streamclone: also stream caches to the client

When stream clone is used over bundle2, relevant cache files are also streamed.
This is expected to be a massive performance win for clone since no important
cache will have to be recomputed.

Some performance numbers:

(All times are wall-clock times in seconds, 2 attempts per case.)

# Mozilla-Central

## Clone over ssh over lan
V1 streaming: 234.3 239.6
V2 streaming: 248.4 243.7

## Clone over ssh over Internet
V1 streaming: 175.5 110.9
V2 streaming: 109.1 111.0

## Clone over HTTP over lan
V1 streaming: 105.3 105.6
V2 streaming: 112.7 111.4

## Clone over HTTP over internet
V1 streaming: 105.6 114.6
V2 streaming: 226.7 225.9

## Hg tags
V1 streaming (no cache): 1.084 1.071
V2 streaming (cache):    0.312 0.325

## Hg branches
V1 streaming (no cache):   14.047 14.148
V2 streaming (with cache):  0.312  0.333

# Pypy

## Clone over ssh over internet
V1 streaming: 29.4 30.1
V2 streaming: 31.2 30.1

## Clone over http over internet
V1 streaming: 29.7 29.7
V2 streaming: 75.2 72.9

(since ssh and lan are not affected, there seems to be an issue with how we
read/write the http stream on connection with latency, unrelated to the format)

## Hg tags
V1 streaming (no cache):   1.752 1.664
V2 streaming (with cache): 0.274 0.260

## Hg branches
V1 streaming (no cache):   4.469 4.728
V2 streaming (with cache): 0.318 0.321

# Private repository:
    * 500K revision revisions
    * 11K topological heads
    * 28K branch heads

## hg tags
no cache:   1543.332
with cache:    4.900

## hg branches
no cache:   91.828
with cache:  2.955

mercurial/streamclone.py

@@ -1,595 +1,634 b''
 # streamclone.py - producing and consuming streaming repository data
 #
 # Copyright 2015 Gregory Szorc <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import contextlib
 import os
 import struct
 import tempfile
+import warnings
 
 from .i18n import _
 from . import (
     branchmap,
+    cacheutil,
     error,
     phases,
     store,
     util,
 )
 
 def canperformstreamclone(pullop, bundle2=False):
     """Whether it is possible to perform a streaming clone as part of pull.
 
     ``bundle2`` will cause the function to consider stream clone through
     bundle2 and only through bundle2.
 
     Returns a tuple of (supported, requirements). ``supported`` is True if
     streaming clone is supported and False otherwise. ``requirements`` is
     a set of repo requirements from the remote, or ``None`` if stream clone
     isn't supported.
     """
     repo = pullop.repo
     remote = pullop.remote
 
     bundle2supported = False
     if pullop.canusebundle2:
         if 'v2' in pullop.remotebundle2caps.get('stream', []):
             bundle2supported = True
         # else
             # Server doesn't support bundle2 stream clone or doesn't support
             # the versions we support. Fall back and possibly allow legacy.
 
     # Ensures legacy code path uses available bundle2.
     if bundle2supported and not bundle2:
         return False, None
     # Ensures bundle2 doesn't try to do a stream clone if it isn't supported.
     elif bundle2 and not bundle2supported:
         return False, None
 
     # Streaming clone only works on empty repositories.
     if len(repo):
         return False, None
 
     # Streaming clone only works if all data is being requested.
     if pullop.heads:
         return False, None
 
     streamrequested = pullop.streamclonerequested
 
     # If we don't have a preference, let the server decide for us. This
     # likely only comes into play in LANs.
     if streamrequested is None:
         # The server can advertise whether to prefer streaming clone.
         streamrequested = remote.capable('stream-preferred')
 
     if not streamrequested:
         return False, None
 
     # In order for stream clone to work, the client has to support all the
     # requirements advertised by the server.
     #
     # The server advertises its requirements via the "stream" and "streamreqs"
     # capability. "stream" (a value-less capability) is advertised if and only
     # if the only requirement is "revlogv1." Else, the "streamreqs" capability
     # is advertised and contains a comma-delimited list of requirements.
     requirements = set()
     if remote.capable('stream'):
         requirements.add('revlogv1')
     else:
         streamreqs = remote.capable('streamreqs')
         # This is weird and shouldn't happen with modern servers.
         if not streamreqs:
             pullop.repo.ui.warn(_(
                 'warning: stream clone requested but server has them '
                 'disabled\n'))
             return False, None
 
         streamreqs = set(streamreqs.split(','))
         # Server requires something we don't support. Bail.
         missingreqs = streamreqs - repo.supportedformats
         if missingreqs:
             pullop.repo.ui.warn(_(
                 'warning: stream clone requested but client is missing '
                 'requirements: %s\n') % ', '.join(sorted(missingreqs)))
             pullop.repo.ui.warn(
                 _('(see https://www.mercurial-scm.org/wiki/MissingRequirement '
                   'for more information)\n'))
             return False, None
         requirements = streamreqs
 
     return True, requirements
 
 def maybeperformlegacystreamclone(pullop):
     """Possibly perform a legacy stream clone operation.
 
     Legacy stream clones are performed as part of pull but before all other
     operations.
 
     A legacy stream clone will not be performed if a bundle2 stream clone is
     supported.
     """
     supported, requirements = canperformstreamclone(pullop)
 
     if not supported:
         return
 
     repo = pullop.repo
     remote = pullop.remote
 
     # Save remote branchmap. We will use it later to speed up branchcache
     # creation.
     rbranchmap = None
     if remote.capable('branchmap'):
         rbranchmap = remote.branchmap()
 
     repo.ui.status(_('streaming all changes\n'))
 
     fp = remote.stream_out()
     l = fp.readline()
     try:
         resp = int(l)
     except ValueError:
         raise error.ResponseError(
             _('unexpected response from remote server:'), l)
     if resp == 1:
         raise error.Abort(_('operation forbidden by server'))
     elif resp == 2:
         raise error.Abort(_('locking the remote repository failed'))
     elif resp != 0:
         raise error.Abort(_('the server sent an unknown error code'))
 
     l = fp.readline()
     try:
         filecount, bytecount = map(int, l.split(' ', 1))
     except (ValueError, TypeError):
         raise error.ResponseError(
             _('unexpected response from remote server:'), l)
 
     with repo.lock():
         consumev1(repo, fp, filecount, bytecount)
 
         # new requirements = old non-format requirements +
         #                    new format-related remote requirements
         # requirements from the streamed-in repository
         repo.requirements = requirements | (
                 repo.requirements - repo.supportedformats)
         repo._applyopenerreqs()
         repo._writerequirements()
 
         if rbranchmap:
             branchmap.replacecache(repo, rbranchmap)
 
         repo.invalidate()
 
 def allowservergeneration(repo):
     """Whether streaming clones are allowed from the server."""
     if not repo.ui.configbool('server', 'uncompressed', untrusted=True):
         return False
 
     # The way stream clone works makes it impossible to hide secret changesets.
     # So don't allow this by default.
     secret = phases.hassecret(repo)
     if secret:
         return repo.ui.configbool('server', 'uncompressedallowsecret')
 
     return True
 
 # This is it's own function so extensions can override it.
 def _walkstreamfiles(repo):
     return repo.store.walk()
 
 def generatev1(repo):
     """Emit content for version 1 of a streaming clone.
 
     This returns a 3-tuple of (file count, byte size, data iterator).
 
     The data iterator consists of N entries for each file being transferred.
     Each file entry starts as a line with the file name and integer size
     delimited by a null byte.
 
     The raw file data follows. Following the raw file data is the next file
     entry, or EOF.
 
     When used on the wire protocol, an additional line indicating protocol
     success will be prepended to the stream. This function is not responsible
     for adding it.
 
     This function will obtain a repository lock to ensure a consistent view of
     the store is captured. It therefore may raise LockError.
     """
     entries = []
     total_bytes = 0
     # Get consistent snapshot of repo, lock during scan.
     with repo.lock():
         repo.ui.debug('scanning\n')
         for name, ename, size in _walkstreamfiles(repo):
             if size:
                 entries.append((name, size))
                 total_bytes += size
 
     repo.ui.debug('%d files, %d bytes to transfer\n' %
                   (len(entries), total_bytes))
 
     svfs = repo.svfs
     debugflag = repo.ui.debugflag
 
     def emitrevlogdata():
         for name, size in entries:
             if debugflag:
                 repo.ui.debug('sending %s (%d bytes)\n' % (name, size))
             # partially encode name over the wire for backwards compat
             yield '%s\0%d\n' % (store.encodedir(name), size)
             # auditing at this stage is both pointless (paths are already
             # trusted by the local repo) and expensive
             with svfs(name, 'rb', auditpath=False) as fp:
                 if size <= 65536:
                     yield fp.read(size)
                 else:
                     for chunk in util.filechunkiter(fp, limit=size):
                         yield chunk
 
     return len(entries), total_bytes, emitrevlogdata()
 
 def generatev1wireproto(repo):
     """Emit content for version 1 of streaming clone suitable for the wire.
 
     This is the data output from ``generatev1()`` with 2 header lines. The
     first line indicates overall success. The 2nd contains the file count and
     byte size of payload.
 
     The success line contains "0" for success, "1" for stream generation not
     allowed, and "2" for error locking the repository (possibly indicating
     a permissions error for the server process).
     """
     if not allowservergeneration(repo):
         yield '1\n'
         return
 
     try:
         filecount, bytecount, it = generatev1(repo)
     except error.LockError:
         yield '2\n'
         return
 
     # Indicates successful response.
     yield '0\n'
     yield '%d %d\n' % (filecount, bytecount)
     for chunk in it:
         yield chunk
 
 def generatebundlev1(repo, compression='UN'):
     """Emit content for version 1 of a stream clone bundle.
 
     The first 4 bytes of the output ("HGS1") denote this as stream clone
     bundle version 1.
 
     The next 2 bytes indicate the compression type. Only "UN" is currently
     supported.
 
     The next 16 bytes are two 64-bit big endian unsigned integers indicating
     file count and byte count, respectively.
 
     The next 2 bytes is a 16-bit big endian unsigned short declaring the length
     of the requirements string, including a trailing \0. The following N bytes
     are the requirements string, which is ASCII containing a comma-delimited
     list of repo requirements that are needed to support the data.
 
     The remaining content is the output of ``generatev1()`` (which may be
     compressed in the future).
 
     Returns a tuple of (requirements, data generator).
     """
     if compression != 'UN':
         raise ValueError('we do not support the compression argument yet')
 
     requirements = repo.requirements & repo.supportedformats
     requires = ','.join(sorted(requirements))
 
     def gen():
         yield 'HGS1'
         yield compression
 
         filecount, bytecount, it = generatev1(repo)
         repo.ui.status(_('writing %d bytes for %d files\n') %
                          (bytecount, filecount))
 
         yield struct.pack('>QQ', filecount, bytecount)
         yield struct.pack('>H', len(requires) + 1)
         yield requires + '\0'
 
         # This is where we'll add compression in the future.
         assert compression == 'UN'
 
         seen = 0
         repo.ui.progress(_('bundle'), 0, total=bytecount, unit=_('bytes'))
 
         for chunk in it:
             seen += len(chunk)
             repo.ui.progress(_('bundle'), seen, total=bytecount,
                              unit=_('bytes'))
             yield chunk
 
         repo.ui.progress(_('bundle'), None)
 
     return requirements, gen()
 
 def consumev1(repo, fp, filecount, bytecount):
     """Apply the contents from version 1 of a streaming clone file handle.
 
     This takes the output from "stream_out" and applies it to the specified
     repository.
 
     Like "stream_out," the status line added by the wire protocol is not
     handled by this function.
     """
     with repo.lock():
         repo.ui.status(_('%d files to transfer, %s of data\n') %
                        (filecount, util.bytecount(bytecount)))
         handled_bytes = 0
         repo.ui.progress(_('clone'), 0, total=bytecount, unit=_('bytes'))
         start = util.timer()
 
         # TODO: get rid of (potential) inconsistency
         #
         # If transaction is started and any @filecache property is
         # changed at this point, it causes inconsistency between
         # in-memory cached property and streamclone-ed file on the
         # disk. Nested transaction prevents transaction scope "clone"
         # below from writing in-memory changes out at the end of it,
         # even though in-memory changes are discarded at the end of it
         # regardless of transaction nesting.
         #
         # But transaction nesting can't be simply prohibited, because
         # nesting occurs also in ordinary case (e.g. enabling
         # clonebundles).
 
         with repo.transaction('clone'):
             with repo.svfs.backgroundclosing(repo.ui, expectedcount=filecount):
                 for i in xrange(filecount):
                     # XXX doesn't support '\n' or '\r' in filenames
                     l = fp.readline()
                     try:
                         name, size = l.split('\0', 1)
                         size = int(size)
                     except (ValueError, TypeError):
                         raise error.ResponseError(
                             _('unexpected response from remote server:'), l)
                     if repo.ui.debugflag:
                         repo.ui.debug('adding %s (%s)\n' %
                                       (name, util.bytecount(size)))
                     # for backwards compat, name was partially encoded
                     path = store.decodedir(name)
                     with repo.svfs(path, 'w', backgroundclose=True) as ofp:
                         for chunk in util.filechunkiter(fp, limit=size):
                             handled_bytes += len(chunk)
                             repo.ui.progress(_('clone'), handled_bytes,
                                              total=bytecount, unit=_('bytes'))
                             ofp.write(chunk)
 
             # force @filecache properties to be reloaded from
             # streamclone-ed file at next access
             repo.invalidate(clearfilecache=True)
 
         elapsed = util.timer() - start
         if elapsed <= 0:
             elapsed = 0.001
         repo.ui.progress(_('clone'), None)
         repo.ui.status(_('transferred %s in %.1f seconds (%s/sec)\n') %
                        (util.bytecount(bytecount), elapsed,
                         util.bytecount(bytecount / elapsed)))
 
 def readbundle1header(fp):
     compression = fp.read(2)
     if compression != 'UN':
         raise error.Abort(_('only uncompressed stream clone bundles are '
             'supported; got %s') % compression)
 
     filecount, bytecount = struct.unpack('>QQ', fp.read(16))
     requireslen = struct.unpack('>H', fp.read(2))[0]
     requires = fp.read(requireslen)
 
     if not requires.endswith('\0'):
         raise error.Abort(_('malformed stream clone bundle: '
                             'requirements not properly encoded'))
 
     requirements = set(requires.rstrip('\0').split(','))
 
     return filecount, bytecount, requirements
 
 def applybundlev1(repo, fp):
     """Apply the content from a stream clone bundle version 1.
 
     We assume the 4 byte header has been read and validated and the file handle
     is at the 2 byte compression identifier.
     """
     if len(repo):
         raise error.Abort(_('cannot apply stream clone bundle on non-empty '
                             'repo'))
 
     filecount, bytecount, requirements = readbundle1header(fp)
     missingreqs = requirements - repo.supportedformats
     if missingreqs:
         raise error.Abort(_('unable to apply stream clone: '
                             'unsupported format: %s') %
                             ', '.join(sorted(missingreqs)))
 
     consumev1(repo, fp, filecount, bytecount)
 
 class streamcloneapplier(object):
     """Class to manage applying streaming clone bundles.
 
     We need to wrap ``applybundlev1()`` in a dedicated type to enable bundle
     readers to perform bundle type-specific functionality.
     """
     def __init__(self, fh):
         self._fh = fh
 
     def apply(self, repo):
         return applybundlev1(repo, self._fh)
 
 # type of file to stream
 _fileappend = 0 # append only file
 _filefull = 1   # full snapshot file
 
+# Source of the file
+_srcstore = 's' # store (svfs)
+_srccache = 'c' # cache (cache)
+
 # This is it's own function so extensions can override it.
 def _walkstreamfullstorefiles(repo):
     """list snapshot file from the store"""
     fnames = []
     if not repo.publishing():
         fnames.append('phaseroots')
     return fnames
 
-def _filterfull(entry, copy, vfs):
+def _filterfull(entry, copy, vfsmap):
     """actually copy the snapshot files"""
-    name, ftype, data = entry
+    src, name, ftype, data = entry
     if ftype != _filefull:
         return entry
-    return (name, ftype, copy(vfs.join(name)))
+    return (src, name, ftype, copy(vfsmap[src].join(name)))
 
 @contextlib.contextmanager
 def maketempcopies():
     """return a function to temporary copy file"""
     files = []
     try:
         def copy(src):
             fd, dst = tempfile.mkstemp()
             os.close(fd)
             files.append(dst)
             util.copyfiles(src, dst, hardlink=True)
             return dst
         yield copy
     finally:
         for tmp in files:
             util.tryunlink(tmp)
 
+def _makemap(repo):
+    """make a (src -> vfs) map for the repo"""
+    vfsmap = {
+        _srcstore: repo.svfs,
+        _srccache: repo.cachevfs,
+    }
+    # we keep repo.vfs out of the on purpose, ther are too many danger there
+    # (eg: .hg/hgrc)
+    assert repo.vfs not in vfsmap.values()
+
+    return vfsmap
+
 def _emit(repo, entries, totalfilesize):
     """actually emit the stream bundle"""
-    vfs = repo.svfs
+    vfsmap = _makemap(repo)
     progress = repo.ui.progress
     progress(_('bundle'), 0, total=totalfilesize, unit=_('bytes'))
     with maketempcopies() as copy:
         try:
             # copy is delayed until we are in the try
-            entries = [_filterfull(e, copy, vfs) for e in entries]
+            entries = [_filterfull(e, copy, vfsmap) for e in entries]
             yield None # this release the lock on the repository
             seen = 0
 
-            for name, ftype, data in entries:
+            for src, name, ftype, data in entries:
+                vfs = vfsmap[src]
+                yield src
                 yield util.uvarintencode(len(name))
                 if ftype == _fileappend:
                     fp = vfs(name)
                     size = data
                 elif ftype == _filefull:
                     fp = open(data, 'rb')
                     size = util.fstat(fp).st_size
                 try:
                     yield util.uvarintencode(size)
                     yield name
                     if size <= 65536:
                         chunks = (fp.read(size),)
                     else:
                         chunks = util.filechunkiter(fp, limit=size)
                     for chunk in chunks:
                         seen += len(chunk)
                         progress(_('bundle'), seen, total=totalfilesize,
                                  unit=_('bytes'))
                         yield chunk
                 finally:
                     fp.close()
         finally:
             progress(_('bundle'), None)
 
 def generatev2(repo):
     """Emit content for version 2 of a streaming clone.
 
     the data stream consists the following entries:
-    1) A varint containing the length of the filename
-    2) A varint containing the length of file data
-    3) N bytes containing the filename (the internal, store-agnostic form)
-    4) N bytes containing the file data
+    1) A char representing the file destination (eg: store or cache)
+    2) A varint containing the length of the filename
+    3) A varint containing the length of file data
+    4) N bytes containing the filename (the internal, store-agnostic form)
+    5) N bytes containing the file data
 
     Returns a 3-tuple of (file count, file size, data iterator).
     """
 
     with repo.lock():
 
         entries = []
         totalfilesize = 0
 
         repo.ui.debug('scanning\n')
         for name, ename, size in _walkstreamfiles(repo):
             if size:
-                entries.append((name, _fileappend, size))
+                entries.append((_srcstore, name, _fileappend, size))
                 totalfilesize += size
         for name in _walkstreamfullstorefiles(repo):
             if repo.svfs.exists(name):
                 totalfilesize += repo.svfs.lstat(name).st_size
-                entries.append((name, _filefull, None))
+                entries.append((_srcstore, name, _filefull, None))
+        for name in cacheutil.cachetocopy(repo):
+            if repo.cachevfs.exists(name):
+                totalfilesize += repo.cachevfs.lstat(name).st_size
+                entries.append((_srccache, name, _filefull, None))
 
         chunks = _emit(repo, entries, totalfilesize)
         first = next(chunks)
         assert first is None
 
     return len(entries), totalfilesize, chunks
 
[e-mail]
+def nested(*ctxs):
+    with warnings.catch_warnings():
+        # For some reason, Python decided 'nested' was deprecated without
+        # replacement. They officially advertised for filtering the deprecation
+        # warning for people who actually need the feature.
+        warnings.filterwarnings("ignore",category=DeprecationWarning)
+        with contextlib.nested(*ctxs):
+            yield
+
 def consumev2(repo, fp, filecount, filesize):
     """Apply the contents from a version 2 streaming clone.
 
     Data is read from an object that only needs to provide a ``read(size)``
     method.
     """
     with repo.lock():
         repo.ui.status(_('%d files to transfer, %s of data\n') %
                        (filecount, util.bytecount(filesize)))
 
         start = util.timer()
         handledbytes = 0
         progress = repo.ui.progress
 
         progress(_('clone'), handledbytes, total=filesize, unit=_('bytes'))
 
-        vfs = repo.svfs
+        vfsmap = _makemap(repo)
 
         with repo.transaction('clone'):
-            with vfs.backgroundclosing(repo.ui):
+            ctxs = (vfs.backgroundclosing(repo.ui)
+                    for vfs in vfsmap.values())
+            with nested(*ctxs):
                 for i in range(filecount):
+                    src = fp.read(1)
+                    vfs = vfsmap[src]
                     namelen = util.uvarintdecodestream(fp)
                     datalen = util.uvarintdecodestream(fp)
 
                     name = fp.read(namelen)
 
                     if repo.ui.debugflag:
-                        repo.ui.debug('adding %s (%s)\n' %
-                                      (name, util.bytecount(datalen)))
+                        repo.ui.debug('adding [%s] %s (%s)\n' %
+                                      (src, name, util.bytecount(datalen)))
 
                     with vfs(name, 'w') as ofp:
                         for chunk in util.filechunkiter(fp, limit=datalen):
                             handledbytes += len(chunk)
                             progress(_('clone'), handledbytes, total=filesize,
                                      unit=_('bytes'))
                             ofp.write(chunk)
 
             # force @filecache properties to be reloaded from
             # streamclone-ed file at next access
             repo.invalidate(clearfilecache=True)
 
         elapsed = util.timer() - start
         if elapsed <= 0:
             elapsed = 0.001
         progress(_('clone'), None)
         repo.ui.status(_('transferred %s in %.1f seconds (%s/sec)\n') %
                        (util.bytecount(handledbytes), elapsed,
                         util.bytecount(handledbytes / elapsed)))
 
 def applybundlev2(repo, fp, filecount, filesize, requirements):
     missingreqs = [r for r in requirements if r not in repo.supported]
     if missingreqs:
         raise error.Abort(_('unable to apply stream clone: '
                             'unsupported format: %s') %
                           ', '.join(sorted(missingreqs)))
 
     consumev2(repo, fp, filecount, filesize)

tests/test-clone-uncompressed.t

@@ -1,330 +1,336 b''
 #require serve
 
 #testcases stream-legacy stream-bundle2
 
 #if stream-bundle2
   $ cat << EOF >> $HGRCPATH
   > [experimental]
   > bundle2.stream = yes
   > EOF
 #endif
 
 Initialize repository
 the status call is to check for issue5130
 
   $ hg init server
   $ cd server
   $ touch foo
   $ hg -q commit -A -m initial
   >>> for i in range(1024):
   ...     with open(str(i), 'wb') as fh:
   ...         fh.write(str(i))
   $ hg -q commit -A -m 'add a lot of files'
   $ hg st
   $ hg serve -p $HGPORT -d --pid-file=hg.pid
   $ cat hg.pid >> $DAEMON_PIDS
   $ cd ..
 
 Basic clone
 
 #if stream-legacy
   $ hg clone --stream -U http://localhost:$HGPORT clone1
   streaming all changes
   1027 files to transfer, 96.3 KB of data
   transferred 96.3 KB in * seconds (*/sec) (glob)
   searching for changes
   no changes found
 #endif
 #if stream-bundle2
   $ hg clone --stream -U http://localhost:$HGPORT clone1
   streaming all changes
-  1027 files to transfer, 96.3 KB of data
-  transferred 96.3 KB in * seconds (* */sec) (glob)
+  1030 files to transfer, 96.4 KB of data
+  transferred 96.4 KB in * seconds (* */sec) (glob)
+
+  $ ls -1 clone1/.hg/cache
+  branch2-served
+  rbc-names-v1
+  rbc-revs-v1
 #endif
 
 --uncompressed is an alias to --stream
 
 #if stream-legacy
   $ hg clone --uncompressed -U http://localhost:$HGPORT clone1-uncompressed
   streaming all changes
   1027 files to transfer, 96.3 KB of data
   transferred 96.3 KB in * seconds (*/sec) (glob)
   searching for changes
   no changes found
 #endif
 #if stream-bundle2
   $ hg clone --uncompressed -U http://localhost:$HGPORT clone1-uncompressed
   streaming all changes
-  1027 files to transfer, 96.3 KB of data
-  transferred 96.3 KB in * seconds (* */sec) (glob)
+  1030 files to transfer, 96.4 KB of data
+  transferred 96.4 KB in * seconds (* */sec) (glob)
 #endif
 
 Clone with background file closing enabled
 
 #if stream-legacy
   $ hg --debug --config worker.backgroundclose=true --config worker.backgroundcloseminfilecount=1 clone --stream -U http://localhost:$HGPORT clone-background | grep -v adding
   using http://localhost:$HGPORT/
   sending capabilities command
   sending branchmap command
   streaming all changes
   sending stream_out command
   1027 files to transfer, 96.3 KB of data
   starting 4 threads for background file closing
   transferred 96.3 KB in * seconds (*/sec) (glob)
   query 1; heads
   sending batch command
   searching for changes
   all remote heads known locally
   no changes found
   sending getbundle command
   bundle2-input-bundle: with-transaction
   bundle2-input-part: "listkeys" (params: 1 mandatory) supported
   bundle2-input-part: "phase-heads" supported
   bundle2-input-part: total payload size 24
   bundle2-input-bundle: 1 parts total
   checking for updated bookmarks
 #endif
 #if stream-bundle2
   $ hg --debug --config worker.backgroundclose=true --config worker.backgroundcloseminfilecount=1 clone --stream -U http://localhost:$HGPORT clone-background | grep -v adding
   using http://localhost:$HGPORT/
   sending capabilities command
   query 1; heads
   sending batch command
   streaming all changes
   sending getbundle command
   bundle2-input-bundle: with-transaction
   bundle2-input-part: "stream" (params: 4 mandatory) supported
   applying stream bundle
-  1027 files to transfer, 96.3 KB of data
+  1030 files to transfer, 96.4 KB of data
+  starting 4 threads for background file closing
   starting 4 threads for background file closing
-  transferred 96.3 KB in * seconds (* */sec) (glob)
-  bundle2-input-part: total payload size 110887
+  transferred 96.4 KB in * seconds (* */sec) (glob)
+  bundle2-input-part: total payload size 112077
   bundle2-input-part: "listkeys" (params: 1 mandatory) supported
   bundle2-input-bundle: 1 parts total
   checking for updated bookmarks
 #endif
 
 Cannot stream clone when there are secret changesets
 
   $ hg -R server phase --force --secret -r tip
   $ hg clone --stream -U http://localhost:$HGPORT secret-denied
   warning: stream clone requested but server has them disabled
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
   new changesets 96ee1d7354c4
 
   $ killdaemons.py
 
 Streaming of secrets can be overridden by server config
 
   $ cd server
   $ hg serve --config server.uncompressedallowsecret=true -p $HGPORT -d --pid-file=hg.pid
   $ cat hg.pid > $DAEMON_PIDS
   $ cd ..
 
 #if stream-legacy
   $ hg clone --stream -U http://localhost:$HGPORT secret-allowed
   streaming all changes
   1027 files to transfer, 96.3 KB of data
   transferred 96.3 KB in * seconds (*/sec) (glob)
   searching for changes
   no changes found
 #endif
 #if stream-bundle2
   $ hg clone --stream -U http://localhost:$HGPORT secret-allowed
   streaming all changes
-  1027 files to transfer, 96.3 KB of data
-  transferred 96.3 KB in * seconds (* */sec) (glob)
+  1030 files to transfer, 96.4 KB of data
+  transferred 96.4 KB in * seconds (* */sec) (glob)
 #endif
 
   $ killdaemons.py
 
 Verify interaction between preferuncompressed and secret presence
 
   $ cd server
   $ hg serve --config server.preferuncompressed=true -p $HGPORT -d --pid-file=hg.pid
   $ cat hg.pid > $DAEMON_PIDS
   $ cd ..
 
   $ hg clone -U http://localhost:$HGPORT preferuncompressed-secret
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
   new changesets 96ee1d7354c4
 
   $ killdaemons.py
 
 Clone not allowed when full bundles disabled and can't serve secrets
 
   $ cd server
   $ hg serve --config server.disablefullbundle=true -p $HGPORT -d --pid-file=hg.pid
   $ cat hg.pid > $DAEMON_PIDS
   $ cd ..
 
   $ hg clone --stream http://localhost:$HGPORT secret-full-disabled
   warning: stream clone requested but server has them disabled
   requesting all changes
   remote: abort: server has pull-based clones disabled
   abort: pull failed on remote
   (remove --pull if specified or upgrade Mercurial)
   [255]
 
 Local stream clone with secrets involved
 (This is just a test over behavior: if you have access to the repo's files,
 there is no security so it isn't important to prevent a clone here.)
 
   $ hg clone -U --stream server local-secret
   warning: stream clone requested but server has them disabled
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 1 changesets with 1 changes to 1 files
   new changesets 96ee1d7354c4
 
 Stream clone while repo is changing:
 
   $ mkdir changing
   $ cd changing
 
 extension for delaying the server process so we reliably can modify the repo
 while cloning
 
   $ cat > delayer.py <<EOF
   > import time
   > from mercurial import extensions, vfs
   > def __call__(orig, self, path, *args, **kwargs):
   >     if path == 'data/f1.i':
   >         time.sleep(2)
   >     return orig(self, path, *args, **kwargs)
   > extensions.wrapfunction(vfs.vfs, '__call__', __call__)
   > EOF
 
 prepare repo with small and big file to cover both code paths in emitrevlogdata
 
   $ hg init repo
   $ touch repo/f1
   $ $TESTDIR/seq.py 50000 > repo/f2
   $ hg -R repo ci -Aqm "0"
   $ hg serve -R repo -p $HGPORT1 -d --pid-file=hg.pid --config extensions.delayer=delayer.py
   $ cat hg.pid >> $DAEMON_PIDS
 
 clone while modifying the repo between stating file with write lock and
 actually serving file content
 
   $ hg clone -q --stream -U http://localhost:$HGPORT1 clone &
   $ sleep 1
   $ echo >> repo/f1
   $ echo >> repo/f2
   $ hg -R repo ci -m "1"
   $ wait
   $ hg -R clone id
   000000000000
   $ cd ..
 
 Stream repository with bookmarks
 --------------------------------
 
 (revert introduction of secret changeset)
 
   $ hg -R server phase --draft 'secret()'
 
 add a bookmark
 
   $ hg -R server bookmark -r tip some-bookmark
 
 clone it
 
 #if stream-legacy
   $ hg clone --stream http://localhost:$HGPORT with-bookmarks
   streaming all changes
   1027 files to transfer, 96.3 KB of data
   transferred 96.3 KB in * seconds (*) (glob)
   searching for changes
   no changes found
   updating to branch default
   1025 files updated, 0 files merged, 0 files removed, 0 files unresolved
 #endif
 #if stream-bundle2
   $ hg clone --stream http://localhost:$HGPORT with-bookmarks
   streaming all changes
-  1027 files to transfer, 96.3 KB of data
-  transferred 96.3 KB in * seconds (* */sec) (glob)
+  1033 files to transfer, 96.6 KB of data
+  transferred 96.6 KB in * seconds (* */sec) (glob)
   updating to branch default
   1025 files updated, 0 files merged, 0 files removed, 0 files unresolved
 #endif
   $ hg -R with-bookmarks bookmarks
      some-bookmark             1:c17445101a72
 
 Stream repository with phases
 -----------------------------
 
 Clone as publishing
 
   $ hg -R server phase -r 'all()'
   0: draft
   1: draft
 
 #if stream-legacy
   $ hg clone --stream http://localhost:$HGPORT phase-publish
   streaming all changes
   1027 files to transfer, 96.3 KB of data
   transferred 96.3 KB in * seconds (*) (glob)
   searching for changes
   no changes found
   updating to branch default
   1025 files updated, 0 files merged, 0 files removed, 0 files unresolved
 #endif
 #if stream-bundle2
   $ hg clone --stream http://localhost:$HGPORT phase-publish
   streaming all changes
-  1027 files to transfer, 96.3 KB of data
-  transferred 96.3 KB in * seconds (* */sec) (glob)
+  1033 files to transfer, 96.6 KB of data
+  transferred 96.6 KB in * seconds (* */sec) (glob)
   updating to branch default
   1025 files updated, 0 files merged, 0 files removed, 0 files unresolved
 #endif
   $ hg -R phase-publish phase -r 'all()'
   0: public
   1: public
 
 Clone as non publishing
 
   $ cat << EOF >> server/.hg/hgrc
   > [phases]
   > publish = False
   > EOF
   $ killdaemons.py
   $ hg -R server serve -p $HGPORT -d --pid-file=hg.pid
   $ cat hg.pid >> $DAEMON_PIDS
 
 #if stream-legacy
   $ hg clone --stream http://localhost:$HGPORT phase-no-publish
   streaming all changes
   1027 files to transfer, 96.3 KB of data
   transferred 96.3 KB in * seconds (*) (glob)
   searching for changes
   no changes found
   updating to branch default
   1025 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R phase-no-publish phase -r 'all()'
   0: public
   1: public
 #endif
 #if stream-bundle2
   $ hg clone --stream http://localhost:$HGPORT phase-no-publish
   streaming all changes
-  1028 files to transfer, 96.4 KB of data
-  transferred 96.4 KB in * seconds (* */sec) (glob)
+  1034 files to transfer, 96.7 KB of data
+  transferred 96.7 KB in * seconds (* */sec) (glob)
   updating to branch default
   1025 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R phase-no-publish phase -r 'all()'
   0: draft
   1: draft
 #endif
 
   $ killdaemons.py