def buildfileblobheader(size, flags, version=None):
"""return the header of a remotefilelog blob.
see remotefilelogserver.createfileblob for the format.
approximately the reverse of parsesizeflags.
version could be 0 or 1, or None (auto decide).
"""
# choose v0 if flags is empty, otherwise v1
if version is None:
version = int(bool(flags))
if version == 1:
header = b'v1\n%s%d\n%s%d' % (
constants.METAKEYSIZE,
size,
constants.METAKEYFLAG,
flags,
)
elif version == 0:
if flags:
raise error.ProgrammingError(b'fileblob v0 does not support flag')
header = b'%d' % size
else:
raise error.ProgrammingError(b'unknown fileblob version %d' % version)
return header
def shlexer(data=None, filepath=None, wordchars=None, whitespace=None):
if data is None:
if pycompat.ispy3:
data = open(filepath, 'r', encoding=r'latin1')
else:
data = open(filepath, 'r')
else:
if filepath is not None:
raise error.ProgrammingError(
'shlexer only accepts data or filepath, not both')
if pycompat.ispy3:
data = data.decode('latin1')
l = shlex.shlex(data, infile=filepath, posix=True)
if whitespace is not None:
l.whitespace_split = True
if pycompat.ispy3:
l.whitespace += whitespace.decode('latin1')
else:
l.whitespace += whitespace
if wordchars is not None:
if pycompat.ispy3:
l.wordchars += wordchars.decode('latin1')
else:
l.wordchars += wordchars
if pycompat.ispy3:
return _shlexpy3proxy(l)
return l
def flags(self, node):
if isinstance(node, int):
raise error.ProgrammingError(
b'remotefilelog does not accept integer rev for flags'
)
store = self.repo.contentstore
return store.getmeta(self.filename, node).get(constants.METAKEYFLAG, 0)
def wrappeddispatch(orig, repo, proto, command):
"""Wraps wireprotov1server.dispatch() to record command requests."""
# TRACKING hg46
# For historical reasons, SSH and HTTP use different log events. With
# the unification of the dispatch code in 4.6, we could likely unify these.
# Keep in mind this function is only called on 4.6+: 4.5 has a different
# code path completely.
if isinstance(proto, wireprotoserver.httpv1protocolhandler):
logevent(repo.ui, repo._serverlog, 'BEGIN_PROTOCOL', command)
elif isinstance(proto, wireprotoserver.sshv1protocolhandler):
logevent(repo.ui, repo._serverlog, 'BEGIN_SSH_COMMAND', command)
startusage = resource.getrusage(resource.RUSAGE_SELF)
repo._serverlog.update({
'requestid': str(uuid.uuid1()),
'startcpu': startusage.ru_utime + startusage.ru_stime,
'starttime': time.time(),
'ui': weakref.ref(repo.ui),
})
else:
raise error.ProgrammingError('unhandled protocol handler: %r' % proto)
return orig(repo, proto, command)
def wrappeddispatch(orig, repo, proto, command):
"""Wraps wireprotov1server.dispatch() to record command requests."""
# TRACKING hg46
# For historical reasons, SSH and HTTP use different log events. With
# the unification of the dispatch code in 4.6, we could likely unify these.
# Keep in mind this function is only called on 4.6+: 4.5 has a different
# code path completely.
if isinstance(proto, wireprotoserver.httpv1protocolhandler):
logevent(repo.ui, repo._serverlog, 'BEGIN_PROTOCOL', command)
elif isinstance(proto, wireprotoserver.sshv1protocolhandler):
logevent(repo.ui, repo._serverlog, 'BEGIN_SSH_COMMAND', command)
startusage = resource.getrusage(resource.RUSAGE_SELF)
repo._serverlog.update({
'requestid': pycompat.bytestr(uuid.uuid1()),
'startcpu': startusage.ru_utime + startusage.ru_stime,
'starttime': time.time(),
'ui': weakref.ref(repo.ui),
})
else:
raise error.ProgrammingError(b'unhandled protocol handler: %r' % proto)
# If the return type is a `pushres`, `_sshv1respondbytes` will be called twice.
# We only want to log a completed SSH event on the second call, so flip the
# `ignorecall` flag here.
res = orig(repo, proto, command)
if isinstance(res, wireprototypes.pushres):
repo._serverlog['ignorecall'] = True
return res
def node(self, rev):
# This is a hack.
if isinstance(rev, int):
raise error.ProgrammingError(
b'remotefilelog does not convert integer rev to node'
)
return rev
def _batch(self, pointers, localstore, action):
if action not in ['upload', 'download']:
raise error.ProgrammingError('invalid Git-LFS action: %s' % action)
response = self._batchrequest(pointers, action)
objects = self._extractobjects(response, pointers, action)
total = sum(x.get('size', 0) for x in objects)
sizes = {}
for obj in objects:
sizes[obj.get('oid')] = obj.get('size', 0)
topic = {
'upload': _('lfs uploading'),
'download': _('lfs downloading')
}[action]
if len(objects) > 1:
self.ui.note(
_('lfs: need to transfer %d objects (%s)\n') %
(len(objects), util.bytecount(total)))
self.ui.progress(topic, 0, total=total)
def transfer(chunk):
for obj in chunk:
objsize = obj.get('size', 0)
if self.ui.verbose:
if action == 'download':
msg = _('lfs: downloading %s (%s)\n')
elif action == 'upload':
msg = _('lfs: uploading %s (%s)\n')
self.ui.note(msg %
(obj.get('oid'), util.bytecount(objsize)))
retry = self.retry
while True:
try:
self._basictransfer(obj, action, localstore)
yield 1, obj.get('oid')
break
except socket.error as ex:
if retry > 0:
self.ui.note(
_('lfs: failed: %r (remaining retry %d)\n') %
(ex, retry))
retry -= 1
continue
raise
# Until https multiplexing gets sorted out
if self.ui.configbool('experimental', 'lfs.worker-enable'):
oids = worker.worker(self.ui, 0.1, transfer, (),
sorted(objects, key=lambda o: o.get('oid')))
else:
oids = transfer(sorted(objects, key=lambda o: o.get('oid')))
processed = 0
for _one, oid in oids:
processed += sizes[oid]
self.ui.progress(topic, processed, total=total)
self.ui.note(_('lfs: processed: %s\n') % oid)
self.ui.progress(topic, pos=None, total=total)
def wrapped_advanceboundary(orig, repo, tr, targetphase, nodes):
if targetphase in moves:
raise error.ProgrammingError('already handled phase %r' %
targetphase)
if targetphase not in supported_phases:
raise error.Abort(_('only draft and public phases are supported'))
moves[targetphase] = nodes
return orig(repo, tr, targetphase, nodes)
def getllrev(f):
"""(fctx) -> int"""
# f should not be a linelog revision
if isinstance(f, int):
raise error.ProgrammingError('f should not be an int')
# f is a fctx, allocate linelog rev on demand
hsh = f.node()
rev = revmap.hsh2rev(hsh)
if rev is None:
rev = revmap.append(hsh, sidebranch=True, path=f.path())
return rev
def extsetup(ui):
# It's easier for tests to define the server behavior via environment
# variables than config options. This is because `hg serve --stdio`
# has to be invoked with a certain form for security reasons and
# `dummyssh` can't just add `--config` flags to the command line.
servermode = ui.environ.get(b'SSHSERVERMODE')
if servermode == b'banner':
wireprotoserver.sshserver = bannerserver
elif servermode == b'no-hello':
wireprotoserver.sshserver = prehelloserver
elif servermode:
raise error.ProgrammingError(b'unknown server mode: %s' % servermode)
peermode = ui.config(b'sshpeer', b'mode')
if peermode == b'extra-handshake-commands':
extensions.wrapfunction(sshpeer, '_performhandshake', performhandshake)
elif peermode:
raise error.ProgrammingError(b'unknown peer mode: %s' % peermode)
def rev(self, node):
validatenode(node)
try:
self._indexbynode[node]
except KeyError:
raise error.LookupError(node, self._indexpath, _('no node'))
for rev, entry in self._indexbyrev.items():
if entry[b'node'] == node:
return rev
raise error.ProgrammingError('this should not occur')
def changelogadd(orig, self, *args):
oldlen = len(self)
node = orig(self, *args)
newlen = len(self)
if oldlen != newlen:
for oldargs in pendingfilecommits:
log, rt, tr, link, p1, p2, n, fl, c, m = oldargs
linknode = self.node(link)
if linknode == node:
log.addrawrevision(rt, tr, linknode, p1, p2, n, fl, c, m)
else:
raise error.ProgrammingError(
'pending multiple integer revisions are not supported')
else:
# "link" is actually wrong here (it is set to len(changelog))
# if changelog remains unchanged, skip writing file revisions
# but still do a sanity check about pending multiple revisions
if len(set(x[3] for x in pendingfilecommits)) > 1:
raise error.ProgrammingError(
'pending multiple integer revisions are not supported')
del pendingfilecommits[:]
return node
def emitrevisions(
self,
nodes,
nodesorder=None,
revisiondata=False,
assumehaveparentrevisions=False,
deltamode=repository.CG_DELTAMODE_STD,
sidedata_helpers=None,
):
if nodesorder not in (b'nodes', b'storage', b'linear', None):
raise error.ProgrammingError(
b'unhandled value for nodesorder: %s' % nodesorder)
nodes = [n for n in nodes if n != sha1nodeconstants.nullid]
if not nodes:
return
# TODO perform in a single query.
res = self._db.execute(
'SELECT revnum, deltaid FROM fileindex '
'WHERE pathid=? '
' AND node in (%s)' % (','.join(['?'] * len(nodes))),
tuple([self._pathid] + nodes),
)
deltabases = {}
for rev, deltaid in res:
res = self._db.execute(
'SELECT revnum from fileindex WHERE pathid=? AND deltaid=?',
(self._pathid, deltaid),
)
deltabases[rev] = res.fetchone()[0]
# TODO define revdifffn so we can use delta from storage.
for delta in storageutil.emitrevisions(
self,
nodes,
nodesorder,
sqliterevisiondelta,
deltaparentfn=deltabases.__getitem__,
revisiondata=revisiondata,
assumehaveparentrevisions=assumehaveparentrevisions,
deltamode=deltamode,
sidedata_helpers=sidedata_helpers,
):
yield delta
def buildpackmeta(metadict):
"""like _buildpackmeta, but typechecks metadict and normalize it.
This means, METAKEYSIZE and METAKEYSIZE should have integers as values,
and METAKEYFLAG will be dropped if its value is 0.
"""
newmeta = {}
for k, v in (metadict or {}).iteritems():
expectedtype = _metaitemtypes.get(k, (bytes,))
if not isinstance(v, expectedtype):
raise error.ProgrammingError('packmeta: wrong type of key %s' % k)
# normalize int to binary buffer
if int in expectedtype:
# optimization: remove flag if it's 0 to save space
if k == constants.METAKEYFLAG and v == 0:
continue
v = int2bin(v)
newmeta[k] = v
return _buildpackmeta(newmeta)
def emitrevisions(self,
nodes,
nodesorder=None,
revisiondata=False,
assumehaveparentrevisions=False,
deltaprevious=False):
if nodesorder not in ('nodes', 'storage', 'linear', None):
raise error.ProgrammingError('unhandled value for nodesorder: %s' %
nodesorder)
nodes = [n for n in nodes if n != nullid]
if not nodes:
return
# TODO perform in a single query.
res = self._db.execute(
r'SELECT revnum, deltaid FROM fileindex '
r'WHERE pathid=? '
r' AND node in (%s)' % (r','.join([r'?'] * len(nodes))),
tuple([self._pathid] + nodes))
deltabases = {}
for rev, deltaid in res:
res = self._db.execute(
r'SELECT revnum from fileindex WHERE pathid=? AND deltaid=?',
(self._pathid, deltaid))
deltabases[rev] = res.fetchone()[0]
# TODO define revdifffn so we can use delta from storage.
for delta in storageutil.emitrevisions(
self,
nodes,
nodesorder,
sqliterevisiondelta,
deltaparentfn=deltabases.__getitem__,
revisiondata=revisiondata,
assumehaveparentrevisions=assumehaveparentrevisions,
deltaprevious=deltaprevious):
yield delta
def performhandshake(orig, ui, stdin, stdout, stderr):
"""Wrapped version of sshpeer._performhandshake to send extra commands."""
mode = ui.config(b'sshpeer', b'handshake-mode')
if mode == b'pre-no-args':
ui.debug(b'sending no-args command\n')
stdin.write(b'no-args\n')
stdin.flush()
return orig(ui, stdin, stdout, stderr)
elif mode == b'pre-multiple-no-args':
ui.debug(b'sending unknown1 command\n')
stdin.write(b'unknown1\n')
ui.debug(b'sending unknown2 command\n')
stdin.write(b'unknown2\n')
ui.debug(b'sending unknown3 command\n')
stdin.write(b'unknown3\n')
stdin.flush()
return orig(ui, stdin, stdout, stderr)
else:
raise error.ProgrammingError(b'unknown HANDSHAKECOMMANDMODE: %s' %
mode)
def walk(tree):
op = tree[0]
if op == b'symbol':
drev = _parsedrev(tree[1])
if drev:
return smartset.baseset([drev])
elif tree[1] in _knownstatusnames:
drevs = [r for r in validids
if _getstatusname(prefetched[r]) == tree[1]]
return smartset.baseset(drevs)
else:
raise error.Abort(_(b'unknown symbol: %s') % tree[1])
elif op in {b'and_', b'add', b'sub'}:
assert len(tree) == 3
return getattr(operator, op)(walk(tree[1]), walk(tree[2]))
elif op == b'group':
return walk(tree[1])
elif op == b'ancestors':
return getstack(walk(tree[1]))
else:
raise error.ProgrammingError(b'illegal tree: %r' % tree)
def _buildpackmeta(metadict):
"""reverse of _parsepackmeta, dict -> bytes (<metadata-list>)
The dict contains raw content - both keys and values are strings.
Upper-level business may want to serialize some of other types (like
integers) to strings before calling this function.
raise ProgrammingError when metadata key is illegal, or ValueError if
length limit is exceeded
"""
metabuf = ''
for k, v in sorted((metadict or {}).iteritems()):
if len(k) != 1:
raise error.ProgrammingError('packmeta: illegal key: %s' % k)
if len(v) > 0xfffe:
raise ValueError('metadata value is too long: 0x%x > 0xfffe'
% len(v))
metabuf += k
metabuf += struct.pack('!H', len(v))
metabuf += v
# len(metabuf) is guaranteed representable in 4 bytes, because there are
# only 256 keys, and for each value, len(value) <= 0xfffe.
return metabuf
def _find_nearest_ancestor_introducing_node(
db, gitrepo, file_path, walk_start, filenode
):
"""Find the nearest ancestor that introduces a file node.
Args:
db: a handle to our sqlite database.
gitrepo: A pygit2.Repository instance.
file_path: the path of a file in the repo
walk_start: a pygit2.Oid that is a commit where we should start walking
for our nearest ancestor.
Returns:
A hexlified SHA that is the commit ID of the next-nearest parent.
"""
assert isinstance(file_path, str), 'file_path must be str, got %r' % type(
file_path
)
assert isinstance(filenode, str), 'filenode must be str, got %r' % type(
filenode
)
parent_options = {
row[0].decode('ascii')
for row in db.execute(
'SELECT node FROM changedfiles '
'WHERE filename = ? AND filenode = ?',
(file_path, filenode),
)
}
inner_walker = gitrepo.walk(walk_start, _OUR_ORDER)
for w in inner_walker:
if w.id.hex in parent_options:
return w.id.hex
raise error.ProgrammingError(
'Unable to find introducing commit for %s node %s from %s',
(file_path, filenode, walk_start),
)
def append(self, hsh, sidebranch=False, path=None, flush=False):
"""add a binary hg hash and return the mapped linelog revision.
if flush is True, incrementally update the file.
"""
if hsh in self._hsh2rev:
raise error.CorruptedFileError(b'%r is in revmap already' %
hex(hsh))
if len(hsh) != _hshlen:
raise hgerror.ProgrammingError(b'hsh must be %d-char long' %
_hshlen)
idx = len(self._rev2hsh)
flag = 0
if sidebranch:
flag |= sidebranchflag
if path is not None and path != self._renamepaths[-1]:
flag |= renameflag
self._renamerevs.append(idx)
self._renamepaths.append(path)
self._rev2hsh.append(hsh)
self._rev2flag.append(flag)
self._hsh2rev[hsh] = idx
if flush:
self.flush()
return idx
def _batch(self, pointers, localstore, action):
if action not in [b'upload', b'download']:
raise error.ProgrammingError(b'invalid Git-LFS action: %s' % action)
response = self._batchrequest(pointers, action)
objects = self._extractobjects(response, pointers, action)
total = sum(x.get(b'size', 0) for x in objects)
sizes = {}
for obj in objects:
sizes[obj.get(b'oid')] = obj.get(b'size', 0)
topic = {
b'upload': _(b'lfs uploading'),
b'download': _(b'lfs downloading'),
}[action]
if len(objects) > 1:
self.ui.note(
_(b'lfs: need to transfer %d objects (%s)\n')
% (len(objects), util.bytecount(total))
)
def transfer(chunk):
for obj in chunk:
objsize = obj.get(b'size', 0)
if self.ui.verbose:
if action == b'download':
msg = _(b'lfs: downloading %s (%s)\n')
elif action == b'upload':
msg = _(b'lfs: uploading %s (%s)\n')
self.ui.note(
msg % (obj.get(b'oid'), util.bytecount(objsize))
)
retry = self.retry
while True:
try:
self._basictransfer(obj, action, localstore)
yield 1, obj.get(b'oid')
break
except socket.error as ex:
if retry > 0:
self.ui.note(
_(b'lfs: failed: %r (remaining retry %d)\n')
% (stringutil.forcebytestr(ex), retry)
)
retry -= 1
continue
raise
# Until https multiplexing gets sorted out
if self.ui.configbool(b'experimental', b'lfs.worker-enable'):
oids = worker.worker(
self.ui,
0.1,
transfer,
(),
sorted(objects, key=lambda o: o.get(b'oid')),
)
else:
oids = transfer(sorted(objects, key=lambda o: o.get(b'oid')))
with self.ui.makeprogress(topic, total=total) as progress:
progress.update(0)
processed = 0
blobs = 0
for _one, oid in oids:
processed += sizes[oid]
blobs += 1
progress.update(processed)
self.ui.note(_(b'lfs: processed: %s\n') % oid)
if blobs > 0:
if action == b'upload':
self.ui.status(
_(b'lfs: uploaded %d files (%s)\n')
% (blobs, util.bytecount(processed))
)
elif action == b'download':
self.ui.status(
_(b'lfs: downloaded %d files (%s)\n')
% (blobs, util.bytecount(processed))
)
def _addrawrevision(
self,
node,
revisiondata,
transaction,
linkrev,
p1,
p2,
storedelta=None,
flags=0,
):
if self._pathid is None:
res = self._db.execute(
'INSERT INTO filepath (path) VALUES (?)', (self._path,)
)
self._pathid = res.lastrowid
# For simplicity, always store a delta against p1.
# TODO we need a lot more logic here to make behavior reasonable.
if storedelta:
deltabase, delta = storedelta
if isinstance(deltabase, int):
deltabase = self.node(deltabase)
else:
assert revisiondata is not None
deltabase = p1
if deltabase == nullid:
delta = revisiondata
else:
delta = mdiff.textdiff(
self.revision(self.rev(deltabase)), revisiondata
)
# File index stores a pointer to its delta and the parent delta.
# The parent delta is stored via a pointer to the fileindex PK.
if deltabase == nullid:
baseid = None
else:
baseid = self._revisions[deltabase].rid
# Deltas are stored with a hash of their content. This allows
# us to de-duplicate. The table is configured to ignore conflicts
# and it is faster to just insert and silently noop than to look
# first.
deltahash = hashutil.sha1(delta).digest()
if self._compengine == b'zstd':
deltablob = self._cctx.compress(delta)
compression = COMPRESSION_ZSTD
elif self._compengine == b'zlib':
deltablob = zlib.compress(delta)
compression = COMPRESSION_ZLIB
elif self._compengine == b'none':
deltablob = delta
compression = COMPRESSION_NONE
else:
raise error.ProgrammingError(
b'unhandled compression engine: %s' % self._compengine
)
# Don't store compressed data if it isn't practical.
if len(deltablob) >= len(delta):
deltablob = delta
compression = COMPRESSION_NONE
deltaid = insertdelta(self._db, compression, deltahash, deltablob)
rev = len(self)
if p1 == nullid:
p1rev = nullrev
else:
p1rev = self._nodetorev[p1]
if p2 == nullid:
p2rev = nullrev
else:
p2rev = self._nodetorev[p2]
rid = self._db.execute(
'INSERT INTO fileindex ('
' pathid, revnum, node, p1rev, p2rev, linkrev, flags, '
' deltaid, deltabaseid) '
' VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)',
(
self._pathid,
rev,
node,
p1rev,
p2rev,
linkrev,
flags,
deltaid,
baseid,
),
).lastrowid
entry = revisionentry(
rid=rid,
rev=rev,
node=node,
p1rev=p1rev,
p2rev=p2rev,
p1node=p1,
p2node=p2,
linkrev=linkrev,
flags=flags,
)
self._nodetorev[node] = rev
self._revtonode[rev] = node
self._revisions[node] = entry
return node
def censorrevision(self, tr, censornode, tombstone=b''):
tombstone = storageutil.packmeta({b'censored': tombstone}, b'')
# This restriction is cargo culted from revlogs and makes no sense for
# SQLite, since columns can be resized at will.
if len(tombstone) > len(self.rawdata(censornode)):
raise error.Abort(
_(b'censor tombstone must be no longer than censored data')
)
# We need to replace the censored revision's data with the tombstone.
# But replacing that data will have implications for delta chains that
# reference it.
#
# While "better," more complex strategies are possible, we do something
# simple: we find delta chain children of the censored revision and we
# replace those incremental deltas with fulltexts of their corresponding
# revision. Then we delete the now-unreferenced delta and original
# revision and insert a replacement.
# Find the delta to be censored.
censoreddeltaid = self._db.execute(
'SELECT deltaid FROM fileindex WHERE id=?',
(self._revisions[censornode].rid,),
).fetchone()[0]
# Find all its delta chain children.
# TODO once we support storing deltas for !files, we'll need to look
# for those delta chains too.
rows = list(
self._db.execute(
'SELECT id, pathid, node FROM fileindex '
'WHERE deltabaseid=? OR deltaid=?',
(censoreddeltaid, censoreddeltaid),
)
)
for row in rows:
rid, pathid, node = row
fulltext = resolvedeltachain(
self._db, pathid, node, {}, {-1: None}, zstddctx=self._dctx
)
deltahash = hashutil.sha1(fulltext).digest()
if self._compengine == b'zstd':
deltablob = self._cctx.compress(fulltext)
compression = COMPRESSION_ZSTD
elif self._compengine == b'zlib':
deltablob = zlib.compress(fulltext)
compression = COMPRESSION_ZLIB
elif self._compengine == b'none':
deltablob = fulltext
compression = COMPRESSION_NONE
else:
raise error.ProgrammingError(
b'unhandled compression engine: %s' % self._compengine
)
if len(deltablob) >= len(fulltext):
deltablob = fulltext
compression = COMPRESSION_NONE
deltaid = insertdelta(self._db, compression, deltahash, deltablob)
self._db.execute(
'UPDATE fileindex SET deltaid=?, deltabaseid=NULL '
'WHERE id=?',
(deltaid, rid),
)
# Now create the tombstone delta and replace the delta on the censored
# node.
deltahash = hashutil.sha1(tombstone).digest()
tombstonedeltaid = insertdelta(
self._db, COMPRESSION_NONE, deltahash, tombstone
)
flags = self._revisions[censornode].flags
flags |= FLAG_CENSORED
self._db.execute(
'UPDATE fileindex SET flags=?, deltaid=?, deltabaseid=NULL '
'WHERE pathid=? AND node=?',
(flags, tombstonedeltaid, self._pathid, censornode),
)
self._db.execute('DELETE FROM delta WHERE id=?', (censoreddeltaid,))
self._refreshindex()
self._revisioncache.clear()
def join(self, path):
"""split the path at first two characters, like: XX/XXXXX..."""
if not _lfsre.match(path):
raise error.ProgrammingError('unexpected lfs path: %s' % path)
return super(lfsvfs, self).join(path[0:2], path[2:])
def _index_repo(
gitrepo,
db,
logfn=lambda x: None,
progress_factory=lambda *args, **kwargs: None,
):
# Identify all references so we can tell the walker to visit all of them.
all_refs = gitrepo.listall_references()
possible_heads = set()
prog = progress_factory(b'refs')
for pos, ref in enumerate(all_refs):
if prog is not None:
prog.update(pos)
if not (
ref.startswith('refs/heads/') # local branch
or ref.startswith('refs/tags/') # tag
or ref.startswith('refs/remotes/') # remote branch
or ref.startswith('refs/hg/') # from this extension
):
continue
try:
start = gitrepo.lookup_reference(ref).peel(pygit2.GIT_OBJ_COMMIT)
except ValueError:
# No commit to be found, so we don't care for hg's purposes.
continue
possible_heads.add(start.id)
# Optimization: if the list of heads hasn't changed, don't
# reindex, the changelog. This doesn't matter on small
# repositories, but on even moderately deep histories (eg cpython)
# this is a very important performance win.
#
# TODO: we should figure out how to incrementally index history
# (preferably by detecting rewinds!) so that we don't have to do a
# full changelog walk every time a new commit is created.
cache_heads = {
pycompat.sysstr(x[0])
for x in db.execute('SELECT node FROM possible_heads')
}
walker = None
cur_cache_heads = {h.hex for h in possible_heads}
if cur_cache_heads == cache_heads:
return
logfn(b'heads mismatch, rebuilding dagcache\n')
for start in possible_heads:
if walker is None:
walker = gitrepo.walk(start, _OUR_ORDER)
else:
walker.push(start)
# Empty out the existing changelog. Even for large-ish histories
# we can do the top-level "walk all the commits" dance very
# quickly as long as we don't need to figure out the changed files
# list.
db.execute('DELETE FROM changelog')
if prog is not None:
prog.complete()
prog = progress_factory(b'commits')
# This walker is sure to visit all the revisions in history, but
# only once.
for pos, commit in enumerate(walker):
if prog is not None:
prog.update(pos)
p1 = p2 = nullhex
if len(commit.parents) > 2:
raise error.ProgrammingError(
(
b"git support can't handle octopus merges, "
b"found a commit with %d parents :("
)
% len(commit.parents)
)
if commit.parents:
p1 = commit.parents[0].id.hex
if len(commit.parents) == 2:
p2 = commit.parents[1].id.hex
db.execute(
'INSERT INTO changelog (rev, node, p1, p2) VALUES(?, ?, ?, ?)',
(pos, commit.id.hex, p1, p2),
)
num_changedfiles = db.execute(
"SELECT COUNT(*) from changedfiles WHERE node = ?",
(commit.id.hex,),
).fetchone()[0]
if not num_changedfiles:
files = {}
# I *think* we only need to check p1 for changed files
# (and therefore linkrevs), because any node that would
# actually have this commit as a linkrev would be
# completely new in this rev.
p1 = commit.parents[0].id.hex if commit.parents else None
if p1 is not None:
patchgen = gitrepo.diff(p1, commit.id.hex, flags=_DIFF_FLAGS)
else:
patchgen = commit.tree.diff_to_tree(
swap=True, flags=_DIFF_FLAGS
)
new_files = (p.delta.new_file for p in patchgen)
files = {
nf.path: nf.id.hex for nf in new_files if nf.id.raw != nullid
}
for p, n in files.items():
# We intentionally set NULLs for any file parentage
# information so it'll get demand-computed later. We
# used to do it right here, and it was _very_ slow.
db.execute(
'INSERT INTO changedfiles ('
'node, filename, filenode, p1node, p1filenode, p2node, '
'p2filenode) VALUES(?, ?, ?, ?, ?, ?, ?)',
(commit.id.hex, p, n, None, None, None, None),
)
db.execute('DELETE FROM heads')
db.execute('DELETE FROM possible_heads')
for hid in possible_heads:
h = hid.hex
db.execute('INSERT INTO possible_heads (node) VALUES(?)', (h,))
haschild = db.execute(
'SELECT COUNT(*) FROM changelog WHERE p1 = ? OR p2 = ?', (h, h)
).fetchone()[0]
if not haschild:
db.execute('INSERT INTO heads (node) VALUES(?)', (h,))
db.commit()
if prog is not None:
prog.complete()
def fill_in_filelog(gitrepo, db, startcommit, path, startfilenode):
"""Given a starting commit and path, fill in a filelog's parent pointers.
Args:
gitrepo: a pygit2.Repository
db: a handle to our sqlite database
startcommit: a hexlified node id for the commit to start at
path: the path of the file whose parent pointers we should fill in.
filenode: the hexlified node id of the file at startcommit
TODO: make filenode optional
"""
assert isinstance(
startcommit, str
), 'startcommit must be str, got %r' % type(startcommit)
assert isinstance(
startfilenode, str
), 'startfilenode must be str, got %r' % type(startfilenode)
visit = collections.deque([(startcommit, startfilenode)])
while visit:
cnode, filenode = visit.popleft()
commit = gitrepo[cnode]
parents = []
for parent in commit.parents:
t = parent.tree
for comp in path.split('/'):
try:
t = gitrepo[t[comp].id]
except KeyError:
break
else:
introducer = _find_nearest_ancestor_introducing_node(
db, gitrepo, path, parent.id, t.id.hex
)
parents.append((introducer, t.id.hex))
p1node = p1fnode = p2node = p2fnode = gitutil.nullgit
for par, parfnode in parents:
found = int(
db.execute(
'SELECT COUNT(*) FROM changedfiles WHERE '
'node = ? AND filename = ? AND filenode = ? AND '
'p1node NOT NULL',
(par, path, parfnode),
).fetchone()[0]
)
if found == 0:
assert par is not None
visit.append((par, parfnode))
if parents:
p1node, p1fnode = parents[0]
if len(parents) == 2:
p2node, p2fnode = parents[1]
if len(parents) > 2:
raise error.ProgrammingError(
b"git support can't handle octopus merges"
)
db.execute(
'UPDATE changedfiles SET '
'p1node = ?, p1filenode = ?, p2node = ?, p2filenode = ? '
'WHERE node = ? AND filename = ? AND filenode = ?',
(p1node, p1fnode, p2node, p2fnode, commit.id.hex, path, filenode),
)
db.commit()
