, added, clean = [], [], []
 
                             for lfile in tocheck:
                                 standin = lfutil.standin(lfile)
                                 if inctx(standin, ctx1):
                                     if ctx1[standin].data().strip() != \
                                             lfutil.hashfile(self.wjoin(lfile)):
                                         modified.append(lfile)
                                     else:
                                         clean.append(lfile)
                                 else:
                                     added.append(lfile)
                         # Replace the original ignore function
                         lfdirstate._ignore = orig_ignore
                     finally:
                         wlock.release()
 
                     for standin in ctx1.manifest():
                         if not lfutil.isstandin(standin):
                             continue
                         lfile = lfutil.splitstandin(standin)
                         if not match(lfile):
                             continue
                         if lfile not in lfdirstate:
                             removed.append(lfile)
 
                     # Filter result lists
                     result = list(result)
 
                     # Largefiles are not really removed when they're
                     # still in the normal dirstate. Likewise, normal
                     # files are not really removed if it's still in
                     # lfdirstate. This happens in merges where files
                     # change type.
                     removed = [f for f in removed if f not in repo.dirstate]
                     result[2] = [f for f in result[2] if f not in lfdirstate]
 
                     # Unknown files
                     unknown = set(unknown).difference(ignored)
                     result[4] = [f for f in unknown
                                  if (repo.dirstate[f] == '?' and
                                      not lfutil.isstandin(f))]
                     # Ignored files were calculated earlier by the dirstate,
                     # and we already stripped out the largefiles from the list
                     result[5] = ignored
                     # combine normal files and largefiles
                     normals = [[fn for fn in filelist
                                 if not lfutil.isstandin(fn)]
                                for filelist in result]
                     lfiles = (modified, added, removed, missing, [], [], clean)
                     result = [sorted(list1 + list2)
                               for (list1, list2) in zip(normals, lfiles)]
                 else:
                     def toname(f):
                         if lfutil.isstandin(f):
                             return lfutil.splitstandin(f)
                         return f
                     result = [[toname(f) for f in items] for items in result]
 
                 if not listunknown:
                     result[4] = []
                 if not listignored:
                     result[5] = []
                 if not listclean:
                     result[6] = []
                 self.lfstatus = True
                 return result
 
         # As part of committing, copy all of the largefiles into the
         # cache.
         def commitctx(self, *args, **kwargs):
             node = super(lfiles_repo, self).commitctx(*args, **kwargs)
             ctx = self[node]
             for filename in ctx.files():
                 if lfutil.isstandin(filename) and filename in ctx.manifest():
                     realfile = lfutil.splitstandin(filename)
                     lfutil.copytostore(self, ctx.node(), realfile)
 
             return node
 
         # Before commit, largefile standins have not had their
         # contents updated to reflect the hash of their largefile.
         # Do that here.
         def commit(self, text="", user=None, date=None, match=None,
                 force=False, editor=False, extra={}):
             orig = super(lfiles_repo, self).commit
 
             wlock = repo.wlock()
             try:
+                # Case 0: Rebase
+                # We have to take the time to pull down the new largefiles now.
+                # Otherwise if we are rebasing, any largefiles that were
+                # modified in the destination changesets get overwritten, either
+                # by the rebase or in the first commit after the rebase.
+                # updatelfiles will update the dirstate to mark any pulled
+                # largefiles as modified
                 if getattr(repo, "_isrebasing", False):
-                    # We have to take the time to pull down the new
-                    # largefiles now. Otherwise if we are rebasing,
-                    # any largefiles that were modified in the
-                    # destination changesets get overwritten, either
-                    # by the rebase or in the first commit after the
-                    # rebase.
                     lfcommands.updatelfiles(repo.ui, repo)
+                    result = orig(text=text, user=user, date=date, match=match,
+                                    force=force, editor=editor, extra=extra)
+                    return result
                 # 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
                     (not match.anypats() and not match.files())):
                     # 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 = lfutil.openlfdirstate(ui, self)
                     dirtymatch = match_.always(repo.root, repo.getcwd())
                     s = lfdirstate.status(dirtymatch, [], False, False, False)
                     modifiedfiles = []
                     for i in s:
                         modifiedfiles.extend(i)
                     lfiles = lfutil.listlfiles(self)
                     # this only loops through largefiles that exist (not
                     # removed/renamed)
                     for lfile in lfiles:
                         if lfile in modifiedfiles:
                             if os.path.exists(self.wjoin(lfutil.standin(lfile))):
                                 # this handles the case where a rebase is being
                                 # performed and the working copy is not updated
                                 # yet.
                                 if os.path.exists(self.wjoin(lfile)):
                                     lfutil.updatestandin(self,
                                         lfutil.standin(lfile))
                                     lfdirstate.normal(lfile)
                     for lfile in lfdirstate:
                         if lfile in modifiedfiles:
                             if not os.path.exists(
                                     repo.wjoin(lfutil.standin(lfile))):
                                 lfdirstate.drop(lfile)
                     lfdirstate.write()
 
                     return orig(text=text, user=user, date=date, match=match,
                                     force=force, editor=editor, extra=extra)
 
                 for f in match.files():
                     if lfutil.isstandin(f):
                         raise util.Abort(
                             _('file "%s" is a largefile standin') % f,
                             hint=('commit the largefile itself instead'))
 
                 # Case 2: user calls commit with specified patterns: refresh
                 # any matching big files.
                 smatcher = lfutil.composestandinmatcher(self, match)
                 standins = lfutil.dirstate_walk(self.dirstate, smatcher)
 
                 # No matching big files: get out of the way and pass control to
                 # the usual commit() method.
                 if not standins:
                     return orig(text=text, user=user, date=date, match=match,
                                     force=force, editor=editor, extra=extra)
 
                 # 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 = lfutil.openlfdirstate(ui, self)
                 for standin in standins:
                     lfile = lfutil.splitstandin(standin)
                     if lfdirstate[lfile] <> 'r':
                         lfutil.updatestandin(self, standin)
                         lfdirstate.normal(lfile)
                     else:
                         lfdirstate.drop(lfile)
                 lfdirstate.write()
 
                 # 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.
                 lfiles = lfutil.listlfiles(repo)
                 match = copy.copy(match)
                 orig_matchfn = 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 = lfutil.standin(f)
 
                     # ignore known largefiles and standins
                     if f in lfiles or fstandin in standins:
                         continue
 
                     # append directory separator to avoid collisions
                     if not fstandin.endswith(os.sep):
                         fstandin += os.sep
 
                     # prevalidate matching standin directories
                     if util.any(st for st in match._files
                                    if st.startswith(fstandin)):
                         continue
                     actualfiles.append(f)
                 match._files = actualfiles
 
                 def matchfn(f):
                     if orig_matchfn(f):
                         return f not in lfiles
                     else:
                         return f in standins
 
                 match.matchfn = matchfn
                 return orig(text=text, user=user, date=date, match=match,
                                 force=force, editor=editor, extra=extra)
             finally:
                 wlock.release()
 
         def push(self, remote, force=False, revs=None, newbranch=False):
             o = lfutil.findoutgoing(repo, remote, force)
             if o:
                 toupload = set()
                 o = repo.changelog.nodesbetween(o, revs)[0]
                 for n in o:
                     parents = [p for p in repo.changelog.parents(n)
                                if p != node_.nullid]
                     ctx = repo[n]
                     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)
 
                     toupload = toupload.union(
                         set([ctx[f].data().strip()
                              for f in files
                              if lfutil.isstandin(f) and f in ctx]))
                 lfcommands.uploadlfiles(ui, self, remote, toupload)
             return super(lfiles_repo, self).push(remote, force, revs,
                 newbranch)
 
     repo.__class__ = lfiles_repo
 
     def checkrequireslfiles(ui, repo, **kwargs):
         if 'largefiles' not in repo.requirements and util.any(
                 lfutil.shortname+'/' in f[0] for f in repo.store.datafiles()):
             repo.requirements.add('largefiles')
             repo._writerequirements()
 
     ui.setconfig('hooks', 'changegroup.lfiles', checkrequireslfiles)
     ui.setconfig('hooks', 'commit.lfiles', checkrequireslfiles)