branching: merge default into stable for 6.8rc0

contrib/all-revsets.txt

@@ -1,159 +1,159 b''
 # All revsets ever used with revsetbenchmarks.py script
 #
 # The goal of this file is to gather all revsets ever used for benchmarking
 # revset's performance. It should be used to gather revsets that test a
 # specific usecase or a specific implementation of revset predicates.
 # If you are working on the smartset implementation itself, check
 # 'base-revsets.txt'.
 #
 # Please update this file with any revsets you use for benchmarking a change so
 # that future contributors can easily find and retest it when doing further
 # modification. Feel free to highlight interesting variants if needed.
 
 
 ## Revset from this section are all extracted from changelog when this file was
 # created. Feel free to dig and improve documentation.
 
 # Used in revision da05fe01170b
 (20000::) - (20000)
 # Used in revision 95af98616aa7
 parents(20000)
 # Used in revision 186fd06283b4
 (_intlist('20000\x0020001')) and merge()
 # Used in revision 911f5a6579d1
 p1(20000)
 p2(10000)
 # Used in revision b6dc3b79bb25
 0::
 # Used in revision faf4f63533ff
 bookmark()
 # Used in revision 22ba2c0825da
 tip~25
 # Used in revision 0cf46b8298fe
 bisect(range)
 # Used in revision 5b65429721d5
 divergent()
 # Used in revision 6261b9c549a2
 file(COPYING)
 # Used in revision 44f471102f3a
 follow(COPYING)
 # Used in revision 8040a44aab1c
 origin(tip)
 # Used in revision bbf4f3dfd700
 rev(25)
 # Used in revision a428db9ab61d
 p1()
 # Used in revision c1546d7400ef
 min(0::)
 # Used in revision 546fa6576815
-author(lmoscovicz) or author(olivia)
-author(olivia) or author(lmoscovicz)
+author(lmoscovicz) or author("pierre-yves")
+author("pierre-yves") or author(lmoscovicz)
 # Used in revision 9bfe68357c01
 public() and id("d82e2223f132")
 # Used in revision ba89f7b542c9
 rev(25)
 # Used in revision eb763217152a
 rev(210000)
 # Used in revision 69524a05a7fa
 10:100
 parents(10):parents(100)
 # Used in revision 6f1b8b3f12fd
 100~5
 parents(100)~5
 (100~5)~5
 # Used in revision 7a42e5d4c418
 children(tip~100)
 # Used in revision 7e8737e6ab08
 100^1
 parents(100)^1
 (100^1)^1
 # Used in revision 30e0dcd7c5ff
 matching(100)
 matching(parents(100))
 # Used in revision aafeaba22826
 0|1|2|3|4|5|6|7|8|9
 # Used in revision 33c7a94d4dd0
 tip:0
 # Used in revision 7d369fae098e
 (0:100000)
 # Used in revision b333ca94403d
 0 + 1 + 2 + ... + 200
 0 + 1 + 2 + ... + 1000
 sort(0 + 1 + 2 + ... + 200)
 sort(0 + 1 + 2 + ... + 1000)
 # Used in revision 7fbef7932af9
 first(0 + 1 + 2 + ... + 1000)
 # Used in revision ceaf04bb14ff
 0:1000
 # Used in revision 262e6ad93885
 not public()
 (tip~1000::) - public()
 not public() and branch("default")
 # Used in revision 15412bba5a68
 0::tip
 
 ## all the revsets from this section have been taken from the former central file
 # for revset's benchmarking, they are undocumented for this reason.
 all()
 draft()
 ::tip
 draft() and ::tip
 ::tip and draft()
 author(lmoscovicz)
-author(olivia)
+author("pierre-yves")
 ::p1(p1(tip))::
 public()
 :10000 and public()
 :10000 and draft()
 (not public() - obsolete())
 
 # The one below is used by rebase
 (children(ancestor(tip~5, tip)) and ::(tip~5))::
 
 # those two `roots(...)` inputs are close to what phase movement use.
 roots((tip~100::) - (tip~100::tip))
 roots((0::) - (0::tip))
 
 # more roots testing
 roots(tip~100:)
 roots(:42)
 roots(not public())
 roots((0:tip)::)
 roots(0::tip)
 42:68 and roots(42:tip)
 # Used in revision f140d6207cca
 roots(0:tip)
 # test disjoint set with multiple roots
 roots((:42) + (tip~42:))
 
 # Testing the behavior of "head()" in various situations
 head()
 head() - public()
 draft() and head()
-head() and author("olivia")
+head() and author("pierre-yves")
 
 # testing the mutable phases set
 draft()
 secret()
 
 # test finding common ancestors
 heads(commonancestors(last(head(), 2)))
 heads(commonancestors(head()))
 
 # more heads testing
 heads(all())
 heads(-10000:-1)
 (-5000:-1000) and heads(-10000:-1)
 heads(matching(tip, "author"))
 heads(matching(tip, "author")) and -10000:-1
 (-10000:-1) and heads(matching(tip, "author"))
 # more roots testing
 roots(all())
 roots(-10000:-1)
 (-5000:-1000) and roots(-10000:-1)
 roots(matching(tip, "author"))
 roots(matching(tip, "author")) and -10000:-1
 (-10000:-1) and roots(matching(tip, "author"))
 only(max(head()))
 only(max(head()), min(head()))
 only(max(head()), limit(head(), 1, 1))

contrib/base-revsets.txt

@@ -1,52 +1,52 b''
 # Base Revsets to be used with revsetbenchmarks.py script
 #
 # The goal of this file is to gather a limited amount of revsets that allow a
 # good coverage of the internal revsets mechanisms.  Revsets included should not
 # be selected for their individual implementation, but for what they reveal of
 # the internal implementation of smartsets classes (and their interactions).
 #
 # Use and update this file when you change internal implementation of these
 # smartsets classes. Please include a comment explaining what each of your
 # addition is testing. Also check if your changes to the smartset class makes
 # some of the tests inadequate and replace them with a new one testing the same
 # behavior.
 #
 # If you want to benchmark revsets predicate itself, check 'all-revsets.txt'.
 #
 # The current content of this file is currently likely not reaching this goal
 # entirely, feel free, to audit its content and comment on each revset to
 # highlight what internal mechanisms they test.
 
 all()
 draft()
 ::tip
 draft() and ::tip
 ::tip and draft()
 0::tip
 roots(0::tip)
 author(lmoscovicz)
-author(olivia)
-author(lmoscovicz) or author(olivia)
-author(olivia) or author(lmoscovicz)
+author("pierre-yves")
+author(lmoscovicz) or author("pierre-yves")
+author("pierre-yves") or author(lmoscovicz)
 tip:0
 0::
 # those two `roots(...)` inputs are close to what phase movement use.
 roots((tip~100::) - (tip~100::tip))
 roots((0::) - (0::tip))
 42:68 and roots(42:tip)
 ::p1(p1(tip))::
 public()
 :10000 and public()
 draft()
 :10000 and draft()
 roots((0:tip)::)
 (not public() - obsolete())
 (_intlist('20000\x0020001')) and merge()
 parents(20000)
 (20000::) - (20000)
 # The one below is used by rebase
 (children(ancestor(tip~5, tip)) and ::(tip~5))::
 heads(commonancestors(last(head(), 2)))
 heads(-10000:-1)
 roots(-10000:-1)
 only(max(head()), min(head()))

contrib/perf.py

@@ -1,4651 +1,4725 b''
 # perf.py - performance test routines
 '''helper extension to measure performance
 
 Configurations
 ==============
 
 ``perf``
 --------
 
 ``all-timing``
     When set, additional statistics will be reported for each benchmark: best,
     worst, median average. If not set only the best timing is reported
     (default: off).
 
 ``presleep``
   number of second to wait before any group of runs (default: 1)
 
 ``pre-run``
   number of run to perform before starting measurement.
 
 ``profile-benchmark``
   Enable profiling for the benchmarked section.
-  (The first iteration is benchmarked)
+  (by default, the first iteration is benchmarked)
+
+``profiled-runs``
+  list of iteration to profile (starting from 0)
 
 ``run-limits``
   Control the number of runs each benchmark will perform. The option value
   should be a list of `<time>-<numberofrun>` pairs. After each run the
   conditions are considered in order with the following logic:
 
       If benchmark has been running for <time> seconds, and we have performed
       <numberofrun> iterations, stop the benchmark,
 
   The default value is: `3.0-100, 10.0-3`
 
 ``stub``
     When set, benchmarks will only be run once, useful for testing
     (default: off)
 '''
 
 # "historical portability" policy of perf.py:
 #
 # We have to do:
 # - make perf.py "loadable" with as wide Mercurial version as possible
 #   This doesn't mean that perf commands work correctly with that Mercurial.
 #   BTW, perf.py itself has been available since 1.1 (or eb240755386d).
 # - make historical perf command work correctly with as wide Mercurial
 #   version as possible
 #
 # We have to do, if possible with reasonable cost:
 # - make recent perf command for historical feature work correctly
 #   with early Mercurial
 #
 # We don't have to do:
 # - make perf command for recent feature work correctly with early
 #   Mercurial
 
 import contextlib
 import functools
 import gc
 import os
 import random
 import shutil
 import struct
 import sys
 import tempfile
 import threading
 import time
 
 import mercurial.revlog
 from mercurial import (
     changegroup,
     cmdutil,
     commands,
     copies,
     error,
     extensions,
     hg,
     mdiff,
     merge,
     util,
 )
 
 # for "historical portability":
 # try to import modules separately (in dict order), and ignore
 # failure, because these aren't available with early Mercurial
 try:
     from mercurial import branchmap  # since 2.5 (or bcee63733aad)
 except ImportError:
     pass
 try:
     from mercurial import obsolete  # since 2.3 (or ad0d6c2b3279)
 except ImportError:
     pass
 try:
     from mercurial import registrar  # since 3.7 (or 37d50250b696)
 
     dir(registrar)  # forcibly load it
 except ImportError:
     registrar = None
 try:
     from mercurial import repoview  # since 2.5 (or 3a6ddacb7198)
 except ImportError:
     pass
 try:
     from mercurial.utils import repoviewutil  # since 5.0
 except ImportError:
     repoviewutil = None
 try:
     from mercurial import scmutil  # since 1.9 (or 8b252e826c68)
 except ImportError:
     pass
 try:
     from mercurial import setdiscovery  # since 1.9 (or cb98fed52495)
 except ImportError:
     pass
 
 try:
     from mercurial import profiling
 except ImportError:
     profiling = None
 
 try:
     from mercurial.revlogutils import constants as revlog_constants
 
     perf_rl_kind = (revlog_constants.KIND_OTHER, b'created-by-perf')
 
     def revlog(opener, *args, **kwargs):
         return mercurial.revlog.revlog(opener, perf_rl_kind, *args, **kwargs)
 
 
 except (ImportError, AttributeError):
     perf_rl_kind = None
 
     def revlog(opener, *args, **kwargs):
         return mercurial.revlog.revlog(opener, *args, **kwargs)
 
 
 def identity(a):
     return a
 
 
 try:
     from mercurial import pycompat
 
     getargspec = pycompat.getargspec  # added to module after 4.5
     _byteskwargs = pycompat.byteskwargs  # since 4.1 (or fbc3f73dc802)
     _sysstr = pycompat.sysstr  # since 4.0 (or 2219f4f82ede)
     _bytestr = pycompat.bytestr  # since 4.2 (or b70407bd84d5)
     _xrange = pycompat.xrange  # since 4.8 (or 7eba8f83129b)
     fsencode = pycompat.fsencode  # since 3.9 (or f4a5e0e86a7e)
     if pycompat.ispy3:
         _maxint = sys.maxsize  # per py3 docs for replacing maxint
     else:
         _maxint = sys.maxint
 except (NameError, ImportError, AttributeError):
     import inspect
 
     getargspec = inspect.getargspec
     _byteskwargs = identity
     _bytestr = str
     fsencode = identity  # no py3 support
     _maxint = sys.maxint  # no py3 support
     _sysstr = lambda x: x  # no py3 support
     _xrange = xrange
 
 try:
     # 4.7+
     queue = pycompat.queue.Queue
 except (NameError, AttributeError, ImportError):
     # <4.7.
     try:
         queue = pycompat.queue
     except (NameError, AttributeError, ImportError):
         import Queue as queue
 
 try:
     from mercurial import logcmdutil
 
     makelogtemplater = logcmdutil.maketemplater
 except (AttributeError, ImportError):
     try:
         makelogtemplater = cmdutil.makelogtemplater
     except (AttributeError, ImportError):
         makelogtemplater = None
 
 # for "historical portability":
 # define util.safehasattr forcibly, because util.safehasattr has been
 # available since 1.9.3 (or 94b200a11cf7)
 _undefined = object()
 
 
 def safehasattr(thing, attr):
     return getattr(thing, _sysstr(attr), _undefined) is not _undefined
 
 
 setattr(util, 'safehasattr', safehasattr)
 
 # for "historical portability":
 # define util.timer forcibly, because util.timer has been available
 # since ae5d60bb70c9
 if safehasattr(time, 'perf_counter'):
     util.timer = time.perf_counter
 elif os.name == b'nt':
     util.timer = time.clock
 else:
     util.timer = time.time
 
 # for "historical portability":
 # use locally defined empty option list, if formatteropts isn't
 # available, because commands.formatteropts has been available since
 # 3.2 (or 7a7eed5176a4), even though formatting itself has been
 # available since 2.2 (or ae5f92e154d3)
 formatteropts = getattr(
     cmdutil, "formatteropts", getattr(commands, "formatteropts", [])
 )
 
 # for "historical portability":
 # use locally defined option list, if debugrevlogopts isn't available,
 # because commands.debugrevlogopts has been available since 3.7 (or
 # 5606f7d0d063), even though cmdutil.openrevlog() has been available
 # since 1.9 (or a79fea6b3e77).
 revlogopts = getattr(
     cmdutil,
     "debugrevlogopts",
     getattr(
         commands,
         "debugrevlogopts",
         [
             (b'c', b'changelog', False, b'open changelog'),
             (b'm', b'manifest', False, b'open manifest'),
             (b'', b'dir', False, b'open directory manifest'),
         ],
     ),
 )
 
 cmdtable = {}
 
 
 # for "historical portability":
 # define parsealiases locally, because cmdutil.parsealiases has been
 # available since 1.5 (or 6252852b4332)
 def parsealiases(cmd):
     return cmd.split(b"|")
 
 
 if safehasattr(registrar, 'command'):
     command = registrar.command(cmdtable)
 elif safehasattr(cmdutil, 'command'):
     command = cmdutil.command(cmdtable)
     if 'norepo' not in getargspec(command).args:
         # for "historical portability":
         # wrap original cmdutil.command, because "norepo" option has
         # been available since 3.1 (or 75a96326cecb)
         _command = command
 
         def command(name, options=(), synopsis=None, norepo=False):
             if norepo:
                 commands.norepo += b' %s' % b' '.join(parsealiases(name))
             return _command(name, list(options), synopsis)
 
 
 else:
     # for "historical portability":
     # define "@command" annotation locally, because cmdutil.command
     # has been available since 1.9 (or 2daa5179e73f)
     def command(name, options=(), synopsis=None, norepo=False):
         def decorator(func):
             if synopsis:
                 cmdtable[name] = func, list(options), synopsis
             else:
                 cmdtable[name] = func, list(options)
             if norepo:
                 commands.norepo += b' %s' % b' '.join(parsealiases(name))
             return func
 
         return decorator
 
 
 try:
     import mercurial.registrar
     import mercurial.configitems
 
     configtable = {}
     configitem = mercurial.registrar.configitem(configtable)
     configitem(
         b'perf',
         b'presleep',
         default=mercurial.configitems.dynamicdefault,
         experimental=True,
     )
     configitem(
         b'perf',
         b'stub',
         default=mercurial.configitems.dynamicdefault,
         experimental=True,
     )
     configitem(
         b'perf',
         b'parentscount',
         default=mercurial.configitems.dynamicdefault,
         experimental=True,
     )
     configitem(
         b'perf',
         b'all-timing',
         default=mercurial.configitems.dynamicdefault,
         experimental=True,
     )
     configitem(
         b'perf',
         b'pre-run',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
         b'profile-benchmark',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
+        b'profiled-runs',
+        default=mercurial.configitems.dynamicdefault,
+    )
+    configitem(
+        b'perf',
         b'run-limits',
         default=mercurial.configitems.dynamicdefault,
         experimental=True,
     )
 except (ImportError, AttributeError):
     pass
 except TypeError:
     # compatibility fix for a11fd395e83f
     # hg version: 5.2
     configitem(
         b'perf',
         b'presleep',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
         b'stub',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
         b'parentscount',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
         b'all-timing',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
         b'pre-run',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
-        b'profile-benchmark',
+        b'profiled-runs',
         default=mercurial.configitems.dynamicdefault,
     )
     configitem(
         b'perf',
         b'run-limits',
         default=mercurial.configitems.dynamicdefault,
     )
 
 
 def getlen(ui):
     if ui.configbool(b"perf", b"stub", False):
         return lambda x: 1
     return len
 
 
 class noop:
     """dummy context manager"""
 
     def __enter__(self):
         pass
 
     def __exit__(self, *args):
         pass
 
 
 NOOPCTX = noop()
 
 
 def gettimer(ui, opts=None):
     """return a timer function and formatter: (timer, formatter)
 
     This function exists to gather the creation of formatter in a single
     place instead of duplicating it in all performance commands."""
 
     # enforce an idle period before execution to counteract power management
     # experimental config: perf.presleep
     time.sleep(getint(ui, b"perf", b"presleep", 1))
 
     if opts is None:
         opts = {}
     # redirect all to stderr unless buffer api is in use
     if not ui._buffers:
         ui = ui.copy()
         uifout = safeattrsetter(ui, b'fout', ignoremissing=True)
         if uifout:
             # for "historical portability":
             # ui.fout/ferr have been available since 1.9 (or 4e1ccd4c2b6d)
             uifout.set(ui.ferr)
 
     # get a formatter
     uiformatter = getattr(ui, 'formatter', None)
     if uiformatter:
         fm = uiformatter(b'perf', opts)
     else:
         # for "historical portability":
         # define formatter locally, because ui.formatter has been
         # available since 2.2 (or ae5f92e154d3)
         from mercurial import node
 
         class defaultformatter:
             """Minimized composition of baseformatter and plainformatter"""
 
             def __init__(self, ui, topic, opts):
                 self._ui = ui
                 if ui.debugflag:
                     self.hexfunc = node.hex
                 else:
                     self.hexfunc = node.short
 
             def __nonzero__(self):
                 return False
 
             __bool__ = __nonzero__
 
             def startitem(self):
                 pass
 
             def data(self, **data):
                 pass
 
             def write(self, fields, deftext, *fielddata, **opts):
                 self._ui.write(deftext % fielddata, **opts)
 
             def condwrite(self, cond, fields, deftext, *fielddata, **opts):
                 if cond:
                     self._ui.write(deftext % fielddata, **opts)
 
             def plain(self, text, **opts):
                 self._ui.write(text, **opts)
 
             def end(self):
                 pass
 
         fm = defaultformatter(ui, b'perf', opts)
 
     # stub function, runs code only once instead of in a loop
     # experimental config: perf.stub
     if ui.configbool(b"perf", b"stub", False):
         return functools.partial(stub_timer, fm), fm
 
     # experimental config: perf.all-timing
     displayall = ui.configbool(b"perf", b"all-timing", True)
 
     # experimental config: perf.run-limits
     limitspec = ui.configlist(b"perf", b"run-limits", [])
     limits = []
     for item in limitspec:
         parts = item.split(b'-', 1)
         if len(parts) < 2:
             ui.warn((b'malformatted run limit entry, missing "-": %s\n' % item))
             continue
         try:
             time_limit = float(_sysstr(parts[0]))
         except ValueError as e:
             ui.warn(
                 (
                     b'malformatted run limit entry, %s: %s\n'
                     % (_bytestr(e), item)
                 )
             )
             continue
         try:
             run_limit = int(_sysstr(parts[1]))
         except ValueError as e:
             ui.warn(
                 (
                     b'malformatted run limit entry, %s: %s\n'
                     % (_bytestr(e), item)
                 )
             )
             continue
         limits.append((time_limit, run_limit))
     if not limits:
         limits = DEFAULTLIMITS
 
     profiler = None
+    profiled_runs = set()
     if profiling is not None:
         if ui.configbool(b"perf", b"profile-benchmark", False):
-            profiler = profiling.profile(ui)
+            profiler = lambda: profiling.profile(ui)
+            for run in ui.configlist(b"perf", b"profiled-runs", [0]):
+                profiled_runs.add(int(run))
 
     prerun = getint(ui, b"perf", b"pre-run", 0)
     t = functools.partial(
         _timer,
         fm,
         displayall=displayall,
         limits=limits,
         prerun=prerun,
         profiler=profiler,
+        profiled_runs=profiled_runs,
     )
     return t, fm
 
 
 def stub_timer(fm, func, setup=None, title=None):
     if setup is not None:
         setup()
     func()
 
 
 @contextlib.contextmanager
 def timeone():
     r = []
     ostart = os.times()
     cstart = util.timer()
     yield r
     cstop = util.timer()
     ostop = os.times()
     a, b = ostart, ostop
     r.append((cstop - cstart, b[0] - a[0], b[1] - a[1]))
 
 
 # list of stop condition (elapsed time, minimal run count)
 DEFAULTLIMITS = (
     (3.0, 100),
     (10.0, 3),
 )
 
 
 @contextlib.contextmanager
 def noop_context():
     yield
 
 
 def _timer(
     fm,
     func,
     setup=None,
     context=noop_context,
     title=None,
     displayall=False,
     limits=DEFAULTLIMITS,
     prerun=0,
     profiler=None,
+    profiled_runs=(0,),
 ):
     gc.collect()
     results = []
-    begin = util.timer()
     count = 0
     if profiler is None:
-        profiler = NOOPCTX
+        profiler = lambda: NOOPCTX
     for i in range(prerun):
         if setup is not None:
             setup()
         with context():
             func()
+    begin = util.timer()
     keepgoing = True
     while keepgoing:
+        if count in profiled_runs:
+            prof = profiler()
+        else:
+            prof = NOOPCTX
         if setup is not None:
             setup()
         with context():
-            with profiler:
+            gc.collect()
+            with prof:
                 with timeone() as item:
                     r = func()
-        profiler = NOOPCTX
         count += 1
         results.append(item[0])
         cstop = util.timer()
         # Look for a stop condition.
         elapsed = cstop - begin
         for t, mincount in limits:
             if elapsed >= t and count >= mincount:
                 keepgoing = False
                 break
 
     formatone(fm, results, title=title, result=r, displayall=displayall)
 
 
 def formatone(fm, timings, title=None, result=None, displayall=False):
     count = len(timings)
 
     fm.startitem()
 
     if title:
         fm.write(b'title', b'! %s\n', title)
     if result:
         fm.write(b'result', b'! result: %s\n', result)
 
     def display(role, entry):
         prefix = b''
         if role != b'best':
             prefix = b'%s.' % role
         fm.plain(b'!')
         fm.write(prefix + b'wall', b' wall %f', entry[0])
         fm.write(prefix + b'comb', b' comb %f', entry[1] + entry[2])
         fm.write(prefix + b'user', b' user %f', entry[1])
         fm.write(prefix + b'sys', b' sys %f', entry[2])
         fm.write(prefix + b'count', b' (%s of %%d)' % role, count)
         fm.plain(b'\n')
 
     timings.sort()
     min_val = timings[0]
     display(b'best', min_val)
     if displayall:
         max_val = timings[-1]
         display(b'max', max_val)
         avg = tuple([sum(x) / count for x in zip(*timings)])
         display(b'avg', avg)
         median = timings[len(timings) // 2]
         display(b'median', median)
 
 
 # utilities for historical portability
 
 
 def getint(ui, section, name, default):
     # for "historical portability":
     # ui.configint has been available since 1.9 (or fa2b596db182)
     v = ui.config(section, name, None)
     if v is None:
         return default
     try:
         return int(v)
     except ValueError:
         raise error.ConfigError(
             b"%s.%s is not an integer ('%s')" % (section, name, v)
         )
 
 
 def safeattrsetter(obj, name, ignoremissing=False):
     """Ensure that 'obj' has 'name' attribute before subsequent setattr
 
     This function is aborted, if 'obj' doesn't have 'name' attribute
     at runtime. This avoids overlooking removal of an attribute, which
     breaks assumption of performance measurement, in the future.
 
     This function returns the object to (1) assign a new value, and
     (2) restore an original value to the attribute.
 
     If 'ignoremissing' is true, missing 'name' attribute doesn't cause
     abortion, and this function returns None. This is useful to
     examine an attribute, which isn't ensured in all Mercurial
     versions.
     """
     if not util.safehasattr(obj, name):
         if ignoremissing:
             return None
         raise error.Abort(
             (
                 b"missing attribute %s of %s might break assumption"
                 b" of performance measurement"
             )
             % (name, obj)
         )
 
     origvalue = getattr(obj, _sysstr(name))
 
     class attrutil:
         def set(self, newvalue):
             setattr(obj, _sysstr(name), newvalue)
 
         def restore(self):
             setattr(obj, _sysstr(name), origvalue)
 
     return attrutil()
 
 
 # utilities to examine each internal API changes
 
 
 def getbranchmapsubsettable():
     # for "historical portability":
     # subsettable is defined in:
     # - branchmap since 2.9 (or 175c6fd8cacc)
     # - repoview since 2.5 (or 59a9f18d4587)
     # - repoviewutil since 5.0
     for mod in (branchmap, repoview, repoviewutil):
         subsettable = getattr(mod, 'subsettable', None)
         if subsettable:
             return subsettable
 
     # bisecting in bcee63733aad::59a9f18d4587 can reach here (both
     # branchmap and repoview modules exist, but subsettable attribute
     # doesn't)
     raise error.Abort(
         b"perfbranchmap not available with this Mercurial",
         hint=b"use 2.5 or later",
     )
 
 
 def getsvfs(repo):
     """Return appropriate object to access files under .hg/store"""
     # for "historical portability":
     # repo.svfs has been available since 2.3 (or 7034365089bf)
     svfs = getattr(repo, 'svfs', None)
     if svfs:
         return svfs
     else:
         return getattr(repo, 'sopener')
 
 
 def getvfs(repo):
     """Return appropriate object to access files under .hg"""
     # for "historical portability":
     # repo.vfs has been available since 2.3 (or 7034365089bf)
     vfs = getattr(repo, 'vfs', None)
     if vfs:
         return vfs
     else:
         return getattr(repo, 'opener')
 
 
 def repocleartagscachefunc(repo):
     """Return the function to clear tags cache according to repo internal API"""
     if util.safehasattr(repo, b'_tagscache'):  # since 2.0 (or 9dca7653b525)
         # in this case, setattr(repo, '_tagscache', None) or so isn't
         # correct way to clear tags cache, because existing code paths
         # expect _tagscache to be a structured object.
         def clearcache():
             # _tagscache has been filteredpropertycache since 2.5 (or
             # 98c867ac1330), and delattr() can't work in such case
             if '_tagscache' in vars(repo):
                 del repo.__dict__['_tagscache']
 
         return clearcache
 
     repotags = safeattrsetter(repo, b'_tags', ignoremissing=True)
     if repotags:  # since 1.4 (or 5614a628d173)
         return lambda: repotags.set(None)
 
     repotagscache = safeattrsetter(repo, b'tagscache', ignoremissing=True)
     if repotagscache:  # since 0.6 (or d7df759d0e97)
         return lambda: repotagscache.set(None)
 
     # Mercurial earlier than 0.6 (or d7df759d0e97) logically reaches
     # this point, but it isn't so problematic, because:
     # - repo.tags of such Mercurial isn't "callable", and repo.tags()
     #   in perftags() causes failure soon
     # - perf.py itself has been available since 1.1 (or eb240755386d)
     raise error.Abort(b"tags API of this hg command is unknown")
 
 
 # utilities to clear cache
 
 
 def clearfilecache(obj, attrname):
     unfiltered = getattr(obj, 'unfiltered', None)
     if unfiltered is not None:
         obj = obj.unfiltered()
     if attrname in vars(obj):
         delattr(obj, attrname)
     obj._filecache.pop(attrname, None)
 
 
 def clearchangelog(repo):
     if repo is not repo.unfiltered():
         object.__setattr__(repo, '_clcachekey', None)
         object.__setattr__(repo, '_clcache', None)
     clearfilecache(repo.unfiltered(), 'changelog')
 
 
 # perf commands
 
 
 @command(b'perf::walk|perfwalk', formatteropts)
 def perfwalk(ui, repo, *pats, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     m = scmutil.match(repo[None], pats, {})
     timer(
         lambda: len(
             list(
                 repo.dirstate.walk(m, subrepos=[], unknown=True, ignored=False)
             )
         )
     )
     fm.end()
 
 
 @command(b'perf::annotate|perfannotate', formatteropts)
 def perfannotate(ui, repo, f, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     fc = repo[b'.'][f]
     timer(lambda: len(fc.annotate(True)))
     fm.end()
 
 
 @command(
     b'perf::status|perfstatus',
     [
         (b'u', b'unknown', False, b'ask status to look for unknown files'),
         (b'', b'dirstate', False, b'benchmark the internal dirstate call'),
     ]
     + formatteropts,
 )
 def perfstatus(ui, repo, **opts):
     """benchmark the performance of a single status call
 
     The repository data are preserved between each call.
 
     By default, only the status of the tracked file are requested. If
     `--unknown` is passed, the "unknown" files are also tracked.
     """
     opts = _byteskwargs(opts)
     # m = match.always(repo.root, repo.getcwd())
     # timer(lambda: sum(map(len, repo.dirstate.status(m, [], False, False,
     #                                                False))))
     timer, fm = gettimer(ui, opts)
     if opts[b'dirstate']:
         dirstate = repo.dirstate
         m = scmutil.matchall(repo)
         unknown = opts[b'unknown']
 
         def status_dirstate():
             s = dirstate.status(
                 m, subrepos=[], ignored=False, clean=False, unknown=unknown
             )
             sum(map(bool, s))
 
         if util.safehasattr(dirstate, 'running_status'):
             with dirstate.running_status(repo):
                 timer(status_dirstate)
                 dirstate.invalidate()
         else:
             timer(status_dirstate)
     else:
         timer(lambda: sum(map(len, repo.status(unknown=opts[b'unknown']))))
     fm.end()
 
 
 @command(b'perf::addremove|perfaddremove', formatteropts)
 def perfaddremove(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     try:
         oldquiet = repo.ui.quiet
         repo.ui.quiet = True
         matcher = scmutil.match(repo[None])
         opts[b'dry_run'] = True
         if 'uipathfn' in getargspec(scmutil.addremove).args:
             uipathfn = scmutil.getuipathfn(repo)
             timer(lambda: scmutil.addremove(repo, matcher, b"", uipathfn, opts))
         else:
             timer(lambda: scmutil.addremove(repo, matcher, b"", opts))
     finally:
         repo.ui.quiet = oldquiet
         fm.end()
 
 
 def clearcaches(cl):
     # behave somewhat consistently across internal API changes
     if util.safehasattr(cl, b'clearcaches'):
         cl.clearcaches()
     elif util.safehasattr(cl, b'_nodecache'):
         # <= hg-5.2
         from mercurial.node import nullid, nullrev
 
         cl._nodecache = {nullid: nullrev}
         cl._nodepos = None
 
 
 @command(b'perf::heads|perfheads', formatteropts)
 def perfheads(ui, repo, **opts):
     """benchmark the computation of a changelog heads"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     cl = repo.changelog
 
     def s():
         clearcaches(cl)
 
     def d():
         len(cl.headrevs())
 
     timer(d, setup=s)
     fm.end()
 
 
 def _default_clear_on_disk_tags_cache(repo):
     from mercurial import tags
 
     repo.cachevfs.tryunlink(tags._filename(repo))
 
 
 def _default_clear_on_disk_tags_fnodes_cache(repo):
     from mercurial import tags
 
     repo.cachevfs.tryunlink(tags._fnodescachefile)
 
 
 def _default_forget_fnodes(repo, revs):
     """function used by the perf extension to prune some entries from the
     fnodes cache"""
     from mercurial import tags
 
     missing_1 = b'\xff' * 4
     missing_2 = b'\xff' * 20
     cache = tags.hgtagsfnodescache(repo.unfiltered())
     for r in revs:
         cache._writeentry(r * tags._fnodesrecsize, missing_1, missing_2)
     cache.write()
 
 
 @command(
     b'perf::tags|perftags',
     formatteropts
     + [
         (b'', b'clear-revlogs', False, b'refresh changelog and manifest'),
         (
             b'',
             b'clear-on-disk-cache',
             False,
             b'clear on disk tags cache (DESTRUCTIVE)',
         ),
         (
             b'',
             b'clear-fnode-cache-all',
             False,
             b'clear on disk file node cache (DESTRUCTIVE),',
         ),
         (
             b'',
             b'clear-fnode-cache-rev',
             [],
             b'clear on disk file node cache (DESTRUCTIVE),',
             b'REVS',
         ),
         (
             b'',
             b'update-last',
             b'',
             b'simulate an update over the last N revisions (DESTRUCTIVE),',
             b'N',
         ),
     ],
 )
 def perftags(ui, repo, **opts):
     """Benchmark tags retrieval in various situation
 
     The option marked as (DESTRUCTIVE) will alter the on-disk cache, possibly
     altering performance after the command was run. However, it does not
     destroy any stored data.
     """
     from mercurial import tags
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     repocleartagscache = repocleartagscachefunc(repo)
     clearrevlogs = opts[b'clear_revlogs']
     clear_disk = opts[b'clear_on_disk_cache']
     clear_fnode = opts[b'clear_fnode_cache_all']
 
     clear_fnode_revs = opts[b'clear_fnode_cache_rev']
     update_last_str = opts[b'update_last']
     update_last = None
     if update_last_str:
         try:
             update_last = int(update_last_str)
         except ValueError:
             msg = b'could not parse value for update-last: "%s"'
             msg %= update_last_str
             hint = b'value should be an integer'
             raise error.Abort(msg, hint=hint)
 
     clear_disk_fn = getattr(
         tags,
         "clear_cache_on_disk",
         _default_clear_on_disk_tags_cache,
     )
     if getattr(tags, 'clear_cache_fnodes_is_working', False):
         clear_fnodes_fn = tags.clear_cache_fnodes
     else:
         clear_fnodes_fn = _default_clear_on_disk_tags_fnodes_cache
     clear_fnodes_rev_fn = getattr(
         tags,
         "forget_fnodes",
         _default_forget_fnodes,
     )
 
     clear_revs = []
     if clear_fnode_revs:
         clear_revs.extend(scmutil.revrange(repo, clear_fnode_revs))
 
     if update_last:
         revset = b'last(all(), %d)' % update_last
         last_revs = repo.unfiltered().revs(revset)
         clear_revs.extend(last_revs)
 
         from mercurial import repoview
 
         rev_filter = {(b'experimental', b'extra-filter-revs'): revset}
         with repo.ui.configoverride(rev_filter, source=b"perf"):
             filter_id = repoview.extrafilter(repo.ui)
 
         filter_name = b'%s%%%s' % (repo.filtername, filter_id)
         pre_repo = repo.filtered(filter_name)
         pre_repo.tags()  # warm the cache
         old_tags_path = repo.cachevfs.join(tags._filename(pre_repo))
         new_tags_path = repo.cachevfs.join(tags._filename(repo))
 
     clear_revs = sorted(set(clear_revs))
 
     def s():
         if update_last:
             util.copyfile(old_tags_path, new_tags_path)
         if clearrevlogs:
             clearchangelog(repo)
             clearfilecache(repo.unfiltered(), 'manifest')
         if clear_disk:
             clear_disk_fn(repo)
         if clear_fnode:
             clear_fnodes_fn(repo)
         elif clear_revs:
             clear_fnodes_rev_fn(repo, clear_revs)
         repocleartagscache()
 
     def t():
         len(repo.tags())
 
     timer(t, setup=s)
     fm.end()
 
 
 @command(b'perf::ancestors|perfancestors', formatteropts)
 def perfancestors(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     heads = repo.changelog.headrevs()
 
     def d():
         for a in repo.changelog.ancestors(heads):
             pass
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::ancestorset|perfancestorset', formatteropts)
 def perfancestorset(ui, repo, revset, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     revs = repo.revs(revset)
     heads = repo.changelog.headrevs()
 
     def d():
         s = repo.changelog.ancestors(heads)
         for rev in revs:
             rev in s
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::delta-find',
     revlogopts + formatteropts,
     b'-c|-m|FILE REV',
 )
 def perf_delta_find(ui, repo, arg_1, arg_2=None, **opts):
     """benchmark the process of finding a valid delta for a revlog revision
 
     When a revlog receives a new revision (e.g. from a commit, or from an
     incoming bundle), it searches for a suitable delta-base to produce a delta.
     This perf command measures how much time we spend in this process. It
     operates on an already stored revision.
 
     See `hg help debug-delta-find` for another related command.
     """
     from mercurial import revlogutils
     import mercurial.revlogutils.deltas as deltautil
 
     opts = _byteskwargs(opts)
     if arg_2 is None:
         file_ = None
         rev = arg_1
     else:
         file_ = arg_1
         rev = arg_2
 
     repo = repo.unfiltered()
 
     timer, fm = gettimer(ui, opts)
 
     rev = int(rev)
 
     revlog = cmdutil.openrevlog(repo, b'perf::delta-find', file_, opts)
 
     deltacomputer = deltautil.deltacomputer(revlog)
 
     node = revlog.node(rev)
     p1r, p2r = revlog.parentrevs(rev)
     p1 = revlog.node(p1r)
     p2 = revlog.node(p2r)
     full_text = revlog.revision(rev)
     textlen = len(full_text)
     cachedelta = None
     flags = revlog.flags(rev)
 
     revinfo = revlogutils.revisioninfo(
         node,
         p1,
         p2,
         [full_text],  # btext
         textlen,
         cachedelta,
         flags,
     )
 
     # Note: we should probably purge the potential caches (like the full
     # manifest cache) between runs.
     def find_one():
         with revlog._datafp() as fh:
             deltacomputer.finddeltainfo(revinfo, fh, target_rev=rev)
 
     timer(find_one)
     fm.end()
 
 
 @command(b'perf::discovery|perfdiscovery', formatteropts, b'PATH')
 def perfdiscovery(ui, repo, path, **opts):
     """benchmark discovery between local repo and the peer at given path"""
     repos = [repo, None]
     timer, fm = gettimer(ui, opts)
 
     try:
         from mercurial.utils.urlutil import get_unique_pull_path_obj
 
         path = get_unique_pull_path_obj(b'perfdiscovery', ui, path)
     except ImportError:
         try:
             from mercurial.utils.urlutil import get_unique_pull_path
 
             path = get_unique_pull_path(b'perfdiscovery', repo, ui, path)[0]
         except ImportError:
             path = ui.expandpath(path)
 
     def s():
         repos[1] = hg.peer(ui, opts, path)
 
     def d():
         setdiscovery.findcommonheads(ui, *repos)
 
     timer(d, setup=s)
     fm.end()
 
 
 @command(
     b'perf::bookmarks|perfbookmarks',
     formatteropts
     + [
         (b'', b'clear-revlogs', False, b'refresh changelog and manifest'),
     ],
 )
 def perfbookmarks(ui, repo, **opts):
     """benchmark parsing bookmarks from disk to memory"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     clearrevlogs = opts[b'clear_revlogs']
 
     def s():
         if clearrevlogs:
             clearchangelog(repo)
         clearfilecache(repo, b'_bookmarks')
 
     def d():
         repo._bookmarks
 
     timer(d, setup=s)
     fm.end()
 
 
 @command(
     b'perf::bundle',
     [
         (
             b'r',
             b'rev',
             [],
             b'changesets to bundle',
             b'REV',
         ),
         (
             b't',
             b'type',
             b'none',
             b'bundlespec to use (see `hg help bundlespec`)',
             b'TYPE',
         ),
     ]
     + formatteropts,
     b'REVS',
 )
 def perfbundle(ui, repo, *revs, **opts):
     """benchmark the creation of a bundle from a repository
 
     For now, this only supports "none" compression.
     """
     try:
         from mercurial import bundlecaches
 
         parsebundlespec = bundlecaches.parsebundlespec
     except ImportError:
         from mercurial import exchange
 
         parsebundlespec = exchange.parsebundlespec
 
     from mercurial import discovery
     from mercurial import bundle2
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     cl = repo.changelog
     revs = list(revs)
     revs.extend(opts.get(b'rev', ()))
     revs = scmutil.revrange(repo, revs)
     if not revs:
         raise error.Abort(b"not revision specified")
     # make it a consistent set (ie: without topological gaps)
     old_len = len(revs)
     revs = list(repo.revs(b"%ld::%ld", revs, revs))
     if old_len != len(revs):
         new_count = len(revs) - old_len
         msg = b"add %d new revisions to make it a consistent set\n"
         ui.write_err(msg % new_count)
 
     targets = [cl.node(r) for r in repo.revs(b"heads(::%ld)", revs)]
     bases = [cl.node(r) for r in repo.revs(b"heads(::%ld - %ld)", revs, revs)]
     outgoing = discovery.outgoing(repo, bases, targets)
 
     bundle_spec = opts.get(b'type')
 
     bundle_spec = parsebundlespec(repo, bundle_spec, strict=False)
 
     cgversion = bundle_spec.params.get(b"cg.version")
     if cgversion is None:
         if bundle_spec.version == b'v1':
             cgversion = b'01'
         if bundle_spec.version == b'v2':
             cgversion = b'02'
     if cgversion not in changegroup.supportedoutgoingversions(repo):
         err = b"repository does not support bundle version %s"
         raise error.Abort(err % cgversion)
 
     if cgversion == b'01':  # bundle1
         bversion = b'HG10' + bundle_spec.wirecompression
         bcompression = None
     elif cgversion in (b'02', b'03'):
         bversion = b'HG20'
         bcompression = bundle_spec.wirecompression
     else:
         err = b'perf::bundle: unexpected changegroup version %s'
         raise error.ProgrammingError(err % cgversion)
 
     if bcompression is None:
         bcompression = b'UN'
 
     if bcompression != b'UN':
         err = b'perf::bundle: compression currently unsupported: %s'
         raise error.ProgrammingError(err % bcompression)
 
     def do_bundle():
         bundle2.writenewbundle(
             ui,
             repo,
             b'perf::bundle',
             os.devnull,
             bversion,
             outgoing,
             bundle_spec.params,
         )
 
     timer(do_bundle)
     fm.end()
 
 
 @command(b'perf::bundleread|perfbundleread', formatteropts, b'BUNDLE')
 def perfbundleread(ui, repo, bundlepath, **opts):
     """Benchmark reading of bundle files.
 
     This command is meant to isolate the I/O part of bundle reading as
     much as possible.
     """
     from mercurial import (
         bundle2,
         exchange,
         streamclone,
     )
 
     opts = _byteskwargs(opts)
 
     def makebench(fn):
         def run():
             with open(bundlepath, b'rb') as fh:
                 bundle = exchange.readbundle(ui, fh, bundlepath)
                 fn(bundle)
 
         return run
 
     def makereadnbytes(size):
         def run():
             with open(bundlepath, b'rb') as fh:
                 bundle = exchange.readbundle(ui, fh, bundlepath)
                 while bundle.read(size):
                     pass
 
         return run
 
     def makestdioread(size):
         def run():
             with open(bundlepath, b'rb') as fh:
                 while fh.read(size):
                     pass
 
         return run
 
     # bundle1
 
     def deltaiter(bundle):
         for delta in bundle.deltaiter():
             pass
 
     def iterchunks(bundle):
         for chunk in bundle.getchunks():
             pass
 
     # bundle2
 
     def forwardchunks(bundle):
         for chunk in bundle._forwardchunks():
             pass
 
     def iterparts(bundle):
         for part in bundle.iterparts():
             pass
 
     def iterpartsseekable(bundle):
         for part in bundle.iterparts(seekable=True):
             pass
 
     def seek(bundle):
         for part in bundle.iterparts(seekable=True):
             part.seek(0, os.SEEK_END)
 
     def makepartreadnbytes(size):
         def run():
             with open(bundlepath, b'rb') as fh:
                 bundle = exchange.readbundle(ui, fh, bundlepath)
                 for part in bundle.iterparts():
                     while part.read(size):
                         pass
 
         return run
 
     benches = [
         (makestdioread(8192), b'read(8k)'),
         (makestdioread(16384), b'read(16k)'),
         (makestdioread(32768), b'read(32k)'),
         (makestdioread(131072), b'read(128k)'),
     ]
 
     with open(bundlepath, b'rb') as fh:
         bundle = exchange.readbundle(ui, fh, bundlepath)
 
         if isinstance(bundle, changegroup.cg1unpacker):
             benches.extend(
                 [
                     (makebench(deltaiter), b'cg1 deltaiter()'),
                     (makebench(iterchunks), b'cg1 getchunks()'),
                     (makereadnbytes(8192), b'cg1 read(8k)'),
                     (makereadnbytes(16384), b'cg1 read(16k)'),
                     (makereadnbytes(32768), b'cg1 read(32k)'),
                     (makereadnbytes(131072), b'cg1 read(128k)'),
                 ]
             )
         elif isinstance(bundle, bundle2.unbundle20):
             benches.extend(
                 [
                     (makebench(forwardchunks), b'bundle2 forwardchunks()'),
                     (makebench(iterparts), b'bundle2 iterparts()'),
                     (
                         makebench(iterpartsseekable),
                         b'bundle2 iterparts() seekable',
                     ),
                     (makebench(seek), b'bundle2 part seek()'),
                     (makepartreadnbytes(8192), b'bundle2 part read(8k)'),
                     (makepartreadnbytes(16384), b'bundle2 part read(16k)'),
                     (makepartreadnbytes(32768), b'bundle2 part read(32k)'),
                     (makepartreadnbytes(131072), b'bundle2 part read(128k)'),
                 ]
             )
         elif isinstance(bundle, streamclone.streamcloneapplier):
             raise error.Abort(b'stream clone bundles not supported')
         else:
             raise error.Abort(b'unhandled bundle type: %s' % type(bundle))
 
     for fn, title in benches:
         timer, fm = gettimer(ui, opts)
         timer(fn, title=title)
         fm.end()
 
 
 @command(
     b'perf::changegroupchangelog|perfchangegroupchangelog',
     formatteropts
     + [
         (b'', b'cgversion', b'02', b'changegroup version'),
         (b'r', b'rev', b'', b'revisions to add to changegroup'),
     ],
 )
 def perfchangegroupchangelog(ui, repo, cgversion=b'02', rev=None, **opts):
     """Benchmark producing a changelog group for a changegroup.
 
     This measures the time spent processing the changelog during a
     bundle operation. This occurs during `hg bundle` and on a server
     processing a `getbundle` wire protocol request (handles clones
     and pull requests).
 
     By default, all revisions are added to the changegroup.
     """
     opts = _byteskwargs(opts)
     cl = repo.changelog
     nodes = [cl.lookup(r) for r in repo.revs(rev or b'all()')]
     bundler = changegroup.getbundler(cgversion, repo)
 
     def d():
         state, chunks = bundler._generatechangelog(cl, nodes)
         for chunk in chunks:
             pass
 
     timer, fm = gettimer(ui, opts)
 
     # Terminal printing can interfere with timing. So disable it.
     with ui.configoverride({(b'progress', b'disable'): True}):
         timer(d)
 
     fm.end()
 
 
 @command(b'perf::dirs|perfdirs', formatteropts)
 def perfdirs(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
     b'a' in dirstate
 
     def d():
         dirstate.hasdir(b'a')
         try:
             del dirstate._map._dirs
         except AttributeError:
             pass
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::dirstate|perfdirstate',
     [
         (
             b'',
             b'iteration',
             None,
             b'benchmark a full iteration for the dirstate',
         ),
         (
             b'',
             b'contains',
             None,
             b'benchmark a large amount of `nf in dirstate` calls',
         ),
     ]
     + formatteropts,
 )
 def perfdirstate(ui, repo, **opts):
     """benchmap the time of various distate operations
 
     By default benchmark the time necessary to load a dirstate from scratch.
     The dirstate is loaded to the point were a "contains" request can be
     answered.
     """
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     b"a" in repo.dirstate
 
     if opts[b'iteration'] and opts[b'contains']:
         msg = b'only specify one of --iteration or --contains'
         raise error.Abort(msg)
 
     if opts[b'iteration']:
         setup = None
         dirstate = repo.dirstate
 
         def d():
             for f in dirstate:
                 pass
 
     elif opts[b'contains']:
         setup = None
         dirstate = repo.dirstate
         allfiles = list(dirstate)
         # also add file path that will be "missing" from the dirstate
         allfiles.extend([f[::-1] for f in allfiles])
 
         def d():
             for f in allfiles:
                 f in dirstate
 
     else:
 
         def setup():
             repo.dirstate.invalidate()
 
         def d():
             b"a" in repo.dirstate
 
     timer(d, setup=setup)
     fm.end()
 
 
 @command(b'perf::dirstatedirs|perfdirstatedirs', formatteropts)
 def perfdirstatedirs(ui, repo, **opts):
     """benchmap a 'dirstate.hasdir' call from an empty `dirs` cache"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     repo.dirstate.hasdir(b"a")
 
     def setup():
         try:
             del repo.dirstate._map._dirs
         except AttributeError:
             pass
 
     def d():
         repo.dirstate.hasdir(b"a")
 
     timer(d, setup=setup)
     fm.end()
 
 
 @command(b'perf::dirstatefoldmap|perfdirstatefoldmap', formatteropts)
 def perfdirstatefoldmap(ui, repo, **opts):
     """benchmap a `dirstate._map.filefoldmap.get()` request
 
     The dirstate filefoldmap cache is dropped between every request.
     """
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
     dirstate._map.filefoldmap.get(b'a')
 
     def setup():
         del dirstate._map.filefoldmap
 
     def d():
         dirstate._map.filefoldmap.get(b'a')
 
     timer(d, setup=setup)
     fm.end()
 
 
 @command(b'perf::dirfoldmap|perfdirfoldmap', formatteropts)
 def perfdirfoldmap(ui, repo, **opts):
     """benchmap a `dirstate._map.dirfoldmap.get()` request
 
     The dirstate dirfoldmap cache is dropped between every request.
     """
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
     dirstate._map.dirfoldmap.get(b'a')
 
     def setup():
         del dirstate._map.dirfoldmap
         try:
             del dirstate._map._dirs
         except AttributeError:
             pass
 
     def d():
         dirstate._map.dirfoldmap.get(b'a')
 
     timer(d, setup=setup)
     fm.end()
 
 
 @command(b'perf::dirstatewrite|perfdirstatewrite', formatteropts)
 def perfdirstatewrite(ui, repo, **opts):
     """benchmap the time it take to write a dirstate on disk"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     ds = repo.dirstate
     b"a" in ds
 
     def setup():
         ds._dirty = True
 
     def d():
         ds.write(repo.currenttransaction())
 
     with repo.wlock():
         timer(d, setup=setup)
     fm.end()
 
 
 def _getmergerevs(repo, opts):
     """parse command argument to return rev involved in merge
 
     input: options dictionnary with `rev`, `from` and `bse`
     output: (localctx, otherctx, basectx)
     """
     if opts[b'from']:
         fromrev = scmutil.revsingle(repo, opts[b'from'])
         wctx = repo[fromrev]
     else:
         wctx = repo[None]
         # we don't want working dir files to be stat'd in the benchmark, so
         # prime that cache
         wctx.dirty()
     rctx = scmutil.revsingle(repo, opts[b'rev'], opts[b'rev'])
     if opts[b'base']:
         fromrev = scmutil.revsingle(repo, opts[b'base'])
         ancestor = repo[fromrev]
     else:
         ancestor = wctx.ancestor(rctx)
     return (wctx, rctx, ancestor)
 
 
 @command(
     b'perf::mergecalculate|perfmergecalculate',
     [
         (b'r', b'rev', b'.', b'rev to merge against'),
         (b'', b'from', b'', b'rev to merge from'),
         (b'', b'base', b'', b'the revision to use as base'),
     ]
     + formatteropts,
 )
 def perfmergecalculate(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     wctx, rctx, ancestor = _getmergerevs(repo, opts)
 
     def d():
         # acceptremote is True because we don't want prompts in the middle of
         # our benchmark
         merge.calculateupdates(
             repo,
             wctx,
             rctx,
             [ancestor],
             branchmerge=False,
             force=False,
             acceptremote=True,
             followcopies=True,
         )
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::mergecopies|perfmergecopies',
     [
         (b'r', b'rev', b'.', b'rev to merge against'),
         (b'', b'from', b'', b'rev to merge from'),
         (b'', b'base', b'', b'the revision to use as base'),
     ]
     + formatteropts,
 )
 def perfmergecopies(ui, repo, **opts):
     """measure runtime of `copies.mergecopies`"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     wctx, rctx, ancestor = _getmergerevs(repo, opts)
 
     def d():
         # acceptremote is True because we don't want prompts in the middle of
         # our benchmark
         copies.mergecopies(repo, wctx, rctx, ancestor)
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::pathcopies|perfpathcopies', [], b"REV REV")
 def perfpathcopies(ui, repo, rev1, rev2, **opts):
     """benchmark the copy tracing logic"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     ctx1 = scmutil.revsingle(repo, rev1, rev1)
     ctx2 = scmutil.revsingle(repo, rev2, rev2)
 
     def d():
         copies.pathcopies(ctx1, ctx2)
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::phases|perfphases',
     [
         (b'', b'full', False, b'include file reading time too'),
     ]
     + formatteropts,
     b"",
 )
 def perfphases(ui, repo, **opts):
     """benchmark phasesets computation"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     _phases = repo._phasecache
     full = opts.get(b'full')
     tip_rev = repo.changelog.tiprev()
 
     def d():
         phases = _phases
         if full:
             clearfilecache(repo, b'_phasecache')
             phases = repo._phasecache
         phases.invalidate()
         phases.phase(repo, tip_rev)
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::phasesremote|perfphasesremote', [], b"[DEST]")
 def perfphasesremote(ui, repo, dest=None, **opts):
     """benchmark time needed to analyse phases of the remote server"""
     from mercurial.node import bin
     from mercurial import (
         exchange,
         hg,
         phases,
     )
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     path = ui.getpath(dest, default=(b'default-push', b'default'))
     if not path:
         raise error.Abort(
             b'default repository not configured!',
             hint=b"see 'hg help config.paths'",
         )
     if util.safehasattr(path, 'main_path'):
         path = path.get_push_variant()
         dest = path.loc
     else:
         dest = path.pushloc or path.loc
     ui.statusnoi18n(b'analysing phase of %s\n' % util.hidepassword(dest))
     other = hg.peer(repo, opts, dest)
 
     # easier to perform discovery through the operation
     op = exchange.pushoperation(repo, other)
     exchange._pushdiscoverychangeset(op)
 
     remotesubset = op.fallbackheads
 
     with other.commandexecutor() as e:
         remotephases = e.callcommand(
             b'listkeys', {b'namespace': b'phases'}
         ).result()
     del other
     publishing = remotephases.get(b'publishing', False)
     if publishing:
         ui.statusnoi18n(b'publishing: yes\n')
     else:
         ui.statusnoi18n(b'publishing: no\n')
 
     has_node = getattr(repo.changelog.index, 'has_node', None)
     if has_node is None:
         has_node = repo.changelog.nodemap.__contains__
     nonpublishroots = 0
     for nhex, phase in remotephases.iteritems():
         if nhex == b'publishing':  # ignore data related to publish option
             continue
         node = bin(nhex)
         if has_node(node) and int(phase):
             nonpublishroots += 1
     ui.statusnoi18n(b'number of roots: %d\n' % len(remotephases))
     ui.statusnoi18n(b'number of known non public roots: %d\n' % nonpublishroots)
 
     def d():
         phases.remotephasessummary(repo, remotesubset, remotephases)
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::manifest|perfmanifest',
     [
         (b'm', b'manifest-rev', False, b'Look up a manifest node revision'),
         (b'', b'clear-disk', False, b'clear on-disk caches too'),
     ]
     + formatteropts,
     b'REV|NODE',
 )
 def perfmanifest(ui, repo, rev, manifest_rev=False, clear_disk=False, **opts):
     """benchmark the time to read a manifest from disk and return a usable
     dict-like object
 
     Manifest caches are cleared before retrieval."""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     if not manifest_rev:
         ctx = scmutil.revsingle(repo, rev, rev)
         t = ctx.manifestnode()
     else:
         from mercurial.node import bin
 
         if len(rev) == 40:
             t = bin(rev)
         else:
             try:
                 rev = int(rev)
 
                 if util.safehasattr(repo.manifestlog, b'getstorage'):
                     t = repo.manifestlog.getstorage(b'').node(rev)
                 else:
                     t = repo.manifestlog._revlog.lookup(rev)
             except ValueError:
                 raise error.Abort(
                     b'manifest revision must be integer or full node'
                 )
 
     def d():
         repo.manifestlog.clearcaches(clear_persisted_data=clear_disk)
         repo.manifestlog[t].read()
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::changeset|perfchangeset', formatteropts)
 def perfchangeset(ui, repo, rev, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     n = scmutil.revsingle(repo, rev).node()
 
     def d():
         repo.changelog.read(n)
         # repo.changelog._cache = None
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::ignore|perfignore', formatteropts)
 def perfignore(ui, repo, **opts):
     """benchmark operation related to computing ignore"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     dirstate = repo.dirstate
 
     def setupone():
         dirstate.invalidate()
         clearfilecache(dirstate, b'_ignore')
 
     def runone():
         dirstate._ignore
 
     timer(runone, setup=setupone, title=b"load")
     fm.end()
 
 
 @command(
     b'perf::index|perfindex',
     [
         (b'', b'rev', [], b'revision to be looked up (default tip)'),
         (b'', b'no-lookup', None, b'do not revision lookup post creation'),
     ]
     + formatteropts,
 )
 def perfindex(ui, repo, **opts):
     """benchmark index creation time followed by a lookup
 
     The default is to look `tip` up. Depending on the index implementation,
     the revision looked up can matters. For example, an implementation
     scanning the index will have a faster lookup time for `--rev tip` than for
     `--rev 0`. The number of looked up revisions and their order can also
     matters.
 
     Example of useful set to test:
 
     * tip
     * 0
     * -10:
     * :10
     * -10: + :10
     * :10: + -10:
     * -10000:
     * -10000: + 0
 
     It is not currently possible to check for lookup of a missing node. For
     deeper lookup benchmarking, checkout the `perfnodemap` command."""
     import mercurial.revlog
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     mercurial.revlog._prereadsize = 2 ** 24  # disable lazy parser in old hg
     if opts[b'no_lookup']:
         if opts['rev']:
             raise error.Abort('--no-lookup and --rev are mutually exclusive')
         nodes = []
     elif not opts[b'rev']:
         nodes = [repo[b"tip"].node()]
     else:
         revs = scmutil.revrange(repo, opts[b'rev'])
         cl = repo.changelog
         nodes = [cl.node(r) for r in revs]
 
     unfi = repo.unfiltered()
     # find the filecache func directly
     # This avoid polluting the benchmark with the filecache logic
     makecl = unfi.__class__.changelog.func
 
     def setup():
         # probably not necessary, but for good measure
         clearchangelog(unfi)
 
     def d():
         cl = makecl(unfi)
         for n in nodes:
             cl.rev(n)
 
     timer(d, setup=setup)
     fm.end()
 
 
 @command(
     b'perf::nodemap|perfnodemap',
     [
         (b'', b'rev', [], b'revision to be looked up (default tip)'),
         (b'', b'clear-caches', True, b'clear revlog cache between calls'),
     ]
     + formatteropts,
 )
 def perfnodemap(ui, repo, **opts):
     """benchmark the time necessary to look up revision from a cold nodemap
 
     Depending on the implementation, the amount and order of revision we look
     up can varies. Example of useful set to test:
     * tip
     * 0
     * -10:
     * :10
     * -10: + :10
     * :10: + -10:
     * -10000:
     * -10000: + 0
 
     The command currently focus on valid binary lookup. Benchmarking for
     hexlookup, prefix lookup and missing lookup would also be valuable.
     """
     import mercurial.revlog
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     mercurial.revlog._prereadsize = 2 ** 24  # disable lazy parser in old hg
 
     unfi = repo.unfiltered()
     clearcaches = opts[b'clear_caches']
     # find the filecache func directly
     # This avoid polluting the benchmark with the filecache logic
     makecl = unfi.__class__.changelog.func
     if not opts[b'rev']:
         raise error.Abort(b'use --rev to specify revisions to look up')
     revs = scmutil.revrange(repo, opts[b'rev'])
     cl = repo.changelog
     nodes = [cl.node(r) for r in revs]
 
     # use a list to pass reference to a nodemap from one closure to the next
     nodeget = [None]
 
     def setnodeget():
         # probably not necessary, but for good measure
         clearchangelog(unfi)
         cl = makecl(unfi)
         if util.safehasattr(cl.index, 'get_rev'):
             nodeget[0] = cl.index.get_rev
         else:
             nodeget[0] = cl.nodemap.get
 
     def d():
         get = nodeget[0]
         for n in nodes:
             get(n)
 
     setup = None
     if clearcaches:
 
         def setup():
             setnodeget()
 
     else:
         setnodeget()
         d()  # prewarm the data structure
     timer(d, setup=setup)
     fm.end()
 
 
 @command(b'perf::startup|perfstartup', formatteropts)
 def perfstartup(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     def d():
         if os.name != 'nt':
             os.system(
                 b"HGRCPATH= %s version -q > /dev/null" % fsencode(sys.argv[0])
             )
         else:
             os.environ['HGRCPATH'] = r' '
             os.system("%s version -q > NUL" % sys.argv[0])
 
     timer(d)
     fm.end()
 
 
+def _clear_store_audit_cache(repo):
+    vfs = getsvfs(repo)
+    # unwrap the fncache proxy
+    if not hasattr(vfs, "audit"):
+        vfs = getattr(vfs, "vfs", vfs)
+    auditor = vfs.audit
+    if hasattr(auditor, "clear_audit_cache"):
+        auditor.clear_audit_cache()
+    elif hasattr(auditor, "audited"):
+        auditor.audited.clear()
+        auditor.auditeddir.clear()
+
+
 def _find_stream_generator(version):
     """find the proper generator function for this stream version"""
     import mercurial.streamclone
 
     available = {}
 
     # try to fetch a v1 generator
     generatev1 = getattr(mercurial.streamclone, "generatev1", None)
     if generatev1 is not None:
 
         def generate(repo):
-            entries, bytes, data = generatev2(repo, None, None, True)
+            entries, bytes, data = generatev1(repo, None, None, True)
             return data
 
         available[b'v1'] = generatev1
     # try to fetch a v2 generator
     generatev2 = getattr(mercurial.streamclone, "generatev2", None)
     if generatev2 is not None:
 
         def generate(repo):
             entries, bytes, data = generatev2(repo, None, None, True)
             return data
 
         available[b'v2'] = generate
     # try to fetch a v3 generator
     generatev3 = getattr(mercurial.streamclone, "generatev3", None)
     if generatev3 is not None:
 
         def generate(repo):
-            entries, bytes, data = generatev3(repo, None, None, True)
-            return data
+            return generatev3(repo, None, None, True)
 
         available[b'v3-exp'] = generate
 
     # resolve the request
     if version == b"latest":
         # latest is the highest non experimental version
         latest_key = max(v for v in available if b'-exp' not in v)
         return available[latest_key]
     elif version in available:
         return available[version]
     else:
         msg = b"unkown or unavailable version: %s"
         msg %= version
         hint = b"available versions: %s"
         hint %= b', '.join(sorted(available))
         raise error.Abort(msg, hint=hint)
 
 
 @command(
     b'perf::stream-locked-section',
     [
         (
             b'',
             b'stream-version',
             b'latest',
-            b'stream version to use ("v1", "v2", "v3" or "latest", (the default))',
+            b'stream version to use ("v1", "v2", "v3-exp" '
+            b'or "latest", (the default))',
         ),
     ]
     + formatteropts,
 )
 def perf_stream_clone_scan(ui, repo, stream_version, **opts):
     """benchmark the initial, repo-locked, section of a stream-clone"""
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     # deletion of the generator may trigger some cleanup that we do not want to
     # measure
     result_holder = [None]
 
     def setupone():
         result_holder[0] = None
+        # This is important for the full generation, even if it does not
+        # currently matters, it seems safer to also real it here.
+        _clear_store_audit_cache(repo)
 
     generate = _find_stream_generator(stream_version)
 
     def runone():
         # the lock is held for the duration the initialisation
         result_holder[0] = generate(repo)
 
     timer(runone, setup=setupone, title=b"load")
     fm.end()
 
 
 @command(
     b'perf::stream-generate',
     [
         (
             b'',
             b'stream-version',
             b'latest',
-            b'stream version to us ("v1", "v2" or "latest", (the default))',
+            b'stream version to us ("v1", "v2", "v3-exp" '
+            b'or "latest", (the default))',
         ),
     ]
     + formatteropts,
 )
 def perf_stream_clone_generate(ui, repo, stream_version, **opts):
     """benchmark the full generation of a stream clone"""
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     # deletion of the generator may trigger some cleanup that we do not want to
     # measure
 
     generate = _find_stream_generator(stream_version)
 
+    def setup():
+        _clear_store_audit_cache(repo)
+
     def runone():
         # the lock is held for the duration the initialisation
         for chunk in generate(repo):
             pass
 
-    timer(runone, title=b"generate")
+    timer(runone, setup=setup, title=b"generate")
     fm.end()
 
 
 @command(
     b'perf::stream-consume',
     formatteropts,
 )
 def perf_stream_clone_consume(ui, repo, filename, **opts):
     """benchmark the full application of a stream clone
 
     This include the creation of the repository
     """
     # try except to appease check code
     msg = b"mercurial too old, missing necessary module: %s"
     try:
         from mercurial import bundle2
     except ImportError as exc:
         msg %= _bytestr(exc)
         raise error.Abort(msg)
     try:
         from mercurial import exchange
     except ImportError as exc:
         msg %= _bytestr(exc)
         raise error.Abort(msg)
     try:
         from mercurial import hg
     except ImportError as exc:
         msg %= _bytestr(exc)
         raise error.Abort(msg)
     try:
         from mercurial import localrepo
     except ImportError as exc:
         msg %= _bytestr(exc)
         raise error.Abort(msg)
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     # deletion of the generator may trigger some cleanup that we do not want to
     # measure
     if not (os.path.isfile(filename) and os.access(filename, os.R_OK)):
         raise error.Abort("not a readable file: %s" % filename)
 
     run_variables = [None, None]
 
+    # we create the new repository next to the other one for two reasons:
+    # - this way we use the same file system, which are relevant for benchmark
+    # - if /tmp/ is small, the operation could overfills it.
+    source_repo_dir = os.path.dirname(repo.root)
+
     @contextlib.contextmanager
     def context():
         with open(filename, mode='rb') as bundle:
-            with tempfile.TemporaryDirectory() as tmp_dir:
+            with tempfile.TemporaryDirectory(
+                prefix=b'hg-perf-stream-consume-',
+                dir=source_repo_dir,
+            ) as tmp_dir:
                 tmp_dir = fsencode(tmp_dir)
                 run_variables[0] = bundle
                 run_variables[1] = tmp_dir
                 yield
                 run_variables[0] = None
                 run_variables[1] = None
 
     def runone():
         bundle = run_variables[0]
         tmp_dir = run_variables[1]
+
+        # we actually wants to copy all config to ensure the repo config is
+        # taken in account during the benchmark
+        new_ui = repo.ui.__class__(repo.ui)
         # only pass ui when no srcrepo
         localrepo.createrepository(
-            repo.ui, tmp_dir, requirements=repo.requirements
+            new_ui, tmp_dir, requirements=repo.requirements
         )
-        target = hg.repository(repo.ui, tmp_dir)
+        target = hg.repository(new_ui, tmp_dir)
         gen = exchange.readbundle(target.ui, bundle, bundle.name)
         # stream v1
         if util.safehasattr(gen, 'apply'):
             gen.apply(target)
         else:
             with target.transaction(b"perf::stream-consume") as tr:
                 bundle2.applybundle(
                     target,
                     gen,
                     tr,
                     source=b'unbundle',
                     url=filename,
                 )
 
     timer(runone, context=context, title=b"consume")
     fm.end()
 
 
 @command(b'perf::parents|perfparents', formatteropts)
 def perfparents(ui, repo, **opts):
     """benchmark the time necessary to fetch one changeset's parents.
 
     The fetch is done using the `node identifier`, traversing all object layers
     from the repository object. The first N revisions will be used for this
     benchmark. N is controlled by the ``perf.parentscount`` config option
     (default: 1000).
     """
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     # control the number of commits perfparents iterates over
     # experimental config: perf.parentscount
     count = getint(ui, b"perf", b"parentscount", 1000)
     if len(repo.changelog) < count:
         raise error.Abort(b"repo needs %d commits for this test" % count)
     repo = repo.unfiltered()
     nl = [repo.changelog.node(i) for i in _xrange(count)]
 
     def d():
         for n in nl:
             repo.changelog.parents(n)
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::ctxfiles|perfctxfiles', formatteropts)
 def perfctxfiles(ui, repo, x, **opts):
     opts = _byteskwargs(opts)
     x = int(x)
     timer, fm = gettimer(ui, opts)
 
     def d():
         len(repo[x].files())
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::rawfiles|perfrawfiles', formatteropts)
 def perfrawfiles(ui, repo, x, **opts):
     opts = _byteskwargs(opts)
     x = int(x)
     timer, fm = gettimer(ui, opts)
     cl = repo.changelog
 
     def d():
         len(cl.read(x)[3])
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::lookup|perflookup', formatteropts)
 def perflookup(ui, repo, rev, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     timer(lambda: len(repo.lookup(rev)))
     fm.end()
 
 
 @command(
     b'perf::linelogedits|perflinelogedits',
     [
         (b'n', b'edits', 10000, b'number of edits'),
         (b'', b'max-hunk-lines', 10, b'max lines in a hunk'),
     ],
     norepo=True,
 )
 def perflinelogedits(ui, **opts):
     from mercurial import linelog
 
     opts = _byteskwargs(opts)
 
     edits = opts[b'edits']
     maxhunklines = opts[b'max_hunk_lines']
 
     maxb1 = 100000
     random.seed(0)
     randint = random.randint
     currentlines = 0
     arglist = []
     for rev in _xrange(edits):
         a1 = randint(0, currentlines)
         a2 = randint(a1, min(currentlines, a1 + maxhunklines))
         b1 = randint(0, maxb1)
         b2 = randint(b1, b1 + maxhunklines)
         currentlines += (b2 - b1) - (a2 - a1)
         arglist.append((rev, a1, a2, b1, b2))
 
     def d():
         ll = linelog.linelog()
         for args in arglist:
             ll.replacelines(*args)
 
     timer, fm = gettimer(ui, opts)
     timer(d)
     fm.end()
 
 
 @command(b'perf::revrange|perfrevrange', formatteropts)
 def perfrevrange(ui, repo, *specs, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     revrange = scmutil.revrange
     timer(lambda: len(revrange(repo, specs)))
     fm.end()
 
 
 @command(b'perf::nodelookup|perfnodelookup', formatteropts)
 def perfnodelookup(ui, repo, rev, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     import mercurial.revlog
 
     mercurial.revlog._prereadsize = 2 ** 24  # disable lazy parser in old hg
     n = scmutil.revsingle(repo, rev).node()
 
     try:
         cl = revlog(getsvfs(repo), radix=b"00changelog")
     except TypeError:
         cl = revlog(getsvfs(repo), indexfile=b"00changelog.i")
 
     def d():
         cl.rev(n)
         clearcaches(cl)
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::log|perflog',
     [(b'', b'rename', False, b'ask log to follow renames')] + formatteropts,
 )
 def perflog(ui, repo, rev=None, **opts):
     opts = _byteskwargs(opts)
     if rev is None:
         rev = []
     timer, fm = gettimer(ui, opts)
     ui.pushbuffer()
     timer(
         lambda: commands.log(
             ui, repo, rev=rev, date=b'', user=b'', copies=opts.get(b'rename')
         )
     )
     ui.popbuffer()
     fm.end()
 
 
 @command(b'perf::moonwalk|perfmoonwalk', formatteropts)
 def perfmoonwalk(ui, repo, **opts):
     """benchmark walking the changelog backwards
 
     This also loads the changelog data for each revision in the changelog.
     """
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
 
     def moonwalk():
         for i in repo.changelog.revs(start=(len(repo) - 1), stop=-1):
             ctx = repo[i]
             ctx.branch()  # read changelog data (in addition to the index)
 
     timer(moonwalk)
     fm.end()
 
 
 @command(
     b'perf::templating|perftemplating',
     [
         (b'r', b'rev', [], b'revisions to run the template on'),
     ]
     + formatteropts,
 )
 def perftemplating(ui, repo, testedtemplate=None, **opts):
     """test the rendering time of a given template"""
     if makelogtemplater is None:
         raise error.Abort(
             b"perftemplating not available with this Mercurial",
             hint=b"use 4.3 or later",
         )
 
     opts = _byteskwargs(opts)
 
     nullui = ui.copy()
     nullui.fout = open(os.devnull, 'wb')
     nullui.disablepager()
     revs = opts.get(b'rev')
     if not revs:
         revs = [b'all()']
     revs = list(scmutil.revrange(repo, revs))
 
     defaulttemplate = (
         b'{date|shortdate} [{rev}:{node|short}]'
         b' {author|person}: {desc|firstline}\n'
     )
     if testedtemplate is None:
         testedtemplate = defaulttemplate
     displayer = makelogtemplater(nullui, repo, testedtemplate)
 
     def format():
         for r in revs:
             ctx = repo[r]
             displayer.show(ctx)
             displayer.flush(ctx)
 
     timer, fm = gettimer(ui, opts)
     timer(format)
     fm.end()
 
 
 def _displaystats(ui, opts, entries, data):
     # use a second formatter because the data are quite different, not sure
     # how it flies with the templater.
     fm = ui.formatter(b'perf-stats', opts)
     for key, title in entries:
         values = data[key]
         nbvalues = len(data)
         values.sort()
         stats = {
             'key': key,
             'title': title,
             'nbitems': len(values),
             'min': values[0][0],
             '10%': values[(nbvalues * 10) // 100][0],
             '25%': values[(nbvalues * 25) // 100][0],
             '50%': values[(nbvalues * 50) // 100][0],
             '75%': values[(nbvalues * 75) // 100][0],
             '80%': values[(nbvalues * 80) // 100][0],
             '85%': values[(nbvalues * 85) // 100][0],
             '90%': values[(nbvalues * 90) // 100][0],
             '95%': values[(nbvalues * 95) // 100][0],
             '99%': values[(nbvalues * 99) // 100][0],
             'max': values[-1][0],
         }
         fm.startitem()
         fm.data(**stats)
         # make node pretty for the human output
         fm.plain('### %s (%d items)\n' % (title, len(values)))
         lines = [
             'min',
             '10%',
             '25%',
             '50%',
             '75%',
             '80%',
             '85%',
             '90%',
             '95%',
             '99%',
             'max',
         ]
         for l in lines:
             fm.plain('%s: %s\n' % (l, stats[l]))
     fm.end()
 
 
 @command(
     b'perf::helper-mergecopies|perfhelper-mergecopies',
     formatteropts
     + [
         (b'r', b'revs', [], b'restrict search to these revisions'),
         (b'', b'timing', False, b'provides extra data (costly)'),
         (b'', b'stats', False, b'provides statistic about the measured data'),
     ],
 )
 def perfhelpermergecopies(ui, repo, revs=[], **opts):
     """find statistics about potential parameters for `perfmergecopies`
 
     This command find (base, p1, p2) triplet relevant for copytracing
     benchmarking in the context of a merge.  It reports values for some of the
     parameters that impact merge copy tracing time during merge.
 
     If `--timing` is set, rename detection is run and the associated timing
     will be reported. The extra details come at the cost of slower command
     execution.
 
     Since rename detection is only run once, other factors might easily
     affect the precision of the timing. However it should give a good
     approximation of which revision triplets are very costly.
     """
     opts = _byteskwargs(opts)
     fm = ui.formatter(b'perf', opts)
     dotiming = opts[b'timing']
     dostats = opts[b'stats']
 
     output_template = [
         ("base", "%(base)12s"),
         ("p1", "%(p1.node)12s"),
         ("p2", "%(p2.node)12s"),
         ("p1.nb-revs", "%(p1.nbrevs)12d"),
         ("p1.nb-files", "%(p1.nbmissingfiles)12d"),
         ("p1.renames", "%(p1.renamedfiles)12d"),
         ("p1.time", "%(p1.time)12.3f"),
         ("p2.nb-revs", "%(p2.nbrevs)12d"),
         ("p2.nb-files", "%(p2.nbmissingfiles)12d"),
         ("p2.renames", "%(p2.renamedfiles)12d"),
         ("p2.time", "%(p2.time)12.3f"),
         ("renames", "%(nbrenamedfiles)12d"),
         ("total.time", "%(time)12.3f"),
     ]
     if not dotiming:
         output_template = [
             i
             for i in output_template
             if not ('time' in i[0] or 'renames' in i[0])
         ]
     header_names = [h for (h, v) in output_template]
     output = ' '.join([v for (h, v) in output_template]) + '\n'
     header = ' '.join(['%12s'] * len(header_names)) + '\n'
     fm.plain(header % tuple(header_names))
 
     if not revs:
         revs = ['all()']
     revs = scmutil.revrange(repo, revs)
 
     if dostats:
         alldata = {
             'nbrevs': [],
             'nbmissingfiles': [],
         }
         if dotiming:
             alldata['parentnbrenames'] = []
             alldata['totalnbrenames'] = []
             alldata['parenttime'] = []
             alldata['totaltime'] = []
 
     roi = repo.revs('merge() and %ld', revs)
     for r in roi:
         ctx = repo[r]
         p1 = ctx.p1()
         p2 = ctx.p2()
         bases = repo.changelog._commonancestorsheads(p1.rev(), p2.rev())
         for b in bases:
             b = repo[b]
             p1missing = copies._computeforwardmissing(b, p1)
             p2missing = copies._computeforwardmissing(b, p2)
             data = {
                 b'base': b.hex(),
                 b'p1.node': p1.hex(),
                 b'p1.nbrevs': len(repo.revs('only(%d, %d)', p1.rev(), b.rev())),
                 b'p1.nbmissingfiles': len(p1missing),
                 b'p2.node': p2.hex(),
                 b'p2.nbrevs': len(repo.revs('only(%d, %d)', p2.rev(), b.rev())),
                 b'p2.nbmissingfiles': len(p2missing),
             }
             if dostats:
                 if p1missing:
                     alldata['nbrevs'].append(
                         (data['p1.nbrevs'], b.hex(), p1.hex())
                     )
                     alldata['nbmissingfiles'].append(
                         (data['p1.nbmissingfiles'], b.hex(), p1.hex())
                     )
                 if p2missing:
                     alldata['nbrevs'].append(
                         (data['p2.nbrevs'], b.hex(), p2.hex())
                     )
                     alldata['nbmissingfiles'].append(
                         (data['p2.nbmissingfiles'], b.hex(), p2.hex())
                     )
             if dotiming:
                 begin = util.timer()
                 mergedata = copies.mergecopies(repo, p1, p2, b)
                 end = util.timer()
                 # not very stable timing since we did only one run
                 data['time'] = end - begin
                 # mergedata contains five dicts: "copy", "movewithdir",
                 # "diverge", "renamedelete" and "dirmove".
                 # The first 4 are about renamed file so lets count that.
                 renames = len(mergedata[0])
                 renames += len(mergedata[1])
                 renames += len(mergedata[2])
                 renames += len(mergedata[3])
                 data['nbrenamedfiles'] = renames
                 begin = util.timer()
                 p1renames = copies.pathcopies(b, p1)
                 end = util.timer()
                 data['p1.time'] = end - begin
                 begin = util.timer()
                 p2renames = copies.pathcopies(b, p2)
                 end = util.timer()
                 data['p2.time'] = end - begin
                 data['p1.renamedfiles'] = len(p1renames)
                 data['p2.renamedfiles'] = len(p2renames)
 
                 if dostats:
                     if p1missing:
                         alldata['parentnbrenames'].append(
                             (data['p1.renamedfiles'], b.hex(), p1.hex())
                         )
                         alldata['parenttime'].append(
                             (data['p1.time'], b.hex(), p1.hex())
                         )
                     if p2missing:
                         alldata['parentnbrenames'].append(
                             (data['p2.renamedfiles'], b.hex(), p2.hex())
                         )
                         alldata['parenttime'].append(
                             (data['p2.time'], b.hex(), p2.hex())
                         )
                     if p1missing or p2missing:
                         alldata['totalnbrenames'].append(
                             (
                                 data['nbrenamedfiles'],
                                 b.hex(),
                                 p1.hex(),
                                 p2.hex(),
                             )
                         )
                         alldata['totaltime'].append(
                             (data['time'], b.hex(), p1.hex(), p2.hex())
                         )
             fm.startitem()
             fm.data(**data)
             # make node pretty for the human output
             out = data.copy()
             out['base'] = fm.hexfunc(b.node())
             out['p1.node'] = fm.hexfunc(p1.node())
             out['p2.node'] = fm.hexfunc(p2.node())
             fm.plain(output % out)
 
     fm.end()
     if dostats:
         # use a second formatter because the data are quite different, not sure
         # how it flies with the templater.
         entries = [
             ('nbrevs', 'number of revision covered'),
             ('nbmissingfiles', 'number of missing files at head'),
         ]
         if dotiming:
             entries.append(
                 ('parentnbrenames', 'rename from one parent to base')
             )
             entries.append(('totalnbrenames', 'total number of renames'))
             entries.append(('parenttime', 'time for one parent'))
             entries.append(('totaltime', 'time for both parents'))
         _displaystats(ui, opts, entries, alldata)
 
 
 @command(
     b'perf::helper-pathcopies|perfhelper-pathcopies',
     formatteropts
     + [
         (b'r', b'revs', [], b'restrict search to these revisions'),
         (b'', b'timing', False, b'provides extra data (costly)'),
         (b'', b'stats', False, b'provides statistic about the measured data'),
     ],
 )
 def perfhelperpathcopies(ui, repo, revs=[], **opts):
     """find statistic about potential parameters for the `perftracecopies`
 
     This command find source-destination pair relevant for copytracing testing.
     It report value for some of the parameters that impact copy tracing time.
 
     If `--timing` is set, rename detection is run and the associated timing
     will be reported. The extra details comes at the cost of a slower command
     execution.
 
     Since the rename detection is only run once, other factors might easily
     affect the precision of the timing. However it should give a good
     approximation of which revision pairs are very costly.
     """
     opts = _byteskwargs(opts)
     fm = ui.formatter(b'perf', opts)
     dotiming = opts[b'timing']
     dostats = opts[b'stats']
 
     if dotiming:
         header = '%12s %12s %12s %12s %12s %12s\n'
         output = (
             "%(source)12s %(destination)12s "
             "%(nbrevs)12d %(nbmissingfiles)12d "
             "%(nbrenamedfiles)12d %(time)18.5f\n"
         )
         header_names = (
             "source",
             "destination",
             "nb-revs",
             "nb-files",
             "nb-renames",
             "time",
         )
         fm.plain(header % header_names)
     else:
         header = '%12s %12s %12s %12s\n'
         output = (
             "%(source)12s %(destination)12s "
             "%(nbrevs)12d %(nbmissingfiles)12d\n"
         )
         fm.plain(header % ("source", "destination", "nb-revs", "nb-files"))
 
     if not revs:
         revs = ['all()']
     revs = scmutil.revrange(repo, revs)
 
     if dostats:
         alldata = {
             'nbrevs': [],
             'nbmissingfiles': [],
         }
         if dotiming:
             alldata['nbrenames'] = []
             alldata['time'] = []
 
     roi = repo.revs('merge() and %ld', revs)
     for r in roi:
         ctx = repo[r]
         p1 = ctx.p1().rev()
         p2 = ctx.p2().rev()
         bases = repo.changelog._commonancestorsheads(p1, p2)
         for p in (p1, p2):
             for b in bases:
                 base = repo[b]
                 parent = repo[p]
                 missing = copies._computeforwardmissing(base, parent)
                 if not missing:
                     continue
                 data = {
                     b'source': base.hex(),
                     b'destination': parent.hex(),
                     b'nbrevs': len(repo.revs('only(%d, %d)', p, b)),
                     b'nbmissingfiles': len(missing),
                 }
                 if dostats:
                     alldata['nbrevs'].append(
                         (
                             data['nbrevs'],
                             base.hex(),
                             parent.hex(),
                         )
                     )
                     alldata['nbmissingfiles'].append(
                         (
                             data['nbmissingfiles'],
                             base.hex(),
                             parent.hex(),
                         )
                     )
                 if dotiming:
                     begin = util.timer()
                     renames = copies.pathcopies(base, parent)
                     end = util.timer()
                     # not very stable timing since we did only one run
                     data['time'] = end - begin
                     data['nbrenamedfiles'] = len(renames)
                     if dostats:
                         alldata['time'].append(
                             (
                                 data['time'],
                                 base.hex(),
                                 parent.hex(),
                             )
                         )
                         alldata['nbrenames'].append(
                             (
                                 data['nbrenamedfiles'],
                                 base.hex(),
                                 parent.hex(),
                             )
                         )
                 fm.startitem()
                 fm.data(**data)
                 out = data.copy()
                 out['source'] = fm.hexfunc(base.node())
                 out['destination'] = fm.hexfunc(parent.node())
                 fm.plain(output % out)
 
     fm.end()
     if dostats:
         entries = [
             ('nbrevs', 'number of revision covered'),
             ('nbmissingfiles', 'number of missing files at head'),
         ]
         if dotiming:
             entries.append(('nbrenames', 'renamed files'))
             entries.append(('time', 'time'))
         _displaystats(ui, opts, entries, alldata)
 
 
 @command(b'perf::cca|perfcca', formatteropts)
 def perfcca(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     timer(lambda: scmutil.casecollisionauditor(ui, False, repo.dirstate))
     fm.end()
 
 
 @command(b'perf::fncacheload|perffncacheload', formatteropts)
 def perffncacheload(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     s = repo.store
 
     def d():
         s.fncache._load()
 
     timer(d)
     fm.end()
 
 
 @command(b'perf::fncachewrite|perffncachewrite', formatteropts)
 def perffncachewrite(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     s = repo.store
     lock = repo.lock()
     s.fncache._load()
     tr = repo.transaction(b'perffncachewrite')
     tr.addbackup(b'fncache')
 
     def d():
         s.fncache._dirty = True
         s.fncache.write(tr)
 
     timer(d)
     tr.close()
     lock.release()
     fm.end()
 
 
 @command(b'perf::fncacheencode|perffncacheencode', formatteropts)
 def perffncacheencode(ui, repo, **opts):
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     s = repo.store
     s.fncache._load()
 
     def d():
         for p in s.fncache.entries:
             s.encode(p)
 
     timer(d)
     fm.end()
 
 
 def _bdiffworker(q, blocks, xdiff, ready, done):
     while not done.is_set():
         pair = q.get()
         while pair is not None:
             if xdiff:
                 mdiff.bdiff.xdiffblocks(*pair)
             elif blocks:
                 mdiff.bdiff.blocks(*pair)
             else:
                 mdiff.textdiff(*pair)
             q.task_done()
             pair = q.get()
         q.task_done()  # for the None one
         with ready:
             ready.wait()
 
 
 def _manifestrevision(repo, mnode):
     ml = repo.manifestlog
 
     if util.safehasattr(ml, b'getstorage'):
         store = ml.getstorage(b'')
     else:
         store = ml._revlog
 
     return store.revision(mnode)
 
 
 @command(
     b'perf::bdiff|perfbdiff',
     revlogopts
     + formatteropts
     + [
         (
             b'',
             b'count',
             1,
             b'number of revisions to test (when using --startrev)',
         ),
         (b'', b'alldata', False, b'test bdiffs for all associated revisions'),
         (b'', b'threads', 0, b'number of thread to use (disable with 0)'),
         (b'', b'blocks', False, b'test computing diffs into blocks'),
         (b'', b'xdiff', False, b'use xdiff algorithm'),
     ],
     b'-c|-m|FILE REV',
 )
 def perfbdiff(ui, repo, file_, rev=None, count=None, threads=0, **opts):
     """benchmark a bdiff between revisions
 
     By default, benchmark a bdiff between its delta parent and itself.
 
     With ``--count``, benchmark bdiffs between delta parents and self for N
     revisions starting at the specified revision.
 
     With ``--alldata``, assume the requested revision is a changeset and
     measure bdiffs for all changes related to that changeset (manifest
     and filelogs).
     """
     opts = _byteskwargs(opts)
 
     if opts[b'xdiff'] and not opts[b'blocks']:
         raise error.CommandError(b'perfbdiff', b'--xdiff requires --blocks')
 
     if opts[b'alldata']:
         opts[b'changelog'] = True
 
     if opts.get(b'changelog') or opts.get(b'manifest'):
         file_, rev = None, file_
     elif rev is None:
         raise error.CommandError(b'perfbdiff', b'invalid arguments')
 
     blocks = opts[b'blocks']
     xdiff = opts[b'xdiff']
     textpairs = []
 
     r = cmdutil.openrevlog(repo, b'perfbdiff', file_, opts)
 
     startrev = r.rev(r.lookup(rev))
     for rev in range(startrev, min(startrev + count, len(r) - 1)):
         if opts[b'alldata']:
             # Load revisions associated with changeset.
             ctx = repo[rev]
             mtext = _manifestrevision(repo, ctx.manifestnode())
             for pctx in ctx.parents():
                 pman = _manifestrevision(repo, pctx.manifestnode())
                 textpairs.append((pman, mtext))
 
             # Load filelog revisions by iterating manifest delta.
             man = ctx.manifest()
             pman = ctx.p1().manifest()
             for filename, change in pman.diff(man).items():
                 fctx = repo.file(filename)
                 f1 = fctx.revision(change[0][0] or -1)
                 f2 = fctx.revision(change[1][0] or -1)
                 textpairs.append((f1, f2))
         else:
             dp = r.deltaparent(rev)
             textpairs.append((r.revision(dp), r.revision(rev)))
 
     withthreads = threads > 0
     if not withthreads:
 
         def d():
             for pair in textpairs:
                 if xdiff:
                     mdiff.bdiff.xdiffblocks(*pair)
                 elif blocks:
                     mdiff.bdiff.blocks(*pair)
                 else:
                     mdiff.textdiff(*pair)
 
     else:
         q = queue()
         for i in _xrange(threads):
             q.put(None)
         ready = threading.Condition()
         done = threading.Event()
         for i in _xrange(threads):
             threading.Thread(
                 target=_bdiffworker, args=(q, blocks, xdiff, ready, done)
             ).start()
         q.join()
 
         def d():
             for pair in textpairs:
                 q.put(pair)
             for i in _xrange(threads):
                 q.put(None)
             with ready:
                 ready.notify_all()
             q.join()
 
     timer, fm = gettimer(ui, opts)
     timer(d)
     fm.end()
 
     if withthreads:
         done.set()
         for i in _xrange(threads):
             q.put(None)
         with ready:
             ready.notify_all()
 
 
 @command(
     b'perf::unbundle',
     [
         (b'', b'as-push', None, b'pretend the bundle comes from a push'),
     ]
     + formatteropts,
     b'BUNDLE_FILE',
 )
 def perf_unbundle(ui, repo, fname, **opts):
     """benchmark application of a bundle in a repository.
 
     This does not include the final transaction processing
 
     The --as-push option make the unbundle operation appears like it comes from
     a client push. It change some aspect of the processing and associated
     performance profile.
     """
 
     from mercurial import exchange
     from mercurial import bundle2
     from mercurial import transaction
 
     opts = _byteskwargs(opts)
 
     ###  some compatibility hotfix
     #
     # the data attribute is dropped in 63edc384d3b7 a changeset introducing a
     # critical regression that break transaction rollback for files that are
     # de-inlined.
     method = transaction.transaction._addentry
     pre_63edc384d3b7 = "data" in getargspec(method).args
     # the `detailed_exit_code` attribute is introduced in 33c0c25d0b0f
     # a changeset that is a close descendant of 18415fc918a1, the changeset
     # that conclude the fix run for the bug introduced in 63edc384d3b7.
     args = getargspec(error.Abort.__init__).args
     post_18415fc918a1 = "detailed_exit_code" in args
 
     unbundle_source = b'perf::unbundle'
     if opts[b'as_push']:
         unbundle_source = b'push'
 
     old_max_inline = None
     try:
         if not (pre_63edc384d3b7 or post_18415fc918a1):
             # disable inlining
             old_max_inline = mercurial.revlog._maxinline
             # large enough to never happen
             mercurial.revlog._maxinline = 2 ** 50
 
         with repo.lock():
             bundle = [None, None]
             orig_quiet = repo.ui.quiet
             try:
                 repo.ui.quiet = True
                 with open(fname, mode="rb") as f:
 
                     def noop_report(*args, **kwargs):
                         pass
 
                     def setup():
                         gen, tr = bundle
                         if tr is not None:
                             tr.abort()
                         bundle[:] = [None, None]
                         f.seek(0)
                         bundle[0] = exchange.readbundle(ui, f, fname)
                         bundle[1] = repo.transaction(b'perf::unbundle')
                         # silence the transaction
                         bundle[1]._report = noop_report
 
                     def apply():
                         gen, tr = bundle
                         bundle2.applybundle(
                             repo,
                             gen,
                             tr,
                             source=unbundle_source,
                             url=fname,
                         )
 
                     timer, fm = gettimer(ui, opts)
                     timer(apply, setup=setup)
                     fm.end()
             finally:
                 repo.ui.quiet == orig_quiet
                 gen, tr = bundle
                 if tr is not None:
                     tr.abort()
     finally:
         if old_max_inline is not None:
             mercurial.revlog._maxinline = old_max_inline
 
 
 @command(
     b'perf::unidiff|perfunidiff',
     revlogopts
     + formatteropts
     + [
         (
             b'',
             b'count',
             1,
             b'number of revisions to test (when using --startrev)',
         ),
         (b'', b'alldata', False, b'test unidiffs for all associated revisions'),
     ],
     b'-c|-m|FILE REV',
 )
 def perfunidiff(ui, repo, file_, rev=None, count=None, **opts):
     """benchmark a unified diff between revisions
 
     This doesn't include any copy tracing - it's just a unified diff
     of the texts.
 
     By default, benchmark a diff between its delta parent and itself.
 
     With ``--count``, benchmark diffs between delta parents and self for N
     revisions starting at the specified revision.
 
     With ``--alldata``, assume the requested revision is a changeset and
     measure diffs for all changes related to that changeset (manifest
     and filelogs).
     """
     opts = _byteskwargs(opts)
     if opts[b'alldata']:
         opts[b'changelog'] = True
 
     if opts.get(b'changelog') or opts.get(b'manifest'):
         file_, rev = None, file_
     elif rev is None:
         raise error.CommandError(b'perfunidiff', b'invalid arguments')
 
     textpairs = []
 
     r = cmdutil.openrevlog(repo, b'perfunidiff', file_, opts)
 
     startrev = r.rev(r.lookup(rev))
     for rev in range(startrev, min(startrev + count, len(r) - 1)):
         if opts[b'alldata']:
             # Load revisions associated with changeset.
             ctx = repo[rev]
             mtext = _manifestrevision(repo, ctx.manifestnode())
             for pctx in ctx.parents():
                 pman = _manifestrevision(repo, pctx.manifestnode())
                 textpairs.append((pman, mtext))
 
             # Load filelog revisions by iterating manifest delta.
             man = ctx.manifest()
             pman = ctx.p1().manifest()
             for filename, change in pman.diff(man).items():
                 fctx = repo.file(filename)
                 f1 = fctx.revision(change[0][0] or -1)
                 f2 = fctx.revision(change[1][0] or -1)
                 textpairs.append((f1, f2))
         else:
             dp = r.deltaparent(rev)
             textpairs.append((r.revision(dp), r.revision(rev)))
 
     def d():
         for left, right in textpairs:
             # The date strings don't matter, so we pass empty strings.
             headerlines, hunks = mdiff.unidiff(
                 left, b'', right, b'', b'left', b'right', binary=False
             )
             # consume iterators in roughly the way patch.py does
             b'\n'.join(headerlines)
             b''.join(sum((list(hlines) for hrange, hlines in hunks), []))
 
     timer, fm = gettimer(ui, opts)
     timer(d)
     fm.end()
 
 
 @command(b'perf::diffwd|perfdiffwd', formatteropts)
 def perfdiffwd(ui, repo, **opts):
     """Profile diff of working directory changes"""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     options = {
         'w': 'ignore_all_space',
         'b': 'ignore_space_change',
         'B': 'ignore_blank_lines',
     }
 
     for diffopt in ('', 'w', 'b', 'B', 'wB'):
         opts = {options[c]: b'1' for c in diffopt}
 
         def d():
             ui.pushbuffer()
             commands.diff(ui, repo, **opts)
             ui.popbuffer()
 
         diffopt = diffopt.encode('ascii')
         title = b'diffopts: %s' % (diffopt and (b'-' + diffopt) or b'none')
         timer(d, title=title)
     fm.end()
 
 
 @command(
     b'perf::revlogindex|perfrevlogindex',
     revlogopts + formatteropts,
     b'-c|-m|FILE',
 )
 def perfrevlogindex(ui, repo, file_=None, **opts):
     """Benchmark operations against a revlog index.
 
     This tests constructing a revlog instance, reading index data,
     parsing index data, and performing various operations related to
     index data.
     """
 
     opts = _byteskwargs(opts)
 
     rl = cmdutil.openrevlog(repo, b'perfrevlogindex', file_, opts)
 
     opener = getattr(rl, 'opener')  # trick linter
     # compat with hg <= 5.8
     radix = getattr(rl, 'radix', None)
     indexfile = getattr(rl, '_indexfile', None)
     if indexfile is None:
         # compatibility with <= hg-5.8
         indexfile = getattr(rl, 'indexfile')
     data = opener.read(indexfile)
 
     header = struct.unpack(b'>I', data[0:4])[0]
     version = header & 0xFFFF
     if version == 1:
         inline = header & (1 << 16)
     else:
         raise error.Abort(b'unsupported revlog version: %d' % version)
 
     parse_index_v1 = getattr(mercurial.revlog, 'parse_index_v1', None)
     if parse_index_v1 is None:
         parse_index_v1 = mercurial.revlog.revlogio().parseindex
 
     rllen = len(rl)
 
     node0 = rl.node(0)
     node25 = rl.node(rllen // 4)
     node50 = rl.node(rllen // 2)
     node75 = rl.node(rllen // 4 * 3)
     node100 = rl.node(rllen - 1)
 
     allrevs = range(rllen)
     allrevsrev = list(reversed(allrevs))
     allnodes = [rl.node(rev) for rev in range(rllen)]
     allnodesrev = list(reversed(allnodes))
 
     def constructor():
         if radix is not None:
             revlog(opener, radix=radix)
         else:
             # hg <= 5.8
             revlog(opener, indexfile=indexfile)
 
     def read():
         with opener(indexfile) as fh:
             fh.read()
 
     def parseindex():
         parse_index_v1(data, inline)
 
     def getentry(revornode):
         index = parse_index_v1(data, inline)[0]
         index[revornode]
 
     def getentries(revs, count=1):
         index = parse_index_v1(data, inline)[0]
 
         for i in range(count):
             for rev in revs:
                 index[rev]
 
     def resolvenode(node):
         index = parse_index_v1(data, inline)[0]
         rev = getattr(index, 'rev', None)
         if rev is None:
             nodemap = getattr(parse_index_v1(data, inline)[0], 'nodemap', None)
             # This only works for the C code.
             if nodemap is None:
                 return
             rev = nodemap.__getitem__
 
         try:
             rev(node)
         except error.RevlogError:
             pass
 
     def resolvenodes(nodes, count=1):
         index = parse_index_v1(data, inline)[0]
         rev = getattr(index, 'rev', None)
         if rev is None:
             nodemap = getattr(parse_index_v1(data, inline)[0], 'nodemap', None)
             # This only works for the C code.
             if nodemap is None:
                 return
             rev = nodemap.__getitem__
 
         for i in range(count):
             for node in nodes:
                 try:
                     rev(node)
                 except error.RevlogError:
                     pass
 
     benches = [
         (constructor, b'revlog constructor'),
         (read, b'read'),
         (parseindex, b'create index object'),
         (lambda: getentry(0), b'retrieve index entry for rev 0'),
         (lambda: resolvenode(b'a' * 20), b'look up missing node'),
         (lambda: resolvenode(node0), b'look up node at rev 0'),
         (lambda: resolvenode(node25), b'look up node at 1/4 len'),
         (lambda: resolvenode(node50), b'look up node at 1/2 len'),
         (lambda: resolvenode(node75), b'look up node at 3/4 len'),
         (lambda: resolvenode(node100), b'look up node at tip'),
         # 2x variation is to measure caching impact.
         (lambda: resolvenodes(allnodes), b'look up all nodes (forward)'),
         (lambda: resolvenodes(allnodes, 2), b'look up all nodes 2x (forward)'),
         (lambda: resolvenodes(allnodesrev), b'look up all nodes (reverse)'),
         (
             lambda: resolvenodes(allnodesrev, 2),
             b'look up all nodes 2x (reverse)',
         ),
         (lambda: getentries(allrevs), b'retrieve all index entries (forward)'),
         (
             lambda: getentries(allrevs, 2),
             b'retrieve all index entries 2x (forward)',
         ),
         (
             lambda: getentries(allrevsrev),
             b'retrieve all index entries (reverse)',
         ),
         (
             lambda: getentries(allrevsrev, 2),
             b'retrieve all index entries 2x (reverse)',
         ),
     ]
 
     for fn, title in benches:
         timer, fm = gettimer(ui, opts)
         timer(fn, title=title)
         fm.end()
 
 
 @command(
     b'perf::revlogrevisions|perfrevlogrevisions',
     revlogopts
     + formatteropts
     + [
         (b'd', b'dist', 100, b'distance between the revisions'),
         (b's', b'startrev', 0, b'revision to start reading at'),
         (b'', b'reverse', False, b'read in reverse'),
     ],
     b'-c|-m|FILE',
 )
 def perfrevlogrevisions(
     ui, repo, file_=None, startrev=0, reverse=False, **opts
 ):
     """Benchmark reading a series of revisions from a revlog.
 
     By default, we read every ``-d/--dist`` revision from 0 to tip of
     the specified revlog.
 
     The start revision can be defined via ``-s/--startrev``.
     """
     opts = _byteskwargs(opts)
 
     rl = cmdutil.openrevlog(repo, b'perfrevlogrevisions', file_, opts)
     rllen = getlen(ui)(rl)
 
     if startrev < 0:
         startrev = rllen + startrev
 
     def d():
         rl.clearcaches()
 
         beginrev = startrev
         endrev = rllen
         dist = opts[b'dist']
 
         if reverse:
             beginrev, endrev = endrev - 1, beginrev - 1
             dist = -1 * dist
 
         for x in _xrange(beginrev, endrev, dist):
             # Old revisions don't support passing int.
             n = rl.node(x)
             rl.revision(n)
 
     timer, fm = gettimer(ui, opts)
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::revlogwrite|perfrevlogwrite',
     revlogopts
     + formatteropts
     + [
         (b's', b'startrev', 1000, b'revision to start writing at'),
         (b'', b'stoprev', -1, b'last revision to write'),
         (b'', b'count', 3, b'number of passes to perform'),
         (b'', b'details', False, b'print timing for every revisions tested'),
         (b'', b'source', b'full', b'the kind of data feed in the revlog'),
         (b'', b'lazydeltabase', True, b'try the provided delta first'),
         (b'', b'clear-caches', True, b'clear revlog cache between calls'),
     ],
     b'-c|-m|FILE',
 )
 def perfrevlogwrite(ui, repo, file_=None, startrev=1000, stoprev=-1, **opts):
     """Benchmark writing a series of revisions to a revlog.
 
     Possible source values are:
     * `full`: add from a full text (default).
     * `parent-1`: add from a delta to the first parent
     * `parent-2`: add from a delta to the second parent if it exists
                   (use a delta from the first parent otherwise)
     * `parent-smallest`: add from the smallest delta (either p1 or p2)
     * `storage`: add from the existing precomputed deltas
 
     Note: This performance command measures performance in a custom way. As a
     result some of the global configuration of the 'perf' command does not
     apply to it:
 
     * ``pre-run``: disabled
 
     * ``profile-benchmark``: disabled
 
     * ``run-limits``: disabled use --count instead
     """
     opts = _byteskwargs(opts)
 
     rl = cmdutil.openrevlog(repo, b'perfrevlogwrite', file_, opts)
     rllen = getlen(ui)(rl)
     if startrev < 0:
         startrev = rllen + startrev
     if stoprev < 0:
         stoprev = rllen + stoprev
 
     lazydeltabase = opts['lazydeltabase']
     source = opts['source']
     clearcaches = opts['clear_caches']
     validsource = (
         b'full',
         b'parent-1',
         b'parent-2',
         b'parent-smallest',
         b'storage',
     )
     if source not in validsource:
         raise error.Abort('invalid source type: %s' % source)
 
     ### actually gather results
     count = opts['count']
     if count <= 0:
         raise error.Abort('invalide run count: %d' % count)
     allresults = []
     for c in range(count):
         timing = _timeonewrite(
             ui,
             rl,
             source,
             startrev,
             stoprev,
             c + 1,
             lazydeltabase=lazydeltabase,
             clearcaches=clearcaches,
         )
         allresults.append(timing)
 
     ### consolidate the results in a single list
     results = []
     for idx, (rev, t) in enumerate(allresults[0]):
         ts = [t]
         for other in allresults[1:]:
             orev, ot = other[idx]
             assert orev == rev
             ts.append(ot)
         results.append((rev, ts))
     resultcount = len(results)
 
     ### Compute and display relevant statistics
 
     # get a formatter
     fm = ui.formatter(b'perf', opts)
     displayall = ui.configbool(b"perf", b"all-timing", True)
 
     # print individual details if requested
     if opts['details']:
         for idx, item in enumerate(results, 1):
             rev, data = item
             title = 'revisions #%d of %d, rev %d' % (idx, resultcount, rev)
             formatone(fm, data, title=title, displayall=displayall)
 
     # sorts results by median time
     results.sort(key=lambda x: sorted(x[1])[len(x[1]) // 2])
     # list of (name, index) to display)
     relevants = [
         ("min", 0),
         ("10%", resultcount * 10 // 100),
         ("25%", resultcount * 25 // 100),
         ("50%", resultcount * 70 // 100),
         ("75%", resultcount * 75 // 100),
         ("90%", resultcount * 90 // 100),
         ("95%", resultcount * 95 // 100),
         ("99%", resultcount * 99 // 100),
         ("99.9%", resultcount * 999 // 1000),
         ("99.99%", resultcount * 9999 // 10000),
         ("99.999%", resultcount * 99999 // 100000),
         ("max", -1),
     ]
     if not ui.quiet:
         for name, idx in relevants:
             data = results[idx]
             title = '%s of %d, rev %d' % (name, resultcount, data[0])
             formatone(fm, data[1], title=title, displayall=displayall)
 
     # XXX summing that many float will not be very precise, we ignore this fact
     # for now
     totaltime = []
     for item in allresults:
         totaltime.append(
             (
                 sum(x[1][0] for x in item),
                 sum(x[1][1] for x in item),
                 sum(x[1][2] for x in item),
             )
         )
     formatone(
         fm,
         totaltime,
         title="total time (%d revs)" % resultcount,
         displayall=displayall,
     )
     fm.end()
 
 
 class _faketr:
     def add(s, x, y, z=None):
         return None
 
 
 def _timeonewrite(
     ui,
     orig,
     source,
     startrev,
     stoprev,
     runidx=None,
     lazydeltabase=True,
     clearcaches=True,
 ):
     timings = []
     tr = _faketr()
     with _temprevlog(ui, orig, startrev) as dest:
         if hasattr(dest, "delta_config"):
             dest.delta_config.lazy_delta_base = lazydeltabase
         else:
             dest._lazydeltabase = lazydeltabase
         revs = list(orig.revs(startrev, stoprev))
         total = len(revs)
         topic = 'adding'
         if runidx is not None:
             topic += ' (run #%d)' % runidx
         # Support both old and new progress API
         if util.safehasattr(ui, 'makeprogress'):
             progress = ui.makeprogress(topic, unit='revs', total=total)
 
             def updateprogress(pos):
                 progress.update(pos)
 
             def completeprogress():
                 progress.complete()
 
         else:
 
             def updateprogress(pos):
                 ui.progress(topic, pos, unit='revs', total=total)
 
             def completeprogress():
                 ui.progress(topic, None, unit='revs', total=total)
 
         for idx, rev in enumerate(revs):
             updateprogress(idx)
             addargs, addkwargs = _getrevisionseed(orig, rev, tr, source)
             if clearcaches:
                 dest.index.clearcaches()
                 dest.clearcaches()
             with timeone() as r:
                 dest.addrawrevision(*addargs, **addkwargs)
             timings.append((rev, r[0]))
         updateprogress(total)
         completeprogress()
     return timings
 
 
 def _getrevisionseed(orig, rev, tr, source):
     from mercurial.node import nullid
 
     linkrev = orig.linkrev(rev)
     node = orig.node(rev)
     p1, p2 = orig.parents(node)
     flags = orig.flags(rev)
     cachedelta = None
     text = None
 
     if source == b'full':
         text = orig.revision(rev)
     elif source == b'parent-1':
         baserev = orig.rev(p1)
         cachedelta = (baserev, orig.revdiff(p1, rev))
     elif source == b'parent-2':
         parent = p2
         if p2 == nullid:
             parent = p1
         baserev = orig.rev(parent)
         cachedelta = (baserev, orig.revdiff(parent, rev))
     elif source == b'parent-smallest':
         p1diff = orig.revdiff(p1, rev)
         parent = p1
         diff = p1diff
         if p2 != nullid:
             p2diff = orig.revdiff(p2, rev)
             if len(p1diff) > len(p2diff):
                 parent = p2
                 diff = p2diff
         baserev = orig.rev(parent)
         cachedelta = (baserev, diff)
     elif source == b'storage':
         baserev = orig.deltaparent(rev)
         cachedelta = (baserev, orig.revdiff(orig.node(baserev), rev))
 
     return (
         (text, tr, linkrev, p1, p2),
         {'node': node, 'flags': flags, 'cachedelta': cachedelta},
     )
 
 
 @contextlib.contextmanager
 def _temprevlog(ui, orig, truncaterev):
     from mercurial import vfs as vfsmod
 
     if orig._inline:
         raise error.Abort('not supporting inline revlog (yet)')
     revlogkwargs = {}
     k = 'upperboundcomp'
     if util.safehasattr(orig, k):
         revlogkwargs[k] = getattr(orig, k)
 
     indexfile = getattr(orig, '_indexfile', None)
     if indexfile is None:
         # compatibility with <= hg-5.8
         indexfile = getattr(orig, 'indexfile')
     origindexpath = orig.opener.join(indexfile)
 
     datafile = getattr(orig, '_datafile', getattr(orig, 'datafile'))
     origdatapath = orig.opener.join(datafile)
     radix = b'revlog'
     indexname = b'revlog.i'
     dataname = b'revlog.d'
 
     tmpdir = tempfile.mkdtemp(prefix='tmp-hgperf-')
     try:
         # copy the data file in a temporary directory
         ui.debug('copying data in %s\n' % tmpdir)
         destindexpath = os.path.join(tmpdir, 'revlog.i')
         destdatapath = os.path.join(tmpdir, 'revlog.d')
         shutil.copyfile(origindexpath, destindexpath)
         shutil.copyfile(origdatapath, destdatapath)
 
         # remove the data we want to add again
         ui.debug('truncating data to be rewritten\n')
         with open(destindexpath, 'ab') as index:
             index.seek(0)
             index.truncate(truncaterev * orig._io.size)
         with open(destdatapath, 'ab') as data:
             data.seek(0)
             data.truncate(orig.start(truncaterev))
 
         # instantiate a new revlog from the temporary copy
         ui.debug('truncating adding to be rewritten\n')
         vfs = vfsmod.vfs(tmpdir)
         vfs.options = getattr(orig.opener, 'options', None)
 
         try:
             dest = revlog(vfs, radix=radix, **revlogkwargs)
         except TypeError:
             dest = revlog(
                 vfs, indexfile=indexname, datafile=dataname, **revlogkwargs
             )
         if dest._inline:
             raise error.Abort('not supporting inline revlog (yet)')
         # make sure internals are initialized
         dest.revision(len(dest) - 1)
         yield dest
         del dest, vfs
     finally:
         shutil.rmtree(tmpdir, True)
 
 
 @command(
     b'perf::revlogchunks|perfrevlogchunks',
     revlogopts
     + formatteropts
     + [
         (b'e', b'engines', b'', b'compression engines to use'),
         (b's', b'startrev', 0, b'revision to start at'),
     ],
     b'-c|-m|FILE',
 )
 def perfrevlogchunks(ui, repo, file_=None, engines=None, startrev=0, **opts):
     """Benchmark operations on revlog chunks.
 
     Logically, each revlog is a collection of fulltext revisions. However,
     stored within each revlog are "chunks" of possibly compressed data. This
     data needs to be read and decompressed or compressed and written.
 
     This command measures the time it takes to read+decompress and recompress
     chunks in a revlog. It effectively isolates I/O and compression performance.
     For measurements of higher-level operations like resolving revisions,
     see ``perfrevlogrevisions`` and ``perfrevlogrevision``.
     """
     opts = _byteskwargs(opts)
 
     rl = cmdutil.openrevlog(repo, b'perfrevlogchunks', file_, opts)
 
     # - _chunkraw was renamed to _getsegmentforrevs
     # - _getsegmentforrevs was moved on the inner object
     try:
         segmentforrevs = rl._inner.get_segment_for_revs
     except AttributeError:
         try:
             segmentforrevs = rl._getsegmentforrevs
         except AttributeError:
             segmentforrevs = rl._chunkraw
 
     # Verify engines argument.
     if engines:
         engines = {e.strip() for e in engines.split(b',')}
         for engine in engines:
             try:
                 util.compressionengines[engine]
             except KeyError:
                 raise error.Abort(b'unknown compression engine: %s' % engine)
     else:
         engines = []
         for e in util.compengines:
             engine = util.compengines[e]
             try:
                 if engine.available():
                     engine.revlogcompressor().compress(b'dummy')
                     engines.append(e)
             except NotImplementedError:
                 pass
 
     revs = list(rl.revs(startrev, len(rl) - 1))
 
     @contextlib.contextmanager
     def reading(rl):
         if getattr(rl, 'reading', None) is not None:
             with rl.reading():
                 yield None
         elif rl._inline:
             indexfile = getattr(rl, '_indexfile', None)
             if indexfile is None:
                 # compatibility with <= hg-5.8
                 indexfile = getattr(rl, 'indexfile')
             yield getsvfs(repo)(indexfile)
         else:
             datafile = getattr(rl, 'datafile', getattr(rl, 'datafile'))
             yield getsvfs(repo)(datafile)
 
     if getattr(rl, 'reading', None) is not None:
 
         @contextlib.contextmanager
         def lazy_reading(rl):
             with rl.reading():
                 yield
 
     else:
 
         @contextlib.contextmanager
         def lazy_reading(rl):
             yield
 
     def doread():
         rl.clearcaches()
         for rev in revs:
             with lazy_reading(rl):
                 segmentforrevs(rev, rev)
 
     def doreadcachedfh():
         rl.clearcaches()
         with reading(rl) as fh:
             if fh is not None:
                 for rev in revs:
                     segmentforrevs(rev, rev, df=fh)
             else:
                 for rev in revs:
                     segmentforrevs(rev, rev)
 
     def doreadbatch():
         rl.clearcaches()
         with lazy_reading(rl):
             segmentforrevs(revs[0], revs[-1])
 
     def doreadbatchcachedfh():
         rl.clearcaches()
         with reading(rl) as fh:
             if fh is not None:
                 segmentforrevs(revs[0], revs[-1], df=fh)
             else:
                 segmentforrevs(revs[0], revs[-1])
 
     def dochunk():
         rl.clearcaches()
         # chunk used to be available directly on the revlog
         _chunk = getattr(rl, '_inner', rl)._chunk
         with reading(rl) as fh:
             if fh is not None:
                 for rev in revs:
                     _chunk(rev, df=fh)
             else:
                 for rev in revs:
                     _chunk(rev)
 
     chunks = [None]
 
     def dochunkbatch():
         rl.clearcaches()
         _chunks = getattr(rl, '_inner', rl)._chunks
         with reading(rl) as fh:
             if fh is not None:
                 # Save chunks as a side-effect.
                 chunks[0] = _chunks(revs, df=fh)
             else:
                 # Save chunks as a side-effect.
                 chunks[0] = _chunks(revs)
 
     def docompress(compressor):
         rl.clearcaches()
 
         compressor_holder = getattr(rl, '_inner', rl)
 
         try:
             # Swap in the requested compression engine.
             oldcompressor = compressor_holder._compressor
             compressor_holder._compressor = compressor
             for chunk in chunks[0]:
                 rl.compress(chunk)
         finally:
             compressor_holder._compressor = oldcompressor
 
     benches = [
         (lambda: doread(), b'read'),
         (lambda: doreadcachedfh(), b'read w/ reused fd'),
         (lambda: doreadbatch(), b'read batch'),
         (lambda: doreadbatchcachedfh(), b'read batch w/ reused fd'),
         (lambda: dochunk(), b'chunk'),
         (lambda: dochunkbatch(), b'chunk batch'),
     ]
 
     for engine in sorted(engines):
         compressor = util.compengines[engine].revlogcompressor()
         benches.append(
             (
                 functools.partial(docompress, compressor),
                 b'compress w/ %s' % engine,
             )
         )
 
     for fn, title in benches:
         timer, fm = gettimer(ui, opts)
         timer(fn, title=title)
         fm.end()
 
 
 @command(
     b'perf::revlogrevision|perfrevlogrevision',
     revlogopts
     + formatteropts
     + [(b'', b'cache', False, b'use caches instead of clearing')],
     b'-c|-m|FILE REV',
 )
 def perfrevlogrevision(ui, repo, file_, rev=None, cache=None, **opts):
     """Benchmark obtaining a revlog revision.
 
     Obtaining a revlog revision consists of roughly the following steps:
 
     1. Compute the delta chain
     2. Slice the delta chain if applicable
     3. Obtain the raw chunks for that delta chain
     4. Decompress each raw chunk
     5. Apply binary patches to obtain fulltext
     6. Verify hash of fulltext
 
     This command measures the time spent in each of these phases.
     """
     opts = _byteskwargs(opts)
 
     if opts.get(b'changelog') or opts.get(b'manifest'):
         file_, rev = None, file_
     elif rev is None:
         raise error.CommandError(b'perfrevlogrevision', b'invalid arguments')
 
     r = cmdutil.openrevlog(repo, b'perfrevlogrevision', file_, opts)
 
     # _chunkraw was renamed to _getsegmentforrevs.
     try:
         segmentforrevs = r._inner.get_segment_for_revs
     except AttributeError:
         try:
             segmentforrevs = r._getsegmentforrevs
         except AttributeError:
             segmentforrevs = r._chunkraw
 
     node = r.lookup(rev)
     rev = r.rev(node)
 
     if getattr(r, 'reading', None) is not None:
 
         @contextlib.contextmanager
         def lazy_reading(r):
             with r.reading():
                 yield
 
     else:
 
         @contextlib.contextmanager
         def lazy_reading(r):
             yield
 
     def getrawchunks(data, chain):
         start = r.start
         length = r.length
         inline = r._inline
         try:
             iosize = r.index.entry_size
         except AttributeError:
             iosize = r._io.size
         buffer = util.buffer
 
         chunks = []
         ladd = chunks.append
         for idx, item in enumerate(chain):
             offset = start(item[0])
             bits = data[idx]
             for rev in item:
                 chunkstart = start(rev)
                 if inline:
                     chunkstart += (rev + 1) * iosize
                 chunklength = length(rev)
                 ladd(buffer(bits, chunkstart - offset, chunklength))
 
         return chunks
 
     def dodeltachain(rev):
         if not cache:
             r.clearcaches()
         r._deltachain(rev)
 
     def doread(chain):
         if not cache:
             r.clearcaches()
         for item in slicedchain:
             with lazy_reading(r):
                 segmentforrevs(item[0], item[-1])
 
     def doslice(r, chain, size):
         for s in slicechunk(r, chain, targetsize=size):
             pass
 
     def dorawchunks(data, chain):
         if not cache:
             r.clearcaches()
         getrawchunks(data, chain)
 
     def dodecompress(chunks):
         decomp = r.decompress
         for chunk in chunks:
             decomp(chunk)
 
     def dopatch(text, bins):
         if not cache:
             r.clearcaches()
         mdiff.patches(text, bins)
 
     def dohash(text):
         if not cache:
             r.clearcaches()
         r.checkhash(text, node, rev=rev)
 
     def dorevision():
         if not cache:
             r.clearcaches()
         r.revision(node)
 
     try:
         from mercurial.revlogutils.deltas import slicechunk
     except ImportError:
         slicechunk = getattr(revlog, '_slicechunk', None)
 
     size = r.length(rev)
     chain = r._deltachain(rev)[0]
 
     with_sparse_read = False
     if hasattr(r, 'data_config'):
         with_sparse_read = r.data_config.with_sparse_read
     elif hasattr(r, '_withsparseread'):
         with_sparse_read = r._withsparseread
     if with_sparse_read:
         slicedchain = (chain,)
     else:
         slicedchain = tuple(slicechunk(r, chain, targetsize=size))
     data = [segmentforrevs(seg[0], seg[-1])[1] for seg in slicedchain]
     rawchunks = getrawchunks(data, slicedchain)
     bins = r._inner._chunks(chain)
     text = bytes(bins[0])
     bins = bins[1:]
     text = mdiff.patches(text, bins)
 
     benches = [
         (lambda: dorevision(), b'full'),
         (lambda: dodeltachain(rev), b'deltachain'),
         (lambda: doread(chain), b'read'),
     ]
 
     if with_sparse_read:
         slicing = (lambda: doslice(r, chain, size), b'slice-sparse-chain')
         benches.append(slicing)
 
     benches.extend(
         [
             (lambda: dorawchunks(data, slicedchain), b'rawchunks'),
             (lambda: dodecompress(rawchunks), b'decompress'),
             (lambda: dopatch(text, bins), b'patch'),
             (lambda: dohash(text), b'hash'),
         ]
     )
 
     timer, fm = gettimer(ui, opts)
     for fn, title in benches:
         timer(fn, title=title)
     fm.end()
 
 
 @command(
     b'perf::revset|perfrevset',
     [
         (b'C', b'clear', False, b'clear volatile cache between each call.'),
         (b'', b'contexts', False, b'obtain changectx for each revision'),
     ]
     + formatteropts,
     b"REVSET",
 )
 def perfrevset(ui, repo, expr, clear=False, contexts=False, **opts):
     """benchmark the execution time of a revset
 
     Use the --clean option if need to evaluate the impact of build volatile
     revisions set cache on the revset execution. Volatile cache hold filtered
     and obsolete related cache."""
     opts = _byteskwargs(opts)
 
     timer, fm = gettimer(ui, opts)
 
     def d():
         if clear:
             repo.invalidatevolatilesets()
         if contexts:
             for ctx in repo.set(expr):
                 pass
         else:
             for r in repo.revs(expr):
                 pass
 
     timer(d)
     fm.end()
 
 
 @command(
     b'perf::volatilesets|perfvolatilesets',
     [
         (b'', b'clear-obsstore', False, b'drop obsstore between each call.'),
     ]
     + formatteropts,
 )
 def perfvolatilesets(ui, repo, *names, **opts):
     """benchmark the computation of various volatile set
 
     Volatile set computes element related to filtering and obsolescence."""
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     repo = repo.unfiltered()
 
     def getobs(name):
         def d():
             repo.invalidatevolatilesets()
             if opts[b'clear_obsstore']:
                 clearfilecache(repo, b'obsstore')
             obsolete.getrevs(repo, name)
 
         return d
 
     allobs = sorted(obsolete.cachefuncs)
     if names:
         allobs = [n for n in allobs if n in names]
 
     for name in allobs:
         timer(getobs(name), title=name)
 
     def getfiltered(name):
         def d():
             repo.invalidatevolatilesets()
             if opts[b'clear_obsstore']:
                 clearfilecache(repo, b'obsstore')
             repoview.filterrevs(repo, name)
 
         return d
 
     allfilter = sorted(repoview.filtertable)
     if names:
         allfilter = [n for n in allfilter if n in names]
 
     for name in allfilter:
         timer(getfiltered(name), title=name)
     fm.end()
 
 
 @command(
     b'perf::branchmap|perfbranchmap',
     [
         (b'f', b'full', False, b'Includes build time of subset'),
         (
             b'',
             b'clear-revbranch',
             False,
             b'purge the revbranch cache between computation',
         ),
     ]
     + formatteropts,
 )
 def perfbranchmap(ui, repo, *filternames, **opts):
     """benchmark the update of a branchmap
 
     This benchmarks the full repo.branchmap() call with read and write disabled
     """
     opts = _byteskwargs(opts)
     full = opts.get(b"full", False)
     clear_revbranch = opts.get(b"clear_revbranch", False)
     timer, fm = gettimer(ui, opts)
 
     def getbranchmap(filtername):
         """generate a benchmark function for the filtername"""
         if filtername is None:
             view = repo
         else:
             view = repo.filtered(filtername)
         if util.safehasattr(view._branchcaches, '_per_filter'):
             filtered = view._branchcaches._per_filter
         else:
             # older versions
             filtered = view._branchcaches
 
         def d():
             if clear_revbranch:
                 repo.revbranchcache()._clear()
             if full:
                 view._branchcaches.clear()
             else:
                 filtered.pop(filtername, None)
             view.branchmap()
 
         return d
 
     # add filter in smaller subset to bigger subset
     possiblefilters = set(repoview.filtertable)
     if filternames:
         possiblefilters &= set(filternames)
     subsettable = getbranchmapsubsettable()
     allfilters = []
     while possiblefilters:
         for name in possiblefilters:
             subset = subsettable.get(name)
             if subset not in possiblefilters:
                 break
         else:
             assert False, b'subset cycle %s!' % possiblefilters
         allfilters.append(name)
         possiblefilters.remove(name)
 
     # warm the cache
     if not full:
         for name in allfilters:
             repo.filtered(name).branchmap()
     if not filternames or b'unfiltered' in filternames:
         # add unfiltered
         allfilters.append(None)
 
-    if util.safehasattr(branchmap.branchcache, 'fromfile'):
+    old_branch_cache_from_file = None
+    branchcacheread = None
+    if util.safehasattr(branchmap, 'branch_cache_from_file'):
+        old_branch_cache_from_file = branchmap.branch_cache_from_file
+        branchmap.branch_cache_from_file = lambda *args: None
+    elif util.safehasattr(branchmap.branchcache, 'fromfile'):
         branchcacheread = safeattrsetter(branchmap.branchcache, b'fromfile')
         branchcacheread.set(classmethod(lambda *args: None))
     else:
         # older versions
         branchcacheread = safeattrsetter(branchmap, b'read')
         branchcacheread.set(lambda *args: None)
+    if util.safehasattr(branchmap, '_LocalBranchCache'):
+        branchcachewrite = safeattrsetter(branchmap._LocalBranchCache, b'write')
+        branchcachewrite.set(lambda *args: None)
+    else:
     branchcachewrite = safeattrsetter(branchmap.branchcache, b'write')
     branchcachewrite.set(lambda *args: None)
     try:
         for name in allfilters:
             printname = name
             if name is None:
                 printname = b'unfiltered'
             timer(getbranchmap(name), title=printname)
     finally:
+        if old_branch_cache_from_file is not None:
+            branchmap.branch_cache_from_file = old_branch_cache_from_file
+        if branchcacheread is not None:
         branchcacheread.restore()
         branchcachewrite.restore()
     fm.end()
 
 
 @command(
     b'perf::branchmapupdate|perfbranchmapupdate',
     [
         (b'', b'base', [], b'subset of revision to start from'),
         (b'', b'target', [], b'subset of revision to end with'),
         (b'', b'clear-caches', False, b'clear cache between each runs'),
     ]
     + formatteropts,
 )
 def perfbranchmapupdate(ui, repo, base=(), target=(), **opts):
     """benchmark branchmap update from for <base> revs to <target> revs
 
     If `--clear-caches` is passed, the following items will be reset before
     each update:
         * the changelog instance and associated indexes
         * the rev-branch-cache instance
 
     Examples:
 
        # update for the one last revision
        $ hg perfbranchmapupdate --base 'not tip' --target 'tip'
 
        $ update for change coming with a new branch
        $ hg perfbranchmapupdate --base 'stable' --target 'default'
     """
     from mercurial import branchmap
     from mercurial import repoview
 
     opts = _byteskwargs(opts)
     timer, fm = gettimer(ui, opts)
     clearcaches = opts[b'clear_caches']
     unfi = repo.unfiltered()
     x = [None]  # used to pass data between closure
 
     # we use a `list` here to avoid possible side effect from smartset
     baserevs = list(scmutil.revrange(repo, base))
     targetrevs = list(scmutil.revrange(repo, target))
     if not baserevs:
         raise error.Abort(b'no revisions selected for --base')
     if not targetrevs:
         raise error.Abort(b'no revisions selected for --target')
 
     # make sure the target branchmap also contains the one in the base
     targetrevs = list(set(baserevs) | set(targetrevs))
     targetrevs.sort()
 
     cl = repo.changelog
     allbaserevs = list(cl.ancestors(baserevs, inclusive=True))
     allbaserevs.sort()
     alltargetrevs = frozenset(cl.ancestors(targetrevs, inclusive=True))
 
     newrevs = list(alltargetrevs.difference(allbaserevs))
     newrevs.sort()
 
     allrevs = frozenset(unfi.changelog.revs())
     basefilterrevs = frozenset(allrevs.difference(allbaserevs))
     targetfilterrevs = frozenset(allrevs.difference(alltargetrevs))
 
     def basefilter(repo, visibilityexceptions=None):
         return basefilterrevs
 
     def targetfilter(repo, visibilityexceptions=None):
         return targetfilterrevs
 
     msg = b'benchmark of branchmap with %d revisions with %d new ones\n'
     ui.status(msg % (len(allbaserevs), len(newrevs)))
     if targetfilterrevs:
         msg = b'(%d revisions still filtered)\n'
         ui.status(msg % len(targetfilterrevs))
 
     try:
         repoview.filtertable[b'__perf_branchmap_update_base'] = basefilter
         repoview.filtertable[b'__perf_branchmap_update_target'] = targetfilter
 
         baserepo = repo.filtered(b'__perf_branchmap_update_base')
         targetrepo = repo.filtered(b'__perf_branchmap_update_target')
 
+        bcache = repo.branchmap()
+        copy_method = 'copy'
+
+        copy_base_kwargs = copy_base_kwargs = {}
+        if hasattr(bcache, 'copy'):
+            if 'repo' in getargspec(bcache.copy).args:
+                copy_base_kwargs = {"repo": baserepo}
+                copy_target_kwargs = {"repo": targetrepo}
+        else:
+            copy_method = 'inherit_for'
+            copy_base_kwargs = {"repo": baserepo}
+            copy_target_kwargs = {"repo": targetrepo}
+
         # try to find an existing branchmap to reuse
         subsettable = getbranchmapsubsettable()
         candidatefilter = subsettable.get(None)
         while candidatefilter is not None:
             candidatebm = repo.filtered(candidatefilter).branchmap()
             if candidatebm.validfor(baserepo):
                 filtered = repoview.filterrevs(repo, candidatefilter)
                 missing = [r for r in allbaserevs if r in filtered]
-                base = candidatebm.copy()
+                base = getattr(candidatebm, copy_method)(**copy_base_kwargs)
                 base.update(baserepo, missing)
                 break
             candidatefilter = subsettable.get(candidatefilter)
         else:
             # no suitable subset where found
             base = branchmap.branchcache()
             base.update(baserepo, allbaserevs)
 
         def setup():
-            x[0] = base.copy()
+            x[0] = getattr(base, copy_method)(**copy_target_kwargs)
             if clearcaches:
                 unfi._revbranchcache = None
                 clearchangelog(repo)
 
         def bench():
             x[0].update(targetrepo, newrevs)
 
         timer(bench, setup=setup)
         fm.end()
     finally:
         repoview.filtertable.pop(b'__perf_branchmap_update_base', None)
         repoview.filtertable.pop(b'__perf_branchmap_update_target', None)
 
 
 @command(
     b'perf::branchmapload|perfbranchmapload',
     [
         (b'f', b'filter', b'', b'Specify repoview filter'),
         (b'', b'list', False, b'List brachmap filter caches'),
         (b'', b'clear-revlogs', False, b'refresh changelog and manifest'),
     ]
     + formatteropts,
 )
 def perfbranchmapload(ui, repo, filter=b'', list=False, **opts):
     """benchmark reading the branchmap"""
     opts = _byteskwargs(opts)
     clearrevlogs = opts[b'clear_revlogs']
 
     if list:
         for name, kind, st in repo.cachevfs.readdir(stat=True):
             if name.startswith(b'branch2'):
                 filtername = name.partition(b'-')[2] or b'unfiltered'
                 ui.status(
                     b'%s - %s\n' % (filtername, util.bytecount(st.st_size))
                 )
         return
     if not filter:
         filter = None
     subsettable = getbranchmapsubsettable()
     if filter is None:
         repo = repo.unfiltered()
     else:
         repo = repoview.repoview(repo, filter)
 
     repo.branchmap()  # make sure we have a relevant, up to date branchmap
 
-    try:
-        fromfile = branchmap.branchcache.fromfile
-    except AttributeError:
-        # older versions
+    fromfile = getattr(branchmap, 'branch_cache_from_file', None)
+    if fromfile is None:
+        fromfile = getattr(branchmap.branchcache, 'fromfile', None)
+    if fromfile is None:
         fromfile = branchmap.read
 
     currentfilter = filter
     # try once without timer, the filter may not be cached
     while fromfile(repo) is None:
         currentfilter = subsettable.get(currentfilter)
         if currentfilter is None:
             raise error.Abort(
                 b'No branchmap cached for %s repo' % (filter or b'unfiltered')
             )
         repo = repo.filtered(currentfilter)
     timer, fm = gettimer(ui, opts)
 
     def setup():
         if clearrevlogs:
             clearchangelog(repo)
 
     def bench():
         fromfile(repo)
 
     timer(bench, setup=setup)
     fm.end()
 
 
 @command(b'perf::loadmarkers|perfloadmarkers')
 def perfloadmarkers(ui, repo):
     """benchmark the time to parse the on-disk markers for a repo
 
     Result is the number of markers in the repo."""
     timer, fm = gettimer(ui)
     svfs = getsvfs(repo)
     timer(lambda: len(obsolete.obsstore(repo, svfs)))
     fm.end()
 
 
 @command(
     b'perf::lrucachedict|perflrucachedict',
     formatteropts
     + [
         (b'', b'costlimit', 0, b'maximum total cost of items in cache'),
         (b'', b'mincost', 0, b'smallest cost of items in cache'),
         (b'', b'maxcost', 100, b'maximum cost of items in cache'),
         (b'', b'size', 4, b'size of cache'),
         (b'', b'gets', 10000, b'number of key lookups'),
         (b'', b'sets', 10000, b'number of key sets'),
         (b'', b'mixed', 10000, b'number of mixed mode operations'),
         (
             b'',
             b'mixedgetfreq',
             50,
             b'frequency of get vs set ops in mixed mode',
         ),
     ],
     norepo=True,
 )
 def perflrucache(
     ui,
     mincost=0,
     maxcost=100,
     costlimit=0,
     size=4,
     gets=10000,
     sets=10000,
     mixed=10000,
     mixedgetfreq=50,
     **opts
 ):
     opts = _byteskwargs(opts)
 
     def doinit():
         for i in _xrange(10000):
             util.lrucachedict(size)
 
     costrange = list(range(mincost, maxcost + 1))
 
     values = []
     for i in _xrange(size):
         values.append(random.randint(0, _maxint))
 
     # Get mode fills the cache and tests raw lookup performance with no
     # eviction.
     getseq = []
     for i in _xrange(gets):
         getseq.append(random.choice(values))
 
     def dogets():
         d = util.lrucachedict(size)
         for v in values:
             d[v] = v
         for key in getseq:
             value = d[key]
             value  # silence pyflakes warning
 
     def dogetscost():
         d = util.lrucachedict(size, maxcost=costlimit)
         for i, v in enumerate(values):
             d.insert(v, v, cost=costs[i])
         for key in getseq:
             try:
                 value = d[key]
                 value  # silence pyflakes warning
             except KeyError:
                 pass
 
     # Set mode tests insertion speed with cache eviction.
     setseq = []
     costs = []
     for i in _xrange(sets):
         setseq.append(random.randint(0, _maxint))
         costs.append(random.choice(costrange))
 
     def doinserts():
         d = util.lrucachedict(size)
         for v in setseq:
             d.insert(v, v)
 
     def doinsertscost():
         d = util.lrucachedict(size, maxcost=costlimit)
         for i, v in enumerate(setseq):
             d.insert(v, v, cost=costs[i])
 
     def dosets():
         d = util.lrucachedict(size)
         for v in setseq:
             d[v] = v
 
     # Mixed mode randomly performs gets and sets with eviction.
     mixedops = []
     for i in _xrange(mixed):
         r = random.randint(0, 100)
         if r < mixedgetfreq:
             op = 0
         else:
             op = 1
 
         mixedops.append(
             (op, random.randint(0, size * 2), random.choice(costrange))
         )
 
     def domixed():
         d = util.lrucachedict(size)
 
         for op, v, cost in mixedops:
             if op == 0:
                 try:
                     d[v]
                 except KeyError:
                     pass
             else:
                 d[v] = v
 
     def domixedcost():
         d = util.lrucachedict(size, maxcost=costlimit)
 
         for op, v, cost in mixedops:
             if op == 0:
                 try:
                     d[v]
                 except KeyError:
                     pass
             else:
                 d.insert(v, v, cost=cost)
 
     benches = [
         (doinit, b'init'),
     ]
 
     if costlimit:
         benches.extend(
             [
                 (dogetscost, b'gets w/ cost limit'),
                 (doinsertscost, b'inserts w/ cost limit'),
                 (domixedcost, b'mixed w/ cost limit'),
             ]
         )
     else:
         benches.extend(
             [
                 (dogets, b'gets'),
                 (doinserts, b'inserts'),
                 (dosets, b'sets'),
                 (domixed, b'mixed'),
             ]
         )
 
     for fn, title in benches:
         timer, fm = gettimer(ui, opts)
         timer(fn, title=title)
         fm.end()
 
 
 @command(
     b'perf::write|perfwrite',
     formatteropts
     + [
         (b'', b'write-method', b'write', b'ui write method'),
         (b'', b'nlines', 100, b'number of lines'),
         (b'', b'nitems', 100, b'number of items (per line)'),
         (b'', b'item', b'x', b'item that is written'),
         (b'', b'batch-line', None, b'pass whole line to write method at once'),
         (b'', b'flush-line', None, b'flush after each line'),
     ],
 )
 def perfwrite(ui, repo, **opts):
     """microbenchmark ui.write (and others)"""
     opts = _byteskwargs(opts)
 
     write = getattr(ui, _sysstr(opts[b'write_method']))
     nlines = int(opts[b'nlines'])
     nitems = int(opts[b'nitems'])
     item = opts[b'item']
     batch_line = opts.get(b'batch_line')
     flush_line = opts.get(b'flush_line')
 
     if batch_line:
         line = item * nitems + b'\n'
 
     def benchmark():
         for i in pycompat.xrange(nlines):
             if batch_line:
                 write(line)
             else:
                 for i in pycompat.xrange(nitems):
                     write(item)
                 write(b'\n')
             if flush_line:
                 ui.flush()
         ui.flush()
 
     timer, fm = gettimer(ui, opts)
     timer(benchmark)
     fm.end()
 
 
 def uisetup(ui):
     if util.safehasattr(cmdutil, b'openrevlog') and not util.safehasattr(
         commands, b'debugrevlogopts'
     ):
         # for "historical portability":
         # In this case, Mercurial should be 1.9 (or a79fea6b3e77) -
         # 3.7 (or 5606f7d0d063). Therefore, '--dir' option for
         # openrevlog() should cause failure, because it has been
         # available since 3.5 (or 49c583ca48c4).
         def openrevlog(orig, repo, cmd, file_, opts):
             if opts.get(b'dir') and not util.safehasattr(repo, b'dirlog'):
                 raise error.Abort(
                     b"This version doesn't support --dir option",
                     hint=b"use 3.5 or later",
                 )
             return orig(repo, cmd, file_, opts)
 
         name = _sysstr(b'openrevlog')
         extensions.wrapfunction(cmdutil, name, openrevlog)
 
 
 @command(
     b'perf::progress|perfprogress',
     formatteropts
     + [
         (b'', b'topic', b'topic', b'topic for progress messages'),
         (b'c', b'total', 1000000, b'total value we are progressing to'),
     ],
     norepo=True,
 )
 def perfprogress(ui, topic=None, total=None, **opts):
     """printing of progress bars"""
     opts = _byteskwargs(opts)
 
     timer, fm = gettimer(ui, opts)
 
     def doprogress():
         with ui.makeprogress(topic, total=total) as progress:
             for i in _xrange(total):
                 progress.increment()
 
     timer(doprogress)
     fm.end()

hgext/largefiles/lfutil.py

@@ -1,823 +1,826 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 contextlib
 import copy
 import os
 import stat
 
 from mercurial.i18n import _
 from mercurial.node import hex
 from mercurial.pycompat import open
 
 from mercurial import (
     dirstate,
     encoding,
     error,
     httpconnection,
     match as matchmod,
     pycompat,
     requirements,
     scmutil,
     sparse,
     util,
     vfs as vfsmod,
 )
 from mercurial.utils import hashutil
 from mercurial.dirstateutils import timestamp
 
 shortname = b'.hglf'
 shortnameslash = shortname + b'/'
 longname = b'largefiles'
 
 # -- Private worker functions ------------------------------------------
 
 
 @contextlib.contextmanager
 def lfstatus(repo, value=True):
     oldvalue = getattr(repo, 'lfstatus', False)
     repo.lfstatus = value
     try:
         yield
     finally:
         repo.lfstatus = oldvalue
 
 
 def getminsize(ui, assumelfiles, opt, default=10):
     lfsize = opt
     if not lfsize and assumelfiles:
         lfsize = ui.config(longname, b'minsize', default=default)
     if lfsize:
         try:
             lfsize = float(lfsize)
         except ValueError:
             raise error.Abort(
                 _(b'largefiles: size must be number (not %s)\n') % lfsize
             )
     if lfsize is None:
         raise error.Abort(_(b'minimum size for largefiles must be specified'))
     return lfsize
 
 
 def link(src, dest):
     """Try to create hardlink - if that fails, efficiently make a copy."""
     util.makedirs(os.path.dirname(dest))
     try:
         util.oslink(src, dest)
     except OSError:
         # if hardlinks fail, fallback on atomic copy
         with open(src, b'rb') as srcf, util.atomictempfile(dest) as dstf:
             for chunk in util.filechunkiter(srcf):
                 dstf.write(chunk)
         os.chmod(dest, os.stat(src).st_mode)
 
 
 def usercachepath(ui, hash):
     """Return the correct location in the "global" largefiles cache for a file
     with the given hash.
     This cache is used for sharing of largefiles across repositories - both
     to preserve download bandwidth and storage space."""
     return os.path.join(_usercachedir(ui), hash)
 
 
 def _usercachedir(ui, name=longname):
     '''Return the location of the "global" largefiles cache.'''
     path = ui.configpath(name, b'usercache')
     if path:
         return path
 
     hint = None
 
     if pycompat.iswindows:
         appdata = encoding.environ.get(
             b'LOCALAPPDATA', encoding.environ.get(b'APPDATA')
         )
         if appdata:
             return os.path.join(appdata, name)
 
         hint = _(b"define %s or %s in the environment, or set %s.usercache") % (
             b"LOCALAPPDATA",
             b"APPDATA",
             name,
         )
     elif pycompat.isdarwin:
         home = encoding.environ.get(b'HOME')
         if home:
             return os.path.join(home, b'Library', b'Caches', name)
 
         hint = _(b"define %s in the environment, or set %s.usercache") % (
             b"HOME",
             name,
         )
     elif pycompat.isposix:
         path = encoding.environ.get(b'XDG_CACHE_HOME')
         if path:
             return os.path.join(path, name)
         home = encoding.environ.get(b'HOME')
         if home:
             return os.path.join(home, b'.cache', name)
 
         hint = _(b"define %s or %s in the environment, or set %s.usercache") % (
             b"XDG_CACHE_HOME",
             b"HOME",
             name,
         )
     else:
         raise error.Abort(
             _(b'unknown operating system: %s\n') % pycompat.osname
         )
 
     raise error.Abort(_(b'unknown %s usercache location') % name, hint=hint)
 
 
 def inusercache(ui, hash):
     path = usercachepath(ui, hash)
     return os.path.exists(path)
 
 
 def findfile(repo, hash):
     """Return store path of the largefile with the specified hash.
     As a side effect, the file might be linked from user cache.
     Return None if the file can't be found locally."""
     path, exists = findstorepath(repo, hash)
     if exists:
         repo.ui.note(_(b'found %s in store\n') % hash)
         return path
     elif inusercache(repo.ui, hash):
         repo.ui.note(_(b'found %s in system cache\n') % hash)
         path = storepath(repo, hash)
         link(usercachepath(repo.ui, hash), path)
         return path
     return None
 
 
 class largefilesdirstate(dirstate.dirstate):
     _large_file_dirstate = True
     _tr_key_suffix = b'-large-files'
 
     def __getitem__(self, key):
         return super(largefilesdirstate, self).__getitem__(unixpath(key))
 
     def set_tracked(self, f):
         return super(largefilesdirstate, self).set_tracked(unixpath(f))
 
     def set_untracked(self, f):
         return super(largefilesdirstate, self).set_untracked(unixpath(f))
 
     def normal(self, f, parentfiledata=None):
         # not sure if we should pass the `parentfiledata` down or throw it
         # away. So throwing it away to stay on the safe side.
         return super(largefilesdirstate, self).normal(unixpath(f))
 
     def remove(self, f):
         return super(largefilesdirstate, self).remove(unixpath(f))
 
     def add(self, f):
         return super(largefilesdirstate, self).add(unixpath(f))
 
     def drop(self, f):
         return super(largefilesdirstate, self).drop(unixpath(f))
 
     def forget(self, f):
         return super(largefilesdirstate, self).forget(unixpath(f))
 
     def normallookup(self, f):
         return super(largefilesdirstate, self).normallookup(unixpath(f))
 
     def _ignore(self, f):
         return False
 
     def write(self, tr):
         # (1) disable PENDING mode always
         #     (lfdirstate isn't yet managed as a part of the transaction)
         # (2) avoid develwarn 'use dirstate.write with ....'
         if tr:
             tr.addbackup(b'largefiles/dirstate', location=b'plain')
         super(largefilesdirstate, self).write(None)
 
 
 def openlfdirstate(ui, repo, create=True):
     """
     Return a dirstate object that tracks largefiles: i.e. its root is
     the repo root, but it is saved in .hg/largefiles/dirstate.
 
     If a dirstate object already exists and is being used for a 'changing_*'
     context, it will be returned.
     """
     sub_dirstate = getattr(repo.dirstate, '_sub_dirstate', None)
     if sub_dirstate is not None:
         return sub_dirstate
     vfs = repo.vfs
     lfstoredir = longname
     opener = vfsmod.vfs(vfs.join(lfstoredir))
     use_dirstate_v2 = requirements.DIRSTATE_V2_REQUIREMENT in repo.requirements
     lfdirstate = largefilesdirstate(
         opener,
         ui,
         repo.root,
         repo.dirstate._validate,
         lambda: sparse.matcher(repo),
         repo.nodeconstants,
         use_dirstate_v2,
     )
 
     # If the largefiles dirstate does not exist, populate and create
     # it. This ensures that we create it on the first meaningful
     # largefiles operation in a new clone.
     if create and not vfs.exists(vfs.join(lfstoredir, b'dirstate')):
         try:
             with repo.wlock(wait=False), lfdirstate.changing_files(repo):
                 matcher = getstandinmatcher(repo)
                 standins = repo.dirstate.walk(
                     matcher, subrepos=[], unknown=False, ignored=False
                 )
 
                 if len(standins) > 0:
                     vfs.makedirs(lfstoredir)
 
                 for standin in standins:
                     lfile = splitstandin(standin)
                     lfdirstate.hacky_extension_update_file(
                         lfile,
                         p1_tracked=True,
                         wc_tracked=True,
                         possibly_dirty=True,
                     )
         except error.LockError:
             # Assume that whatever was holding the lock was important.
             # If we were doing something important, we would already have
             # either the lock or a largefile dirstate.
             pass
     return lfdirstate
 
 
 def lfdirstatestatus(lfdirstate, repo):
     pctx = repo[b'.']
     match = matchmod.always()
     unsure, s, mtime_boundary = lfdirstate.status(
         match, subrepos=[], ignored=False, clean=False, unknown=False
     )
     modified, clean = s.modified, s.clean
     wctx = repo[None]
     for lfile in unsure:
         try:
             fctx = pctx[standin(lfile)]
         except LookupError:
             fctx = None
         if not fctx or readasstandin(fctx) != hashfile(repo.wjoin(lfile)):
             modified.append(lfile)
         else:
             clean.append(lfile)
             st = wctx[lfile].lstat()
             mode = st.st_mode
             size = st.st_size
             mtime = timestamp.reliable_mtime_of(st, mtime_boundary)
             if mtime is not None:
                 cache_data = (mode, size, mtime)
                 lfdirstate.set_clean(lfile, cache_data)
     return s
 
 
 def listlfiles(repo, rev=None, matcher=None):
     """return a list of largefiles in the working copy or the
     specified changeset"""
 
     if matcher is None:
         matcher = getstandinmatcher(repo)
 
     # ignore unknown files in working directory
     return [
         splitstandin(f)
         for f in repo[rev].walk(matcher)
         if rev is not None or repo.dirstate.get_entry(f).any_tracked
     ]
 
 
 def instore(repo, hash, forcelocal=False):
     '''Return true if a largefile with the given hash exists in the store'''
     return os.path.exists(storepath(repo, hash, forcelocal))
 
 
 def storepath(repo, hash, forcelocal=False):
     """Return the correct location in the repository largefiles store for a
     file with the given hash."""
     if not forcelocal and repo.shared():
         return repo.vfs.reljoin(repo.sharedpath, longname, hash)
     return repo.vfs.join(longname, hash)
 
 
 def findstorepath(repo, hash):
     """Search through the local store path(s) to find the file for the given
     hash.  If the file is not found, its path in the primary store is returned.
     The return value is a tuple of (path, exists(path)).
     """
     # For shared repos, the primary store is in the share source.  But for
     # backward compatibility, force a lookup in the local store if it wasn't
     # found in the share source.
     path = storepath(repo, hash, False)
 
     if instore(repo, hash):
         return (path, True)
     elif repo.shared() and instore(repo, hash, True):
         return storepath(repo, hash, True), True
 
     return (path, False)
 
 
 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)."""
     wvfs = repo.wvfs
     path = findfile(repo, hash)
     if path is None:
         return False
     wvfs.makedirs(wvfs.dirname(wvfs.join(filename)))
     # The write may fail before the file is fully written, but we
     # don't use atomic writes in the working copy.
     with open(path, b'rb') as srcfd, wvfs(filename, b'wb') as destfd:
         gothash = copyandhash(util.filechunkiter(srcfd), destfd)
     if gothash != hash:
         repo.ui.warn(
             _(b'%s: data corruption in %s with hash %s\n')
             % (filename, path, gothash)
         )
         wvfs.unlink(filename)
         return False
     return True
 
 
 def copytostore(repo, ctx, file, fstandin):
     wvfs = repo.wvfs
     hash = readasstandin(ctx[fstandin])
     if instore(repo, hash):
         return
     if wvfs.exists(file):
         copytostoreabsolute(repo, wvfs.join(file), hash)
     else:
         repo.ui.warn(
             _(b"%s: largefile %s not available from local store\n")
             % (file, hash)
         )
 
 
 def copyalltostore(repo, node):
     '''Copy all largefiles in a given revision to the store'''
 
     ctx = repo[node]
     for filename in ctx.files():
         realfile = splitstandin(filename)
         if realfile is not None and filename in ctx.manifest():
             copytostore(repo, ctx, realfile, filename)
 
 
 def copytostoreabsolute(repo, file, hash):
     if inusercache(repo.ui, hash):
         link(usercachepath(repo.ui, hash), storepath(repo, hash))
     else:
         util.makedirs(os.path.dirname(storepath(repo, hash)))
         with open(file, b'rb') as srcf:
             with util.atomictempfile(
                 storepath(repo, hash), createmode=repo.store.createmode
             ) as dstf:
                 for chunk in util.filechunkiter(srcf):
                     dstf.write(chunk)
         linktousercache(repo, hash)
 
 
 def linktousercache(repo, hash):
     """Link / copy the largefile with the specified hash from the store
     to the cache."""
     path = usercachepath(repo.ui, hash)
     link(storepath(repo, hash), path)
 
 
 def getstandinmatcher(repo, rmatcher=None):
     '''Return a match object that applies rmatcher to the standin directory'''
     wvfs = repo.wvfs
     standindir = shortname
 
     # no warnings about missing files or directories
     badfn = lambda f, msg: None
 
     if rmatcher and not rmatcher.always():
         pats = [wvfs.join(standindir, pat) for pat in rmatcher.files()]
         if not pats:
             pats = [wvfs.join(standindir)]
         match = scmutil.match(repo[None], pats, badfn=badfn)
     else:
         # no patterns: relative to repo root
         match = scmutil.match(repo[None], [wvfs.join(standindir)], badfn=badfn)
     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)
     isstandin = smatcher.matchfn
 
     def composedmatchfn(f):
         return isstandin(f) and rmatcher.matchfn(splitstandin(f))
 
+    smatcher._was_tampered_with = True
     smatcher.matchfn = composedmatchfn
 
     return smatcher
 
 
 def standin(filename):
     """Return the repo-relative path to the standin for the specified big
     file."""
     # Notes:
     # 1) Some callers want an absolute path, but for instance addlargefiles
     #    needs it repo-relative so it can be passed to repo[None].add().  So
     #    leave it up to the caller to use repo.wjoin() to get an absolute path.
     # 2) Join with '/' because that's what dirstate always uses, even on
     #    Windows. Change existing separator to '/' first in case we are
     #    passed filenames from an external source (like the command line).
     return shortnameslash + util.pconvert(filename)
 
 
 def isstandin(filename):
     """Return true if filename is a big file standin. filename must be
     in Mercurial's internal form (slash-separated)."""
     return filename.startswith(shortnameslash)
 
 
 def splitstandin(filename):
     # Split on / because that's what dirstate always uses, even on Windows.
     # Change local separator to / first just in case we are passed filenames
     # from an external source (like the command line).
     bits = util.pconvert(filename).split(b'/', 1)
     if len(bits) == 2 and bits[0] == shortname:
         return bits[1]
     else:
         return None
 
 
 def updatestandin(repo, lfile, standin):
     """Re-calculate hash value of lfile and write it into standin
 
     This assumes that "lfutil.standin(lfile) == standin", for efficiency.
     """
     file = repo.wjoin(lfile)
     if repo.wvfs.exists(lfile):
         hash = hashfile(file)
         executable = getexecutable(file)
         writestandin(repo, standin, hash, executable)
     else:
         raise error.Abort(_(b'%s: file not found!') % lfile)
 
 
 def readasstandin(fctx):
     """read hex hash from given filectx of standin file
 
     This encapsulates how "standin" data is stored into storage layer."""
     return fctx.data().strip()
 
 
 def writestandin(repo, standin, hash, executable):
     '''write hash to <repo.root>/<standin>'''
     repo.wwrite(standin, hash + b'\n', executable and b'x' or b'')
 
 
 def copyandhash(instream, outfile):
     """Read bytes from instream (iterable) and write them to outfile,
     computing the SHA-1 hash of the data along the way. Return the hash."""
     hasher = hashutil.sha1(b'')
     for data in instream:
         hasher.update(data)
         outfile.write(data)
     return hex(hasher.digest())
 
 
 def hashfile(file):
     if not os.path.exists(file):
         return b''
     with open(file, b'rb') as fd:
         return hexsha1(fd)
 
 
 def getexecutable(filename):
     mode = os.stat(filename).st_mode
     return (
         (mode & stat.S_IXUSR)
         and (mode & stat.S_IXGRP)
         and (mode & stat.S_IXOTH)
     )
 
 
 def urljoin(first, second, *arg):
     def join(left, right):
         if not left.endswith(b'/'):
             left += b'/'
         if right.startswith(b'/'):
             right = right[1:]
         return left + right
 
     url = join(first, second)
     for a in arg:
         url = join(url, a)
     return url
 
 
 def hexsha1(fileobj):
     """hexsha1 returns the hex-encoded sha1 sum of the data in the file-like
     object data"""
     h = hashutil.sha1()
     for chunk in util.filechunkiter(fileobj):
         h.update(chunk)
     return hex(h.digest())
 
 
 def httpsendfile(ui, filename):
     return httpconnection.httpsendfile(ui, filename, b'rb')
 
 
 def unixpath(path):
     '''Return a version of path normalized for use with the lfdirstate.'''
     return util.pconvert(os.path.normpath(path))
 
 
 def islfilesrepo(repo):
     '''Return true if the repo is a largefile repo.'''
     if b'largefiles' in repo.requirements:
         for entry in repo.store.data_entries():
             if entry.is_revlog and shortnameslash in entry.target_id:
                 return True
 
     return any(openlfdirstate(repo.ui, repo, False))
 
 
 class storeprotonotcapable(Exception):
     def __init__(self, storetypes):
         self.storetypes = storetypes
 
 
 def getstandinsstate(repo):
     standins = []
     matcher = getstandinmatcher(repo)
     wctx = repo[None]
     for standin in repo.dirstate.walk(
         matcher, subrepos=[], unknown=False, ignored=False
     ):
         lfile = splitstandin(standin)
         try:
             hash = readasstandin(wctx[standin])
         except IOError:
             hash = None
         standins.append((lfile, hash))
     return standins
 
 
 def synclfdirstate(repo, lfdirstate, lfile, normallookup):
     lfstandin = standin(lfile)
     if lfstandin not in repo.dirstate:
         lfdirstate.hacky_extension_update_file(
             lfile,
             p1_tracked=False,
             wc_tracked=False,
         )
     else:
         entry = repo.dirstate.get_entry(lfstandin)
         lfdirstate.hacky_extension_update_file(
             lfile,
             wc_tracked=entry.tracked,
             p1_tracked=entry.p1_tracked,
             p2_info=entry.p2_info,
             possibly_dirty=True,
         )
 
 
 def markcommitted(orig, ctx, node):
     repo = ctx.repo()
 
     with repo.dirstate.changing_parents(repo):
         orig(node)
 
         # ATTENTION: "ctx.files()" may differ from "repo[node].files()"
         # because files coming from the 2nd parent are omitted in the latter.
         #
         # The former should be used to get targets of "synclfdirstate",
         # because such files:
         # - are marked as "a" by "patch.patch()" (e.g. via transplant), and
         # - have to be marked as "n" after commit, but
         # - aren't listed in "repo[node].files()"
 
         lfdirstate = openlfdirstate(repo.ui, repo)
         for f in ctx.files():
             lfile = splitstandin(f)
             if lfile is not None:
                 synclfdirstate(repo, lfdirstate, lfile, False)
 
     # As part of committing, copy all of the largefiles into the cache.
     #
     # Using "node" instead of "ctx" implies additional "repo[node]"
     # lookup while copyalltostore(), but can omit redundant check for
     # files comming from the 2nd parent, which should exist in store
     # at merging.
     copyalltostore(repo, node)
 
 
 def getlfilestoupdate(oldstandins, newstandins):
     changedstandins = set(oldstandins).symmetric_difference(set(newstandins))
     filelist = []
     for f in changedstandins:
         if f[0] not in filelist:
             filelist.append(f[0])
     return filelist
 
 
 def getlfilestoupload(repo, missing, addfunc):
     makeprogress = repo.ui.makeprogress
     with makeprogress(
         _(b'finding outgoing largefiles'),
         unit=_(b'revisions'),
         total=len(missing),
     ) as progress:
         for i, n in enumerate(missing):
             progress.update(i)
             parents = [p for p in repo[n].parents() if p != repo.nullid]
 
             with lfstatus(repo, value=False):
                 ctx = repo[n]
 
             files = set(ctx.files())
             if len(parents) == 2:
                 mc = ctx.manifest()
                 mp1 = ctx.p1().manifest()
                 mp2 = ctx.p2().manifest()
                 for f in mp1:
                     if f not in mc:
                         files.add(f)
                 for f in mp2:
                     if f not in mc:
                         files.add(f)
                 for f in mc:
                     if mc[f] != mp1.get(f, None) or mc[f] != mp2.get(f, None):
                         files.add(f)
             for fn in files:
                 if isstandin(fn) and fn in ctx:
                     addfunc(fn, readasstandin(ctx[fn]))
 
 
 def updatestandinsbymatch(repo, match):
     """Update standins in the working directory according to specified match
 
     This returns (possibly modified) ``match`` object to be used for
     subsequent commit process.
     """
 
     ui = repo.ui
 
     # Case 1: user calls commit with no specific files or
     # include/exclude patterns: refresh and commit all files that
     # are "dirty".
     if match is None or match.always():
         # Spend a bit of time here to get a list of files we know
         # are modified so we can compare only against those.
         # It can cost a lot of time (several seconds)
         # otherwise to update all standins if the largefiles are
         # large.
         dirtymatch = matchmod.always()
         with repo.dirstate.running_status(repo):
             lfdirstate = openlfdirstate(ui, repo)
             unsure, s, mtime_boundary = lfdirstate.status(
                 dirtymatch,
                 subrepos=[],
                 ignored=False,
                 clean=False,
                 unknown=False,
             )
         modifiedfiles = unsure + s.modified + s.added + s.removed
         lfiles = listlfiles(repo)
         # this only loops through largefiles that exist (not
         # removed/renamed)
         for lfile in lfiles:
             if lfile in modifiedfiles:
                 fstandin = standin(lfile)
                 if repo.wvfs.exists(fstandin):
                     # this handles the case where a rebase is being
                     # performed and the working copy is not updated
                     # yet.
                     if repo.wvfs.exists(lfile):
                         updatestandin(repo, lfile, fstandin)
 
         return match
 
     lfiles = listlfiles(repo)
+    match._was_tampered_with = True
     match._files = repo._subdirlfs(match.files(), lfiles)
 
     # Case 2: user calls commit with specified patterns: refresh
     # any matching big files.
     smatcher = composestandinmatcher(repo, match)
     standins = repo.dirstate.walk(
         smatcher, subrepos=[], unknown=False, ignored=False
     )
 
     # No matching big files: get out of the way and pass control to
     # the usual commit() method.
     if not standins:
         return match
 
     # Refresh all matching big files.  It's possible that the
     # commit will end up failing, in which case the big files will
     # stay refreshed.  No harm done: the user modified them and
     # asked to commit them, so sooner or later we're going to
     # refresh the standins.  Might as well leave them refreshed.
     lfdirstate = openlfdirstate(ui, repo)
     for fstandin in standins:
         lfile = splitstandin(fstandin)
         if lfdirstate.get_entry(lfile).tracked:
             updatestandin(repo, lfile, fstandin)
 
     # Cook up a new matcher that only matches regular files or
     # standins corresponding to the big files requested by the
     # user.  Have to modify _files to prevent commit() from
     # complaining "not tracked" for big files.
     match = copy.copy(match)
+    match._was_tampered_with = True
     origmatchfn = match.matchfn
 
     # Check both the list of largefiles and the list of
     # standins because if a largefile was removed, it
     # won't be in the list of largefiles at this point
     match._files += sorted(standins)
 
     actualfiles = []
     for f in match._files:
         fstandin = standin(f)
 
         # For largefiles, only one of the normal and standin should be
         # committed (except if one of them is a remove).  In the case of a
         # standin removal, drop the normal file if it is unknown to dirstate.
         # Thus, skip plain largefile names but keep the standin.
         if f in lfiles or fstandin in standins:
             if not repo.dirstate.get_entry(fstandin).removed:
                 if not repo.dirstate.get_entry(f).removed:
                     continue
             elif not repo.dirstate.get_entry(f).any_tracked:
                 continue
 
         actualfiles.append(f)
     match._files = actualfiles
 
     def matchfn(f):
         if origmatchfn(f):
             return f not in lfiles
         else:
             return f in standins
 
     match.matchfn = matchfn
 
     return match
 
 
 class automatedcommithook:
     """Stateful hook to update standins at the 1st commit of resuming
 
     For efficiency, updating standins in the working directory should
     be avoided while automated committing (like rebase, transplant and
     so on), because they should be updated before committing.
 
     But the 1st commit of resuming automated committing (e.g. ``rebase
     --continue``) should update them, because largefiles may be
     modified manually.
     """
 
     def __init__(self, resuming):
         self.resuming = resuming
 
     def __call__(self, repo, match):
         if self.resuming:
             self.resuming = False  # avoids updating at subsequent commits
             return updatestandinsbymatch(repo, match)
         else:
             return match
 
 
 def getstatuswriter(ui, repo, forcibly=None):
     """Return the function to write largefiles specific status out
 
     If ``forcibly`` is ``None``, this returns the last element of
     ``repo._lfstatuswriters`` as "default" writer function.
 
     Otherwise, this returns the function to always write out (or
     ignore if ``not forcibly``) status.
     """
     if forcibly is None and hasattr(repo, '_largefilesenabled'):
         return repo._lfstatuswriters[-1]
     else:
         if forcibly:
             return ui.status  # forcibly WRITE OUT
         else:
             return lambda *msg, **opts: None  # forcibly IGNORE

hgext/largefiles/overrides.py

@@ -1,1924 +1,1932 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.
 
 '''Overridden Mercurial commands and functions for the largefiles extension'''
 
 import contextlib
 import copy
 import os
 
 from mercurial.i18n import _
 
 from mercurial.pycompat import open
 
 from mercurial.hgweb import webcommands
 
 from mercurial import (
     archival,
     cmdutil,
     copies as copiesmod,
     dirstate,
     error,
     exchange,
     extensions,
     exthelper,
     filemerge,
     hg,
     logcmdutil,
     match as matchmod,
     merge,
     mergestate as mergestatemod,
     pathutil,
     pycompat,
     scmutil,
     smartset,
     subrepo,
     url as urlmod,
     util,
 )
 
 from mercurial.upgrade_utils import (
     actions as upgrade_actions,
 )
 
 from . import (
     lfcommands,
     lfutil,
     storefactory,
 )
 
 ACTION_ADD = mergestatemod.ACTION_ADD
 ACTION_DELETED_CHANGED = mergestatemod.ACTION_DELETED_CHANGED
 ACTION_GET = mergestatemod.ACTION_GET
 ACTION_KEEP = mergestatemod.ACTION_KEEP
 ACTION_REMOVE = mergestatemod.ACTION_REMOVE
 
 eh = exthelper.exthelper()
 
 lfstatus = lfutil.lfstatus
 
 MERGE_ACTION_LARGEFILE_MARK_REMOVED = mergestatemod.MergeAction('lfmr')
 
 # -- Utility functions: commonly/repeatedly needed functionality ---------------
 
 
 def composelargefilematcher(match, manifest):
     """create a matcher that matches only the largefiles in the original
     matcher"""
     m = copy.copy(match)
+    m._was_tampered_with = True
     lfile = lambda f: lfutil.standin(f) in manifest
     m._files = [lf for lf in m._files if lfile(lf)]
     m._fileset = set(m._files)
     m.always = lambda: False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: lfile(f) and origmatchfn(f)
     return m
 
 
 def composenormalfilematcher(match, manifest, exclude=None):
     excluded = set()
     if exclude is not None:
         excluded.update(exclude)
 
     m = copy.copy(match)
+    m._was_tampered_with = True
     notlfile = lambda f: not (
         lfutil.isstandin(f) or lfutil.standin(f) in manifest or f in excluded
     )
     m._files = [lf for lf in m._files if notlfile(lf)]
     m._fileset = set(m._files)
     m.always = lambda: False
     origmatchfn = m.matchfn
     m.matchfn = lambda f: notlfile(f) and origmatchfn(f)
     return m
 
 
 def addlargefiles(ui, repo, isaddremove, matcher, uipathfn, **opts):
     large = opts.get('large')
     lfsize = lfutil.getminsize(
         ui, lfutil.islfilesrepo(repo), opts.get('lfsize')
     )
 
     lfmatcher = None
     if lfutil.islfilesrepo(repo):
         lfpats = ui.configlist(lfutil.longname, b'patterns')
         if lfpats:
             lfmatcher = matchmod.match(repo.root, b'', list(lfpats))
 
     lfnames = []
     m = matcher
 
     wctx = repo[None]
     for f in wctx.walk(matchmod.badmatch(m, lambda x, y: None)):
         exact = m.exact(f)
         lfile = lfutil.standin(f) in wctx
         nfile = f in wctx
         exists = lfile or nfile
 
         # Don't warn the user when they attempt to add a normal tracked file.
         # The normal add code will do that for us.
         if exact and exists:
             if lfile:
                 ui.warn(_(b'%s already a largefile\n') % uipathfn(f))
             continue
 
         if (exact or not exists) and not lfutil.isstandin(f):
             # In case the file was removed previously, but not committed
             # (issue3507)
             if not repo.wvfs.exists(f):
                 continue
 
             abovemin = (
                 lfsize and repo.wvfs.lstat(f).st_size >= lfsize * 1024 * 1024
             )
             if large or abovemin or (lfmatcher and lfmatcher(f)):
                 lfnames.append(f)
                 if ui.verbose or not exact:
                     ui.status(_(b'adding %s as a largefile\n') % uipathfn(f))
 
     bad = []
 
     # Need to lock, otherwise there could be a race condition between
     # when standins are created and added to the repo.
     with repo.wlock():
         if not opts.get('dry_run'):
             standins = []
             lfdirstate = lfutil.openlfdirstate(ui, repo)
             for f in lfnames:
                 standinname = lfutil.standin(f)
                 lfutil.writestandin(
                     repo,
                     standinname,
                     hash=b'',
                     executable=lfutil.getexecutable(repo.wjoin(f)),
                 )
                 standins.append(standinname)
                 lfdirstate.set_tracked(f)
             lfdirstate.write(repo.currenttransaction())
             bad += [
                 lfutil.splitstandin(f)
                 for f in repo[None].add(standins)
                 if f in m.files()
             ]
 
         added = [f for f in lfnames if f not in bad]
     return added, bad
 
 
 def removelargefiles(ui, repo, isaddremove, matcher, uipathfn, dryrun, **opts):
     after = opts.get('after')
     m = composelargefilematcher(matcher, repo[None].manifest())
     with lfstatus(repo):
         s = repo.status(match=m, clean=not isaddremove)
     manifest = repo[None].manifest()
     modified, added, deleted, clean = [
         [f for f in list if lfutil.standin(f) in manifest]
         for list in (s.modified, s.added, s.deleted, s.clean)
     ]
 
     def warn(files, msg):
         for f in files:
             ui.warn(msg % uipathfn(f))
         return int(len(files) > 0)
 
     if after:
         remove = deleted
         result = warn(
             modified + added + clean, _(b'not removing %s: file still exists\n')
         )
     else:
         remove = deleted + clean
         result = warn(
             modified,
             _(
                 b'not removing %s: file is modified (use -f'
                 b' to force removal)\n'
             ),
         )
         result = (
             warn(
                 added,
                 _(
                     b'not removing %s: file has been marked for add'
                     b' (use forget to undo)\n'
                 ),
             )
             or result
         )
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in sorted(remove):
             if ui.verbose or not m.exact(f):
                 ui.status(_(b'removing %s\n') % uipathfn(f))
 
             if not dryrun:
                 if not after:
                     repo.wvfs.unlinkpath(f, ignoremissing=True)
 
         if dryrun:
             return result
 
         remove = [lfutil.standin(f) for f in remove]
         # If this is being called by addremove, let the original addremove
         # function handle this.
         if not isaddremove:
             for f in remove:
                 repo.wvfs.unlinkpath(f, ignoremissing=True)
         repo[None].forget(remove)
 
         for f in remove:
             lfdirstate.set_untracked(lfutil.splitstandin(f))
 
         lfdirstate.write(repo.currenttransaction())
 
     return result
 
 
 # For overriding mercurial.hgweb.webcommands so that largefiles will
 # appear at their right place in the manifests.
 @eh.wrapfunction(webcommands, 'decodepath')
 def decodepath(orig, path):
     return lfutil.splitstandin(path) or path
 
 
 # -- Wrappers: modify existing commands --------------------------------
 
 
 @eh.wrapcommand(
     b'add',
     opts=[
         (b'', b'large', None, _(b'add as largefile')),
         (b'', b'normal', None, _(b'add as normal file')),
         (
             b'',
             b'lfsize',
             b'',
             _(
                 b'add all files above this size (in megabytes) '
                 b'as largefiles (default: 10)'
             ),
         ),
     ],
 )
 def overrideadd(orig, ui, repo, *pats, **opts):
     if opts.get('normal') and opts.get('large'):
         raise error.Abort(_(b'--normal cannot be used with --large'))
     return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(cmdutil, 'add')
 def cmdutiladd(orig, ui, repo, matcher, prefix, uipathfn, explicitonly, **opts):
     # The --normal flag short circuits this override
     if opts.get('normal'):
         return orig(ui, repo, matcher, prefix, uipathfn, explicitonly, **opts)
 
     ladded, lbad = addlargefiles(ui, repo, False, matcher, uipathfn, **opts)
     normalmatcher = composenormalfilematcher(
         matcher, repo[None].manifest(), ladded
     )
     bad = orig(ui, repo, normalmatcher, prefix, uipathfn, explicitonly, **opts)
 
     bad.extend(f for f in lbad)
     return bad
 
 
 @eh.wrapfunction(cmdutil, 'remove')
 def cmdutilremove(
     orig, ui, repo, matcher, prefix, uipathfn, after, force, subrepos, dryrun
 ):
     normalmatcher = composenormalfilematcher(matcher, repo[None].manifest())
     result = orig(
         ui,
         repo,
         normalmatcher,
         prefix,
         uipathfn,
         after,
         force,
         subrepos,
         dryrun,
     )
     return (
         removelargefiles(
             ui, repo, False, matcher, uipathfn, dryrun, after=after, force=force
         )
         or result
     )
 
 
 @eh.wrapfunction(dirstate.dirstate, '_changing')
 @contextlib.contextmanager
 def _changing(orig, self, repo, change_type):
     pre = sub_dirstate = getattr(self, '_sub_dirstate', None)
     try:
         lfd = getattr(self, '_large_file_dirstate', False)
         if sub_dirstate is None and not lfd:
             sub_dirstate = lfutil.openlfdirstate(repo.ui, repo)
             self._sub_dirstate = sub_dirstate
         if not lfd:
             assert self._sub_dirstate is not None
         with orig(self, repo, change_type):
             if sub_dirstate is None:
                 yield
             else:
                 with sub_dirstate._changing(repo, change_type):
                     yield
     finally:
         self._sub_dirstate = pre
 
 
 @eh.wrapfunction(dirstate.dirstate, 'running_status')
 @contextlib.contextmanager
 def running_status(orig, self, repo):
     pre = sub_dirstate = getattr(self, '_sub_dirstate', None)
     try:
         lfd = getattr(self, '_large_file_dirstate', False)
         if sub_dirstate is None and not lfd:
             sub_dirstate = lfutil.openlfdirstate(repo.ui, repo)
             self._sub_dirstate = sub_dirstate
         if not lfd:
             assert self._sub_dirstate is not None
         with orig(self, repo):
             if sub_dirstate is None:
                 yield
             else:
                 with sub_dirstate.running_status(repo):
                     yield
     finally:
         self._sub_dirstate = pre
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, 'status')
 def overridestatusfn(orig, repo, rev2, **opts):
     with lfstatus(repo._repo):
         return orig(repo, rev2, **opts)
 
 
 @eh.wrapcommand(b'status')
 def overridestatus(orig, ui, repo, *pats, **opts):
     with lfstatus(repo):
         return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, 'dirty')
 def overridedirty(orig, repo, ignoreupdate=False, missing=False):
     with lfstatus(repo._repo):
         return orig(repo, ignoreupdate=ignoreupdate, missing=missing)
 
 
 @eh.wrapcommand(b'log')
 def overridelog(orig, ui, repo, *pats, **opts):
     def overridematchandpats(
         orig,
         ctx,
         pats=(),
         opts=None,
         globbed=False,
         default=b'relpath',
         badfn=None,
     ):
         """Matcher that merges root directory with .hglf, suitable for log.
         It is still possible to match .hglf directly.
         For any listed files run log on the standin too.
         matchfn tries both the given filename and with .hglf stripped.
         """
         if opts is None:
             opts = {}
         matchandpats = orig(ctx, pats, opts, globbed, default, badfn=badfn)
         m, p = copy.copy(matchandpats)
 
         if m.always():
             # We want to match everything anyway, so there's no benefit trying
             # to add standins.
             return matchandpats
 
         pats = set(p)
 
         def fixpats(pat, tostandin=lfutil.standin):
             if pat.startswith(b'set:'):
                 return pat
 
             kindpat = matchmod._patsplit(pat, None)
 
             if kindpat[0] is not None:
                 return kindpat[0] + b':' + tostandin(kindpat[1])
             return tostandin(kindpat[1])
 
         cwd = repo.getcwd()
         if cwd:
             hglf = lfutil.shortname
             back = util.pconvert(repo.pathto(hglf)[: -len(hglf)])
 
             def tostandin(f):
                 # The file may already be a standin, so truncate the back
                 # prefix and test before mangling it.  This avoids turning
                 # 'glob:../.hglf/foo*' into 'glob:../.hglf/../.hglf/foo*'.
                 if f.startswith(back) and lfutil.splitstandin(f[len(back) :]):
                     return f
 
                 # An absolute path is from outside the repo, so truncate the
                 # path to the root before building the standin.  Otherwise cwd
                 # is somewhere in the repo, relative to root, and needs to be
                 # prepended before building the standin.
                 if os.path.isabs(cwd):
                     f = f[len(back) :]
                 else:
                     f = cwd + b'/' + f
                 return back + lfutil.standin(f)
 
         else:
 
             def tostandin(f):
                 if lfutil.isstandin(f):
                     return f
                 return lfutil.standin(f)
 
         pats.update(fixpats(f, tostandin) for f in p)
 
+        m._was_tampered_with = True
+
         for i in range(0, len(m._files)):
             # Don't add '.hglf' to m.files, since that is already covered by '.'
             if m._files[i] == b'.':
                 continue
             standin = lfutil.standin(m._files[i])
             # If the "standin" is a directory, append instead of replace to
             # support naming a directory on the command line with only
             # largefiles.  The original directory is kept to support normal
             # files.
             if standin in ctx:
                 m._files[i] = standin
             elif m._files[i] not in ctx and repo.wvfs.isdir(standin):
                 m._files.append(standin)
 
         m._fileset = set(m._files)
         m.always = lambda: False
         origmatchfn = m.matchfn
 
         def lfmatchfn(f):
             lf = lfutil.splitstandin(f)
             if lf is not None and origmatchfn(lf):
                 return True
             r = origmatchfn(f)
             return r
 
         m.matchfn = lfmatchfn
 
         ui.debug(b'updated patterns: %s\n' % b', '.join(sorted(pats)))
         return m, pats
 
     # For hg log --patch, the match object is used in two different senses:
     # (1) to determine what revisions should be printed out, and
     # (2) to determine what files to print out diffs for.
     # The magic matchandpats override should be used for case (1) but not for
     # case (2).
     oldmatchandpats = scmutil.matchandpats
 
     def overridemakefilematcher(orig, repo, pats, opts, badfn=None):
         wctx = repo[None]
         match, pats = oldmatchandpats(wctx, pats, opts, badfn=badfn)
         return lambda ctx: match
 
     wrappedmatchandpats = extensions.wrappedfunction(
         scmutil, 'matchandpats', overridematchandpats
     )
     wrappedmakefilematcher = extensions.wrappedfunction(
         logcmdutil, '_makenofollowfilematcher', overridemakefilematcher
     )
     with wrappedmatchandpats, wrappedmakefilematcher:
         return orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapcommand(
     b'verify',
     opts=[
         (
             b'',
             b'large',
             None,
             _(b'verify that all largefiles in current revision exists'),
         ),
         (
             b'',
             b'lfa',
             None,
             _(b'verify largefiles in all revisions, not just current'),
         ),
         (
             b'',
             b'lfc',
             None,
             _(b'verify local largefile contents, not just existence'),
         ),
     ],
 )
 def overrideverify(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     all = opts.pop('lfa', False)
     contents = opts.pop('lfc', False)
 
     result = orig(ui, repo, *pats, **opts)
     if large or all or contents:
         result = result or lfcommands.verifylfiles(ui, repo, all, contents)
     return result
 
 
 @eh.wrapcommand(
     b'debugstate',
     opts=[(b'', b'large', None, _(b'display largefiles dirstate'))],
 )
 def overridedebugstate(orig, ui, repo, *pats, **opts):
     large = opts.pop('large', False)
     if large:
 
         class fakerepo:
             dirstate = lfutil.openlfdirstate(ui, repo)
 
         orig(ui, fakerepo, *pats, **opts)
     else:
         orig(ui, repo, *pats, **opts)
 
 
 # Before starting the manifest merge, merge.updates will call
 # _checkunknownfile to check if there are any files in the merged-in
 # changeset that collide with unknown files in the working copy.
 #
 # The largefiles are seen as unknown, so this prevents us from merging
 # in a file 'foo' if we already have a largefile with the same name.
 #
 # The overridden function filters the unknown files by removing any
 # largefiles. This makes the merge proceed and we can then handle this
 # case further in the overridden calculateupdates function below.
 @eh.wrapfunction(merge, '_checkunknownfile')
 def overridecheckunknownfile(
     origfn, dirstate, wvfs, dircache, wctx, mctx, f, f2=None
 ):
     if lfutil.standin(dirstate.normalize(f)) in wctx:
         return False
     return origfn(dirstate, wvfs, dircache, wctx, mctx, f, f2)
 
 
 # The manifest merge handles conflicts on the manifest level. We want
 # to handle changes in largefile-ness of files at this level too.
 #
 # The strategy is to run the original calculateupdates and then process
 # the action list it outputs. There are two cases we need to deal with:
 #
 # 1. Normal file in p1, largefile in p2. Here the largefile is
 #    detected via its standin file, which will enter the working copy
 #    with a "get" action. It is not "merge" since the standin is all
 #    Mercurial is concerned with at this level -- the link to the
 #    existing normal file is not relevant here.
 #
 # 2. Largefile in p1, normal file in p2. Here we get a "merge" action
 #    since the largefile will be present in the working copy and
 #    different from the normal file in p2. Mercurial therefore
 #    triggers a merge action.
 #
 # In both cases, we prompt the user and emit new actions to either
 # remove the standin (if the normal file was kept) or to remove the
 # normal file and get the standin (if the largefile was kept). The
 # default prompt answer is to use the largefile version since it was
 # presumably changed on purpose.
 #
 # Finally, the merge.applyupdates function will then take care of
 # writing the files into the working copy and lfcommands.updatelfiles
 # will update the largefiles.
 @eh.wrapfunction(merge, 'calculateupdates')
 def overridecalculateupdates(
     origfn, repo, p1, p2, pas, branchmerge, force, acceptremote, *args, **kwargs
 ):
     overwrite = force and not branchmerge
     mresult = origfn(
         repo, p1, p2, pas, branchmerge, force, acceptremote, *args, **kwargs
     )
 
     if overwrite:
         return mresult
 
     # Convert to dictionary with filename as key and action as value.
     lfiles = set()
     for f in mresult.files():
         splitstandin = lfutil.splitstandin(f)
         if splitstandin is not None and splitstandin in p1:
             lfiles.add(splitstandin)
         elif lfutil.standin(f) in p1:
             lfiles.add(f)
 
     for lfile in sorted(lfiles):
         standin = lfutil.standin(lfile)
         (lm, largs, lmsg) = mresult.getfile(lfile, (None, None, None))
         (sm, sargs, smsg) = mresult.getfile(standin, (None, None, None))
 
         if sm in (ACTION_GET, ACTION_DELETED_CHANGED) and lm != ACTION_REMOVE:
             if sm == ACTION_DELETED_CHANGED:
                 f1, f2, fa, move, anc = sargs
                 sargs = (p2[f2].flags(), False)
             # Case 1: normal file in the working copy, largefile in
             # the second parent
             usermsg = (
                 _(
                     b'remote turned local normal file %s into a largefile\n'
                     b'use (l)argefile or keep (n)ormal file?'
                     b'$$ &Largefile $$ &Normal file'
                 )
                 % lfile
             )
             if repo.ui.promptchoice(usermsg, 0) == 0:  # pick remote largefile
                 mresult.addfile(
                     lfile, ACTION_REMOVE, None, b'replaced by standin'
                 )
                 mresult.addfile(standin, ACTION_GET, sargs, b'replaces standin')
             else:  # keep local normal file
                 mresult.addfile(lfile, ACTION_KEEP, None, b'replaces standin')
                 if branchmerge:
                     mresult.addfile(
                         standin,
                         ACTION_KEEP,
                         None,
                         b'replaced by non-standin',
                     )
                 else:
                     mresult.addfile(
                         standin,
                         ACTION_REMOVE,
                         None,
                         b'replaced by non-standin',
                     )
         if lm in (ACTION_GET, ACTION_DELETED_CHANGED) and sm != ACTION_REMOVE:
             if lm == ACTION_DELETED_CHANGED:
                 f1, f2, fa, move, anc = largs
                 largs = (p2[f2].flags(), False)
             # Case 2: largefile in the working copy, normal file in
             # the second parent
             usermsg = (
                 _(
                     b'remote turned local largefile %s into a normal file\n'
                     b'keep (l)argefile or use (n)ormal file?'
                     b'$$ &Largefile $$ &Normal file'
                 )
                 % lfile
             )
             if repo.ui.promptchoice(usermsg, 0) == 0:  # keep local largefile
                 if branchmerge:
                     # largefile can be restored from standin safely
                     mresult.addfile(
                         lfile,
                         ACTION_KEEP,
                         None,
                         b'replaced by standin',
                     )
                     mresult.addfile(
                         standin, ACTION_KEEP, None, b'replaces standin'
                     )
                 else:
                     # "lfile" should be marked as "removed" without
                     # removal of itself
                     mresult.addfile(
                         lfile,
                         MERGE_ACTION_LARGEFILE_MARK_REMOVED,
                         None,
                         b'forget non-standin largefile',
                     )
 
                     # linear-merge should treat this largefile as 're-added'
                     mresult.addfile(standin, ACTION_ADD, None, b'keep standin')
             else:  # pick remote normal file
                 mresult.addfile(lfile, ACTION_GET, largs, b'replaces standin')
                 mresult.addfile(
                     standin,
                     ACTION_REMOVE,
                     None,
                     b'replaced by non-standin',
                 )
 
     return mresult
 
 
 @eh.wrapfunction(mergestatemod, 'recordupdates')
 def mergerecordupdates(orig, repo, actions, branchmerge, getfiledata):
     if MERGE_ACTION_LARGEFILE_MARK_REMOVED in actions:
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         for lfile, args, msg in actions[MERGE_ACTION_LARGEFILE_MARK_REMOVED]:
             # this should be executed before 'orig', to execute 'remove'
             # before all other actions
             repo.dirstate.update_file(lfile, p1_tracked=True, wc_tracked=False)
             # make sure lfile doesn't get synclfdirstate'd as normal
             lfdirstate.update_file(lfile, p1_tracked=False, wc_tracked=True)
 
     return orig(repo, actions, branchmerge, getfiledata)
 
 
 # Override filemerge to prompt the user about how they wish to merge
 # largefiles. This will handle identical edits without prompting the user.
 @eh.wrapfunction(filemerge, 'filemerge')
 def overridefilemerge(
     origfn, repo, wctx, mynode, orig, fcd, fco, fca, labels=None
 ):
     if not lfutil.isstandin(orig) or fcd.isabsent() or fco.isabsent():
         return origfn(repo, wctx, mynode, orig, fcd, fco, fca, labels=labels)
 
     ahash = lfutil.readasstandin(fca).lower()
     dhash = lfutil.readasstandin(fcd).lower()
     ohash = lfutil.readasstandin(fco).lower()
     if (
         ohash != ahash
         and ohash != dhash
         and (
             dhash == ahash
             or repo.ui.promptchoice(
                 _(
                     b'largefile %s has a merge conflict\nancestor was %s\n'
                     b'you can keep (l)ocal %s or take (o)ther %s.\n'
                     b'what do you want to do?'
                     b'$$ &Local $$ &Other'
                 )
                 % (lfutil.splitstandin(orig), ahash, dhash, ohash),
                 0,
             )
             == 1
         )
     ):
         repo.wwrite(fcd.path(), fco.data(), fco.flags())
     return 0, False
 
 
 @eh.wrapfunction(copiesmod, 'pathcopies')
 def copiespathcopies(orig, ctx1, ctx2, match=None):
     copies = orig(ctx1, ctx2, match=match)
     updated = {}
 
     for k, v in copies.items():
         updated[lfutil.splitstandin(k) or k] = lfutil.splitstandin(v) or v
 
     return updated
 
 
 # Copy first changes the matchers to match standins instead of
 # largefiles.  Then it overrides util.copyfile in that function it
 # checks if the destination largefile already exists. It also keeps a
 # list of copied files so that the largefiles can be copied and the
 # dirstate updated.
 @eh.wrapfunction(cmdutil, 'copy')
 def overridecopy(orig, ui, repo, pats, opts, rename=False):
     # doesn't remove largefile on rename
     if len(pats) < 2:
         # this isn't legal, let the original function deal with it
         return orig(ui, repo, pats, opts, rename)
 
     # This could copy both lfiles and normal files in one command,
     # but we don't want to do that. First replace their matcher to
     # only match normal files and run it, then replace it to just
     # match largefiles and run it again.
     nonormalfiles = False
     nolfiles = False
     manifest = repo[None].manifest()
 
     def normalfilesmatchfn(
         orig,
         ctx,
         pats=(),
         opts=None,
         globbed=False,
         default=b'relpath',
         badfn=None,
     ):
         if opts is None:
             opts = {}
         match = orig(ctx, pats, opts, globbed, default, badfn=badfn)
         return composenormalfilematcher(match, manifest)
 
     with extensions.wrappedfunction(scmutil, 'match', normalfilesmatchfn):
         try:
             result = orig(ui, repo, pats, opts, rename)
         except error.Abort as e:
             if e.message != _(b'no files to copy'):
                 raise e
             else:
                 nonormalfiles = True
             result = 0
 
     # The first rename can cause our current working directory to be removed.
     # In that case there is nothing left to copy/rename so just quit.
     try:
         repo.getcwd()
     except OSError:
         return result
 
     def makestandin(relpath):
         path = pathutil.canonpath(repo.root, repo.getcwd(), relpath)
         return repo.wvfs.join(lfutil.standin(path))
 
     fullpats = scmutil.expandpats(pats)
     dest = fullpats[-1]
 
     if os.path.isdir(dest):
         if not os.path.isdir(makestandin(dest)):
             os.makedirs(makestandin(dest))
 
     try:
         # When we call orig below it creates the standins but we don't add
         # them to the dir state until later so lock during that time.
         wlock = repo.wlock()
 
         manifest = repo[None].manifest()
 
         def overridematch(
             orig,
             ctx,
             pats=(),
             opts=None,
             globbed=False,
             default=b'relpath',
             badfn=None,
         ):
             if opts is None:
                 opts = {}
             newpats = []
             # The patterns were previously mangled to add the standin
             # directory; we need to remove that now
             for pat in pats:
                 if matchmod.patkind(pat) is None and lfutil.shortname in pat:
                     newpats.append(pat.replace(lfutil.shortname, b''))
                 else:
                     newpats.append(pat)
             match = orig(ctx, newpats, opts, globbed, default, badfn=badfn)
             m = copy.copy(match)
+            m._was_tampered_with = True
             lfile = lambda f: lfutil.standin(f) in manifest
             m._files = [lfutil.standin(f) for f in m._files if lfile(f)]
             m._fileset = set(m._files)
             origmatchfn = m.matchfn
 
             def matchfn(f):
                 lfile = lfutil.splitstandin(f)
                 return (
                     lfile is not None
                     and (f in manifest)
                     and origmatchfn(lfile)
                     or None
                 )
 
             m.matchfn = matchfn
             return m
 
         listpats = []
         for pat in pats:
             if matchmod.patkind(pat) is not None:
                 listpats.append(pat)
             else:
                 listpats.append(makestandin(pat))
 
         copiedfiles = []
 
         def overridecopyfile(orig, src, dest, *args, **kwargs):
             if lfutil.shortname in src and dest.startswith(
                 repo.wjoin(lfutil.shortname)
             ):
                 destlfile = dest.replace(lfutil.shortname, b'')
                 if not opts[b'force'] and os.path.exists(destlfile):
                     raise IOError(
                         b'', _(b'destination largefile already exists')
                     )
             copiedfiles.append((src, dest))
             orig(src, dest, *args, **kwargs)
 
         with extensions.wrappedfunction(util, 'copyfile', overridecopyfile):
             with extensions.wrappedfunction(scmutil, 'match', overridematch):
                 result += orig(ui, repo, listpats, opts, rename)
 
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for (src, dest) in copiedfiles:
             if lfutil.shortname in src and dest.startswith(
                 repo.wjoin(lfutil.shortname)
             ):
                 srclfile = src.replace(repo.wjoin(lfutil.standin(b'')), b'')
                 destlfile = dest.replace(repo.wjoin(lfutil.standin(b'')), b'')
                 destlfiledir = repo.wvfs.dirname(repo.wjoin(destlfile)) or b'.'
                 if not os.path.isdir(destlfiledir):
                     os.makedirs(destlfiledir)
                 if rename:
                     os.rename(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                     # The file is gone, but this deletes any empty parent
                     # directories as a side-effect.
                     repo.wvfs.unlinkpath(srclfile, ignoremissing=True)
                     lfdirstate.set_untracked(srclfile)
                 else:
                     util.copyfile(repo.wjoin(srclfile), repo.wjoin(destlfile))
 
                 lfdirstate.set_tracked(destlfile)
         lfdirstate.write(repo.currenttransaction())
     except error.Abort as e:
         if e.message != _(b'no files to copy'):
             raise e
         else:
             nolfiles = True
     finally:
         wlock.release()
 
     if nolfiles and nonormalfiles:
         raise error.Abort(_(b'no files to copy'))
 
     return result
 
 
 # When the user calls revert, we have to be careful to not revert any
 # changes to other largefiles accidentally. This means we have to keep
 # track of the largefiles that are being reverted so we only pull down
 # the necessary largefiles.
 #
 # Standins are only updated (to match the hash of largefiles) before
 # commits. Update the standins then run the original revert, changing
 # the matcher to hit standins instead of largefiles. Based on the
 # resulting standins update the largefiles.
 @eh.wrapfunction(cmdutil, 'revert')
 def overriderevert(orig, ui, repo, ctx, *pats, **opts):
     # Because we put the standins in a bad state (by updating them)
     # and then return them to a correct state we need to lock to
     # prevent others from changing them in their incorrect state.
     with repo.wlock(), repo.dirstate.running_status(repo):
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         s = lfutil.lfdirstatestatus(lfdirstate, repo)
         lfdirstate.write(repo.currenttransaction())
         for lfile in s.modified:
             lfutil.updatestandin(repo, lfile, lfutil.standin(lfile))
         for lfile in s.deleted:
             fstandin = lfutil.standin(lfile)
             if repo.wvfs.exists(fstandin):
                 repo.wvfs.unlink(fstandin)
 
         oldstandins = lfutil.getstandinsstate(repo)
 
         def overridematch(
             orig,
             mctx,
             pats=(),
             opts=None,
             globbed=False,
             default=b'relpath',
             badfn=None,
         ):
             if opts is None:
                 opts = {}
             match = orig(mctx, pats, opts, globbed, default, badfn=badfn)
             m = copy.copy(match)
+            m._was_tampered_with = True
 
             # revert supports recursing into subrepos, and though largefiles
             # currently doesn't work correctly in that case, this match is
             # called, so the lfdirstate above may not be the correct one for
             # this invocation of match.
             lfdirstate = lfutil.openlfdirstate(
                 mctx.repo().ui, mctx.repo(), False
             )
 
             wctx = repo[None]
             matchfiles = []
             for f in m._files:
                 standin = lfutil.standin(f)
                 if standin in ctx or standin in mctx:
                     matchfiles.append(standin)
                 elif standin in wctx or lfdirstate.get_entry(f).removed:
                     continue
                 else:
                     matchfiles.append(f)
             m._files = matchfiles
             m._fileset = set(m._files)
             origmatchfn = m.matchfn
 
             def matchfn(f):
                 lfile = lfutil.splitstandin(f)
                 if lfile is not None:
                     return origmatchfn(lfile) and (f in ctx or f in mctx)
                 return origmatchfn(f)
 
             m.matchfn = matchfn
             return m
 
         with extensions.wrappedfunction(scmutil, 'match', overridematch):
             orig(ui, repo, ctx, *pats, **opts)
 
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
         # lfdirstate should be 'normallookup'-ed for updated files,
         # because reverting doesn't touch dirstate for 'normal' files
         # when target revision is explicitly specified: in such case,
         # 'n' and valid timestamp in dirstate doesn't ensure 'clean'
         # of target (standin) file.
         lfcommands.updatelfiles(
             ui, repo, filelist, printmessage=False, normallookup=True
         )
 
 
 # after pulling changesets, we need to take some extra care to get
 # largefiles updated remotely
 @eh.wrapcommand(
     b'pull',
     opts=[
         (
             b'',
             b'all-largefiles',
             None,
             _(b'download all pulled versions of largefiles (DEPRECATED)'),
         ),
         (
             b'',
             b'lfrev',
             [],
             _(b'download largefiles for these revisions'),
             _(b'REV'),
         ),
     ],
 )
 def overridepull(orig, ui, repo, source=None, **opts):
     revsprepull = len(repo)
     if not source:
         source = b'default'
     repo.lfpullsource = source
     result = orig(ui, repo, source, **opts)
     revspostpull = len(repo)
     lfrevs = opts.get('lfrev', [])
     if opts.get('all_largefiles'):
         lfrevs.append(b'pulled()')
     if lfrevs and revspostpull > revsprepull:
         numcached = 0
         repo.firstpulled = revsprepull  # for pulled() revset expression
         try:
             for rev in logcmdutil.revrange(repo, lfrevs):
                 ui.note(_(b'pulling largefiles for revision %d\n') % rev)
                 (cached, missing) = lfcommands.cachelfiles(ui, repo, rev)
                 numcached += len(cached)
         finally:
             del repo.firstpulled
         ui.status(_(b"%d largefiles cached\n") % numcached)
     return result
 
 
 @eh.wrapcommand(
     b'push',
     opts=[
         (
             b'',
             b'lfrev',
             [],
             _(b'upload largefiles for these revisions'),
             _(b'REV'),
         )
     ],
 )
 def overridepush(orig, ui, repo, *args, **kwargs):
     """Override push command and store --lfrev parameters in opargs"""
     lfrevs = kwargs.pop('lfrev', None)
     if lfrevs:
         opargs = kwargs.setdefault('opargs', {})
         opargs[b'lfrevs'] = logcmdutil.revrange(repo, lfrevs)
     return orig(ui, repo, *args, **kwargs)
 
 
 @eh.wrapfunction(exchange, 'pushoperation')
 def exchangepushoperation(orig, *args, **kwargs):
     """Override pushoperation constructor and store lfrevs parameter"""
     lfrevs = kwargs.pop('lfrevs', None)
     pushop = orig(*args, **kwargs)
     pushop.lfrevs = lfrevs
     return pushop
 
 
 @eh.revsetpredicate(b'pulled()')
 def pulledrevsetsymbol(repo, subset, x):
     """Changesets that just has been pulled.
 
     Only available with largefiles from pull --lfrev expressions.
 
     .. container:: verbose
 
       Some examples:
 
       - pull largefiles for all new changesets::
 
           hg pull -lfrev "pulled()"
 
       - pull largefiles for all new branch heads::
 
           hg pull -lfrev "head(pulled()) and not closed()"
 
     """
 
     try:
         firstpulled = repo.firstpulled
     except AttributeError:
         raise error.Abort(_(b"pulled() only available in --lfrev"))
     return smartset.baseset([r for r in subset if r >= firstpulled])
 
 
 @eh.wrapcommand(
     b'clone',
     opts=[
         (
             b'',
             b'all-largefiles',
             None,
             _(b'download all versions of all largefiles'),
         )
     ],
 )
 def overrideclone(orig, ui, source, dest=None, **opts):
     d = dest
     if d is None:
         d = hg.defaultdest(source)
     if opts.get('all_largefiles') and not hg.islocal(d):
         raise error.Abort(
             _(b'--all-largefiles is incompatible with non-local destination %s')
             % d
         )
 
     return orig(ui, source, dest, **opts)
 
 
 @eh.wrapfunction(hg, 'clone')
 def hgclone(orig, ui, opts, *args, **kwargs):
     result = orig(ui, opts, *args, **kwargs)
 
     if result is not None:
         sourcerepo, destrepo = result
         repo = destrepo.local()
 
         # When cloning to a remote repo (like through SSH), no repo is available
         # from the peer.   Therefore the largefiles can't be downloaded and the
         # hgrc can't be updated.
         if not repo:
             return result
 
         # Caching is implicitly limited to 'rev' option, since the dest repo was
         # truncated at that point.  The user may expect a download count with
         # this option, so attempt whether or not this is a largefile repo.
         if opts.get(b'all_largefiles'):
             success, missing = lfcommands.downloadlfiles(ui, repo)
 
             if missing != 0:
                 return None
 
     return result
 
 
 @eh.wrapcommand(b'rebase', extension=b'rebase')
 def overriderebasecmd(orig, ui, repo, **opts):
     if not hasattr(repo, '_largefilesenabled'):
         return orig(ui, repo, **opts)
 
     resuming = opts.get('continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         with ui.configoverride(
             {(b'rebase', b'experimental.inmemory'): False}, b"largefiles"
         ):
             return orig(ui, repo, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
 
 
 @eh.extsetup
 def overriderebase(ui):
     try:
         rebase = extensions.find(b'rebase')
     except KeyError:
         pass
     else:
 
         def _dorebase(orig, *args, **kwargs):
             kwargs['inmemory'] = False
             return orig(*args, **kwargs)
 
         extensions.wrapfunction(rebase, '_dorebase', _dorebase)
 
 
 @eh.wrapcommand(b'archive')
 def overridearchivecmd(orig, ui, repo, dest, **opts):
     with lfstatus(repo.unfiltered()):
         return orig(ui, repo.unfiltered(), dest, **opts)
 
 
 @eh.wrapfunction(webcommands, 'archive')
 def hgwebarchive(orig, web):
     with lfstatus(web.repo):
         return orig(web)
 
 
 @eh.wrapfunction(archival, 'archive')
 def overridearchive(
     orig,
     repo,
     dest,
     node,
     kind,
     decode=True,
     match=None,
     prefix=b'',
     mtime=None,
     subrepos=None,
 ):
     # For some reason setting repo.lfstatus in hgwebarchive only changes the
     # unfiltered repo's attr, so check that as well.
     if not repo.lfstatus and not repo.unfiltered().lfstatus:
         return orig(
             repo, dest, node, kind, decode, match, prefix, mtime, subrepos
         )
 
     # No need to lock because we are only reading history and
     # largefile caches, neither of which are modified.
     if node is not None:
         lfcommands.cachelfiles(repo.ui, repo, node)
 
     if kind not in archival.archivers:
         raise error.Abort(_(b"unknown archive type '%s'") % kind)
 
     ctx = repo[node]
 
     if kind == b'files':
         if prefix:
             raise error.Abort(_(b'cannot give prefix when archiving to files'))
     else:
         prefix = archival.tidyprefix(dest, kind, prefix)
 
     def write(name, mode, islink, getdata):
         if match and not match(name):
             return
         data = getdata()
         if decode:
             data = repo.wwritedata(name, data)
         archiver.addfile(prefix + name, mode, islink, data)
 
     archiver = archival.archivers[kind](dest, mtime or ctx.date()[0])
 
     if repo.ui.configbool(b"ui", b"archivemeta"):
         write(
             b'.hg_archival.txt',
             0o644,
             False,
             lambda: archival.buildmetadata(ctx),
         )
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         lfile = lfutil.splitstandin(f)
         if lfile is not None:
             if node is not None:
                 path = lfutil.findfile(repo, getdata().strip())
 
                 if path is None:
                     raise error.Abort(
                         _(
                             b'largefile %s not found in repo store or system cache'
                         )
                         % lfile
                     )
             else:
                 path = lfile
 
             f = lfile
 
             getdata = lambda: util.readfile(path)
         write(f, b'x' in ff and 0o755 or 0o644, b'l' in ff, getdata)
 
     if subrepos:
         for subpath in sorted(ctx.substate):
             sub = ctx.workingsub(subpath)
             submatch = matchmod.subdirmatcher(subpath, match)
             subprefix = prefix + subpath + b'/'
 
             # TODO: Only hgsubrepo instances have `_repo`, so figure out how to
             # infer and possibly set lfstatus in hgsubrepoarchive.  That would
             # allow only hgsubrepos to set this, instead of the current scheme
             # where the parent sets this for the child.
             with (
                 hasattr(sub, '_repo')
                 and lfstatus(sub._repo)
                 or util.nullcontextmanager()
             ):
                 sub.archive(archiver, subprefix, submatch)
 
     archiver.done()
 
 
 @eh.wrapfunction(subrepo.hgsubrepo, 'archive')
 def hgsubrepoarchive(orig, repo, archiver, prefix, match=None, decode=True):
     lfenabled = hasattr(repo._repo, '_largefilesenabled')
     if not lfenabled or not repo._repo.lfstatus:
         return orig(repo, archiver, prefix, match, decode)
 
     repo._get(repo._state + (b'hg',))
     rev = repo._state[1]
     ctx = repo._repo[rev]
 
     if ctx.node() is not None:
         lfcommands.cachelfiles(repo.ui, repo._repo, ctx.node())
 
     def write(name, mode, islink, getdata):
         # At this point, the standin has been replaced with the largefile name,
         # so the normal matcher works here without the lfutil variants.
         if match and not match(f):
             return
         data = getdata()
         if decode:
             data = repo._repo.wwritedata(name, data)
 
         archiver.addfile(prefix + name, mode, islink, data)
 
     for f in ctx:
         ff = ctx.flags(f)
         getdata = ctx[f].data
         lfile = lfutil.splitstandin(f)
         if lfile is not None:
             if ctx.node() is not None:
                 path = lfutil.findfile(repo._repo, getdata().strip())
 
                 if path is None:
                     raise error.Abort(
                         _(
                             b'largefile %s not found in repo store or system cache'
                         )
                         % lfile
                     )
             else:
                 path = lfile
 
             f = lfile
 
             getdata = lambda: util.readfile(os.path.join(prefix, path))
 
         write(f, b'x' in ff and 0o755 or 0o644, b'l' in ff, getdata)
 
     for subpath in sorted(ctx.substate):
         sub = ctx.workingsub(subpath)
         submatch = matchmod.subdirmatcher(subpath, match)
         subprefix = prefix + subpath + b'/'
         # TODO: Only hgsubrepo instances have `_repo`, so figure out how to
         # infer and possibly set lfstatus at the top of this function.  That
         # would allow only hgsubrepos to set this, instead of the current scheme
         # where the parent sets this for the child.
         with (
             hasattr(sub, '_repo')
             and lfstatus(sub._repo)
             or util.nullcontextmanager()
         ):
             sub.archive(archiver, subprefix, submatch, decode)
 
 
 # If a largefile is modified, the change is not reflected in its
 # standin until a commit. cmdutil.bailifchanged() raises an exception
 # if the repo has uncommitted changes. Wrap it to also check if
 # largefiles were changed. This is used by bisect, backout and fetch.
 @eh.wrapfunction(cmdutil, 'bailifchanged')
 def overridebailifchanged(orig, repo, *args, **kwargs):
     orig(repo, *args, **kwargs)
     with lfstatus(repo):
         s = repo.status()
     if s.modified or s.added or s.removed or s.deleted:
         raise error.Abort(_(b'uncommitted changes'))
 
 
 @eh.wrapfunction(cmdutil, 'postcommitstatus')
 def postcommitstatus(orig, repo, *args, **kwargs):
     with lfstatus(repo):
         return orig(repo, *args, **kwargs)
 
 
 @eh.wrapfunction(cmdutil, 'forget')
 def cmdutilforget(
     orig, ui, repo, match, prefix, uipathfn, explicitonly, dryrun, interactive
 ):
     normalmatcher = composenormalfilematcher(match, repo[None].manifest())
     bad, forgot = orig(
         ui,
         repo,
         normalmatcher,
         prefix,
         uipathfn,
         explicitonly,
         dryrun,
         interactive,
     )
     m = composelargefilematcher(match, repo[None].manifest())
 
     with lfstatus(repo):
         s = repo.status(match=m, clean=True)
     manifest = repo[None].manifest()
     forget = sorted(s.modified + s.added + s.deleted + s.clean)
     forget = [f for f in forget if lfutil.standin(f) in manifest]
 
     for f in forget:
         fstandin = lfutil.standin(f)
         if fstandin not in repo.dirstate and not repo.wvfs.isdir(fstandin):
             ui.warn(
                 _(b'not removing %s: file is already untracked\n') % uipathfn(f)
             )
             bad.append(f)
 
     for f in forget:
         if ui.verbose or not m.exact(f):
             ui.status(_(b'removing %s\n') % uipathfn(f))
 
     # Need to lock because standin files are deleted then removed from the
     # repository and we could race in-between.
     with repo.wlock():
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         for f in forget:
             lfdirstate.set_untracked(f)
         lfdirstate.write(repo.currenttransaction())
         standins = [lfutil.standin(f) for f in forget]
         for f in standins:
             repo.wvfs.unlinkpath(f, ignoremissing=True)
         rejected = repo[None].forget(standins)
 
     bad.extend(f for f in rejected if f in m.files())
     forgot.extend(f for f in forget if f not in rejected)
     return bad, forgot
 
 
 def _getoutgoings(repo, other, missing, addfunc):
     """get pairs of filename and largefile hash in outgoing revisions
     in 'missing'.
 
     largefiles already existing on 'other' repository are ignored.
 
     'addfunc' is invoked with each unique pairs of filename and
     largefile hash value.
     """
     knowns = set()
     lfhashes = set()
 
     def dedup(fn, lfhash):
         k = (fn, lfhash)
         if k not in knowns:
             knowns.add(k)
             lfhashes.add(lfhash)
 
     lfutil.getlfilestoupload(repo, missing, dedup)
     if lfhashes:
         lfexists = storefactory.openstore(repo, other).exists(lfhashes)
         for fn, lfhash in knowns:
             if not lfexists[lfhash]:  # lfhash doesn't exist on "other"
                 addfunc(fn, lfhash)
 
 
 def outgoinghook(ui, repo, other, opts, missing):
     if opts.pop(b'large', None):
         lfhashes = set()
         if ui.debugflag:
             toupload = {}
 
             def addfunc(fn, lfhash):
                 if fn not in toupload:
                     toupload[fn] = []  # pytype: disable=unsupported-operands
                 toupload[fn].append(lfhash)
                 lfhashes.add(lfhash)
 
             def showhashes(fn):
                 for lfhash in sorted(toupload[fn]):
                     ui.debug(b'    %s\n' % lfhash)
 
         else:
             toupload = set()
 
             def addfunc(fn, lfhash):
                 toupload.add(fn)
                 lfhashes.add(lfhash)
 
             def showhashes(fn):
                 pass
 
         _getoutgoings(repo, other, missing, addfunc)
 
         if not toupload:
             ui.status(_(b'largefiles: no files to upload\n'))
         else:
             ui.status(
                 _(b'largefiles to upload (%d entities):\n') % (len(lfhashes))
             )
             for file in sorted(toupload):
                 ui.status(lfutil.splitstandin(file) + b'\n')
                 showhashes(file)
             ui.status(b'\n')
 
 
 @eh.wrapcommand(
     b'outgoing', opts=[(b'', b'large', None, _(b'display outgoing largefiles'))]
 )
 def _outgoingcmd(orig, *args, **kwargs):
     # Nothing to do here other than add the extra help option- the hook above
     # processes it.
     return orig(*args, **kwargs)
 
 
 def summaryremotehook(ui, repo, opts, changes):
     largeopt = opts.get(b'large', False)
     if changes is None:
         if largeopt:
             return (False, True)  # only outgoing check is needed
         else:
             return (False, False)
     elif largeopt:
         url, branch, peer, outgoing = changes[1]
         if peer is None:
             # i18n: column positioning for "hg summary"
             ui.status(_(b'largefiles: (no remote repo)\n'))
             return
 
         toupload = set()
         lfhashes = set()
 
         def addfunc(fn, lfhash):
             toupload.add(fn)
             lfhashes.add(lfhash)
 
         _getoutgoings(repo, peer, outgoing.missing, addfunc)
 
         if not toupload:
             # i18n: column positioning for "hg summary"
             ui.status(_(b'largefiles: (no files to upload)\n'))
         else:
             # i18n: column positioning for "hg summary"
             ui.status(
                 _(b'largefiles: %d entities for %d files to upload\n')
                 % (len(lfhashes), len(toupload))
             )
 
 
 @eh.wrapcommand(
     b'summary', opts=[(b'', b'large', None, _(b'display outgoing largefiles'))]
 )
 def overridesummary(orig, ui, repo, *pats, **opts):
     with lfstatus(repo):
         orig(ui, repo, *pats, **opts)
 
 
 @eh.wrapfunction(scmutil, 'addremove')
 def scmutiladdremove(
     orig,
     repo,
     matcher,
     prefix,
     uipathfn,
     opts=None,
     open_tr=None,
 ):
     if opts is None:
         opts = {}
     if not lfutil.islfilesrepo(repo):
         return orig(repo, matcher, prefix, uipathfn, opts, open_tr=open_tr)
 
     # open the transaction and changing_files context
     if open_tr is not None:
         open_tr()
 
     # Get the list of missing largefiles so we can remove them
     with repo.dirstate.running_status(repo):
         lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
         unsure, s, mtime_boundary = lfdirstate.status(
             matchmod.always(),
             subrepos=[],
             ignored=False,
             clean=False,
             unknown=False,
         )
 
     # Call into the normal remove code, but the removing of the standin, we want
     # to have handled by original addremove.  Monkey patching here makes sure
     # we don't remove the standin in the largefiles code, preventing a very
     # confused state later.
     if s.deleted:
         m = copy.copy(matcher)
+        m._was_tampered_with = True
 
         # The m._files and m._map attributes are not changed to the deleted list
         # because that affects the m.exact() test, which in turn governs whether
         # or not the file name is printed, and how.  Simply limit the original
         # matches to those in the deleted status list.
         matchfn = m.matchfn
         m.matchfn = lambda f: f in s.deleted and matchfn(f)
 
         removelargefiles(
             repo.ui,
             repo,
             True,
             m,
             uipathfn,
             opts.get(b'dry_run'),
             **pycompat.strkwargs(opts)
         )
     # Call into the normal add code, and any files that *should* be added as
     # largefiles will be
     added, bad = addlargefiles(
         repo.ui, repo, True, matcher, uipathfn, **pycompat.strkwargs(opts)
     )
     # Now that we've handled largefiles, hand off to the original addremove
     # function to take care of the rest.  Make sure it doesn't do anything with
     # largefiles by passing a matcher that will ignore them.
     matcher = composenormalfilematcher(matcher, repo[None].manifest(), added)
 
     return orig(repo, matcher, prefix, uipathfn, opts, open_tr=open_tr)
 
 
 # Calling purge with --all will cause the largefiles to be deleted.
 # Override repo.status to prevent this from happening.
 @eh.wrapcommand(b'purge')
 def overridepurge(orig, ui, repo, *dirs, **opts):
     # XXX Monkey patching a repoview will not work. The assigned attribute will
     # be set on the unfiltered repo, but we will only lookup attributes in the
     # unfiltered repo if the lookup in the repoview object itself fails. As the
     # monkey patched method exists on the repoview class the lookup will not
     # fail. As a result, the original version will shadow the monkey patched
     # one, defeating the monkey patch.
     #
     # As a work around we use an unfiltered repo here. We should do something
     # cleaner instead.
     repo = repo.unfiltered()
     oldstatus = repo.status
 
     def overridestatus(
         node1=b'.',
         node2=None,
         match=None,
         ignored=False,
         clean=False,
         unknown=False,
         listsubrepos=False,
     ):
         r = oldstatus(
             node1, node2, match, ignored, clean, unknown, listsubrepos
         )
         lfdirstate = lfutil.openlfdirstate(ui, repo)
         unknown = [
             f for f in r.unknown if not lfdirstate.get_entry(f).any_tracked
         ]
         ignored = [
             f for f in r.ignored if not lfdirstate.get_entry(f).any_tracked
         ]
         return scmutil.status(
             r.modified, r.added, r.removed, r.deleted, unknown, ignored, r.clean
         )
 
     repo.status = overridestatus
     orig(ui, repo, *dirs, **opts)
     repo.status = oldstatus
 
 
 @eh.wrapcommand(b'rollback')
 def overriderollback(orig, ui, repo, **opts):
     with repo.wlock():
         before = repo.dirstate.parents()
         orphans = {
             f
             for f in repo.dirstate
             if lfutil.isstandin(f) and not repo.dirstate.get_entry(f).removed
         }
         result = orig(ui, repo, **opts)
         after = repo.dirstate.parents()
         if before == after:
             return result  # no need to restore standins
 
         pctx = repo[b'.']
         for f in repo.dirstate:
             if lfutil.isstandin(f):
                 orphans.discard(f)
                 if repo.dirstate.get_entry(f).removed:
                     repo.wvfs.unlinkpath(f, ignoremissing=True)
                 elif f in pctx:
                     fctx = pctx[f]
                     repo.wwrite(f, fctx.data(), fctx.flags())
                 else:
                     # content of standin is not so important in 'a',
                     # 'm' or 'n' (coming from the 2nd parent) cases
                     lfutil.writestandin(repo, f, b'', False)
         for standin in orphans:
             repo.wvfs.unlinkpath(standin, ignoremissing=True)
 
     return result
 
 
 @eh.wrapcommand(b'transplant', extension=b'transplant')
 def overridetransplant(orig, ui, repo, *revs, **opts):
     resuming = opts.get('continue')
     repo._lfcommithooks.append(lfutil.automatedcommithook(resuming))
     repo._lfstatuswriters.append(lambda *msg, **opts: None)
     try:
         result = orig(ui, repo, *revs, **opts)
     finally:
         repo._lfstatuswriters.pop()
         repo._lfcommithooks.pop()
     return result
 
 
 @eh.wrapcommand(b'cat')
 def overridecat(orig, ui, repo, file1, *pats, **opts):
     ctx = logcmdutil.revsingle(repo, opts.get('rev'))
     err = 1
     notbad = set()
     m = scmutil.match(ctx, (file1,) + pats, pycompat.byteskwargs(opts))
+    m._was_tampered_with = True
     origmatchfn = m.matchfn
 
     def lfmatchfn(f):
         if origmatchfn(f):
             return True
         lf = lfutil.splitstandin(f)
         if lf is None:
             return False
         notbad.add(lf)
         return origmatchfn(lf)
 
     m.matchfn = lfmatchfn
     origbadfn = m.bad
 
     def lfbadfn(f, msg):
         if not f in notbad:
             origbadfn(f, msg)
 
     m.bad = lfbadfn
 
     origvisitdirfn = m.visitdir
 
     def lfvisitdirfn(dir):
         if dir == lfutil.shortname:
             return True
         ret = origvisitdirfn(dir)
         if ret:
             return ret
         lf = lfutil.splitstandin(dir)
         if lf is None:
             return False
         return origvisitdirfn(lf)
 
     m.visitdir = lfvisitdirfn
 
     for f in ctx.walk(m):
         with cmdutil.makefileobj(ctx, opts.get('output'), pathname=f) as fp:
             lf = lfutil.splitstandin(f)
             if lf is None or origmatchfn(f):
                 # duplicating unreachable code from commands.cat
                 data = ctx[f].data()
                 if opts.get('decode'):
                     data = repo.wwritedata(f, data)
                 fp.write(data)
             else:
                 hash = lfutil.readasstandin(ctx[f])
                 if not lfutil.inusercache(repo.ui, hash):
                     store = storefactory.openstore(repo)
                     success, missing = store.get([(lf, hash)])
                     if len(success) != 1:
                         raise error.Abort(
                             _(
                                 b'largefile %s is not in cache and could not be '
                                 b'downloaded'
                             )
                             % lf
                         )
                 path = lfutil.usercachepath(repo.ui, hash)
                 with open(path, b"rb") as fpin:
                     for chunk in util.filechunkiter(fpin):
                         fp.write(chunk)
         err = 0
     return err
 
 
 @eh.wrapfunction(merge, '_update')
 def mergeupdate(orig, repo, node, branchmerge, force, *args, **kwargs):
     matcher = kwargs.get('matcher', None)
     # note if this is a partial update
     partial = matcher and not matcher.always()
     with repo.wlock(), repo.dirstate.changing_parents(repo):
         # branch |       |         |
         #  merge | force | partial | action
         # -------+-------+---------+--------------
         #    x   |   x   |    x    | linear-merge
         #    o   |   x   |    x    | branch-merge
         #    x   |   o   |    x    | overwrite (as clean update)
         #    o   |   o   |    x    | force-branch-merge (*1)
         #    x   |   x   |    o    |   (*)
         #    o   |   x   |    o    |   (*)
         #    x   |   o   |    o    | overwrite (as revert)
         #    o   |   o   |    o    |   (*)
         #
         # (*) don't care
         # (*1) deprecated, but used internally (e.g: "rebase --collapse")
         with repo.dirstate.running_status(repo):
             lfdirstate = lfutil.openlfdirstate(repo.ui, repo)
             unsure, s, mtime_boundary = lfdirstate.status(
                 matchmod.always(),
                 subrepos=[],
                 ignored=False,
                 clean=True,
                 unknown=False,
             )
         oldclean = set(s.clean)
         pctx = repo[b'.']
         dctx = repo[node]
         for lfile in unsure + s.modified:
             lfileabs = repo.wvfs.join(lfile)
             if not repo.wvfs.exists(lfileabs):
                 continue
             lfhash = lfutil.hashfile(lfileabs)
             standin = lfutil.standin(lfile)
             lfutil.writestandin(
                 repo, standin, lfhash, lfutil.getexecutable(lfileabs)
             )
             if standin in pctx and lfhash == lfutil.readasstandin(
                 pctx[standin]
             ):
                 oldclean.add(lfile)
         for lfile in s.added:
             fstandin = lfutil.standin(lfile)
             if fstandin not in dctx:
                 # in this case, content of standin file is meaningless
                 # (in dctx, lfile is unknown, or normal file)
                 continue
             lfutil.updatestandin(repo, lfile, fstandin)
         # mark all clean largefiles as dirty, just in case the update gets
         # interrupted before largefiles and lfdirstate are synchronized
         for lfile in oldclean:
             entry = lfdirstate.get_entry(lfile)
             lfdirstate.hacky_extension_update_file(
                 lfile,
                 wc_tracked=entry.tracked,
                 p1_tracked=entry.p1_tracked,
                 p2_info=entry.p2_info,
                 possibly_dirty=True,
             )
         lfdirstate.write(repo.currenttransaction())
 
         oldstandins = lfutil.getstandinsstate(repo)
         wc = kwargs.get('wc')
         if wc and wc.isinmemory():
             # largefiles is not a good candidate for in-memory merge (large
             # files, custom dirstate, matcher usage).
             raise error.ProgrammingError(
                 b'largefiles is not compatible with in-memory merge'
             )
         result = orig(repo, node, branchmerge, force, *args, **kwargs)
 
         newstandins = lfutil.getstandinsstate(repo)
         filelist = lfutil.getlfilestoupdate(oldstandins, newstandins)
 
         # to avoid leaving all largefiles as dirty and thus rehash them, mark
         # all the ones that didn't change as clean
         for lfile in oldclean.difference(filelist):
             lfdirstate.update_file(lfile, p1_tracked=True, wc_tracked=True)
 
         if branchmerge or force or partial:
             filelist.extend(s.deleted + s.removed)
 
         lfcommands.updatelfiles(
             repo.ui, repo, filelist=filelist, normallookup=partial
         )
 
         return result
 
 
 @eh.wrapfunction(scmutil, 'marktouched')
 def scmutilmarktouched(orig, repo, files, *args, **kwargs):
     result = orig(repo, files, *args, **kwargs)
 
     filelist = []
     for f in files:
         lf = lfutil.splitstandin(f)
         if lf is not None:
             filelist.append(lf)
     if filelist:
         lfcommands.updatelfiles(
             repo.ui,
             repo,
             filelist=filelist,
             printmessage=False,
             normallookup=True,
         )
 
     return result
 
 
 @eh.wrapfunction(upgrade_actions, 'preservedrequirements')
 @eh.wrapfunction(upgrade_actions, 'supporteddestrequirements')
 def upgraderequirements(orig, repo):
     reqs = orig(repo)
     if b'largefiles' in repo.requirements:
         reqs.add(b'largefiles')
     return reqs
 
 
 _lfscheme = b'largefile://'
 
 
 @eh.wrapfunction(urlmod, 'open')
 def openlargefile(orig, ui, url_, data=None, **kwargs):
     if url_.startswith(_lfscheme):
         if data:
             msg = b"cannot use data on a 'largefile://' url"
             raise error.ProgrammingError(msg)
         lfid = url_[len(_lfscheme) :]
         return storefactory.getlfile(ui, lfid)
     else:
         return orig(ui, url_, data=data, **kwargs)

hgext/largefiles/reposetup.py

@@ -1,474 +1,476 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
 
 from mercurial.i18n import _
 
 from mercurial import (
     error,
     extensions,
     localrepo,
     match as matchmod,
     scmutil,
     util,
 )
 
 from mercurial.dirstateutils import timestamp
 
 from . import (
     lfcommands,
     lfutil,
 )
 
 
 def reposetup(ui, repo):
     # wire repositories should be given new wireproto functions
     # by "proto.wirereposetup()" via "hg.wirepeersetupfuncs"
     if not repo.local():
         return
 
     class lfilesrepo(repo.__class__):
         # the mark to examine whether "repo" object enables largefiles or not
         _largefilesenabled = True
 
         lfstatus = False
 
         # 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(lfilesrepo, self).__getitem__(changeid)
             if self.lfstatus:
 
                 def files(orig):
                     filenames = orig()
                     return [lfutil.splitstandin(f) or f for f in filenames]
 
                 extensions.wrapfunction(ctx, 'files', files)
 
                 def manifest(orig):
                     man1 = orig()
 
                     class lfilesmanifest(man1.__class__):
                         def __contains__(self, filename):
                             orig = super(lfilesmanifest, self).__contains__
                             return orig(filename) or orig(
                                 lfutil.standin(filename)
                             )
 
                     man1.__class__ = lfilesmanifest
                     return man1
 
                 extensions.wrapfunction(ctx, 'manifest', manifest)
 
                 def filectx(orig, path, fileid=None, filelog=None):
                     try:
                         if filelog is not None:
                             result = orig(path, fileid, filelog)
                         else:
                             result = orig(path, fileid)
                     except error.LookupError:
                         # Adding a null character will cause Mercurial to
                         # identify this as a binary file.
                         if filelog is not None:
                             result = orig(lfutil.standin(path), fileid, filelog)
                         else:
                             result = orig(lfutil.standin(path), fileid)
                         olddata = result.data
                         result.data = lambda: olddata() + b'\0'
                     return result
 
                 extensions.wrapfunction(ctx, 'filectx', filectx)
 
             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.
         # XXX large file status is buggy when used on repo proxy.
         # XXX this needs to be investigated.
         @localrepo.unfilteredmethod
         def status(
             self,
             node1=b'.',
             node2=None,
             match=None,
             ignored=False,
             clean=False,
             unknown=False,
             listsubrepos=False,
         ):
             listignored, listclean, listunknown = ignored, clean, unknown
             orig = super(lfilesrepo, self).status
             if not self.lfstatus:
                 return orig(
                     node1,
                     node2,
                     match,
                     listignored,
                     listclean,
                     listunknown,
                     listsubrepos,
                 )
 
             # some calls in this function rely on the old version of status
             self.lfstatus = False
             ctx1 = self[node1]
             ctx2 = self[node2]
             working = ctx2.rev() is None
             parentworking = working and ctx1 == self[b'.']
 
             if match is None:
                 match = matchmod.always()
 
             try:
                 # updating the dirstate is optional
                 # so we don't wait on the lock
                 wlock = self.wlock(False)
                 gotlock = True
             except error.LockError:
                 wlock = util.nullcontextmanager()
                 gotlock = False
             with wlock, self.dirstate.running_status(self):
 
                 # First check if paths or patterns were specified on the
                 # command line.  If there were, and they don't match any
                 # largefiles, we should just bail here and let super
                 # handle it -- thus gaining a big performance boost.
                 lfdirstate = lfutil.openlfdirstate(ui, self)
                 if not match.always():
                     for f in lfdirstate:
                         if match(f):
                             break
                     else:
                         return orig(
                             node1,
                             node2,
                             match,
                             listignored,
                             listclean,
                             listunknown,
                             listsubrepos,
                         )
 
                 # Create a copy of match that matches standins instead
                 # of largefiles.
                 def tostandins(files):
                     if not working:
                         return files
                     newfiles = []
                     dirstate = self.dirstate
                     for f in files:
                         sf = lfutil.standin(f)
                         if sf in dirstate:
                             newfiles.append(sf)
                         elif dirstate.hasdir(sf):
                             # Directory entries could be regular or
                             # standin, check both
                             newfiles.extend((f, sf))
                         else:
                             newfiles.append(f)
                     return newfiles
 
                 m = copy.copy(match)
+                m._was_tampered_with = True
                 m._files = tostandins(m._files)
 
                 result = orig(
                     node1, node2, m, ignored, clean, unknown, listsubrepos
                 )
                 if working:
 
                     def sfindirstate(f):
                         sf = lfutil.standin(f)
                         dirstate = self.dirstate
                         return sf in dirstate or dirstate.hasdir(sf)
 
+                    match._was_tampered_with = True
                     match._files = [f for f in match._files if sfindirstate(f)]
                     # Don't waste time getting the ignored and unknown
                     # files from lfdirstate
                     unsure, s, mtime_boundary = lfdirstate.status(
                         match,
                         subrepos=[],
                         ignored=False,
                         clean=listclean,
                         unknown=False,
                     )
                     (modified, added, removed, deleted, clean) = (
                         s.modified,
                         s.added,
                         s.removed,
                         s.deleted,
                         s.clean,
                     )
                     if parentworking:
                         wctx = repo[None]
                         for lfile in unsure:
                             standin = lfutil.standin(lfile)
                             if standin not in ctx1:
                                 # from second parent
                                 modified.append(lfile)
                             elif lfutil.readasstandin(
                                 ctx1[standin]
                             ) != lfutil.hashfile(self.wjoin(lfile)):
                                 modified.append(lfile)
                             else:
                                 if listclean:
                                     clean.append(lfile)
                                 s = wctx[lfile].lstat()
                                 mode = s.st_mode
                                 size = s.st_size
                                 mtime = timestamp.reliable_mtime_of(
                                     s, mtime_boundary
                                 )
                                 if mtime is not None:
                                     cache_data = (mode, size, mtime)
                                     lfdirstate.set_clean(lfile, cache_data)
                     else:
                         tocheck = unsure + modified + added + clean
                         modified, added, clean = [], [], []
                         checkexec = self.dirstate._checkexec
 
                         for lfile in tocheck:
                             standin = lfutil.standin(lfile)
                             if standin in ctx1:
                                 abslfile = self.wjoin(lfile)
                                 if (
                                     lfutil.readasstandin(ctx1[standin])
                                     != lfutil.hashfile(abslfile)
                                 ) or (
                                     checkexec
                                     and (b'x' in ctx1.flags(standin))
                                     != bool(lfutil.getexecutable(abslfile))
                                 ):
                                     modified.append(lfile)
                                 elif listclean:
                                     clean.append(lfile)
                             else:
                                 added.append(lfile)
 
                         # at this point, 'removed' contains largefiles
                         # marked as 'R' in the working context.
                         # then, largefiles not managed also in the target
                         # context should be excluded from 'removed'.
                         removed = [
                             lfile
                             for lfile in removed
                             if lfutil.standin(lfile) in ctx1
                         ]
 
                     # Standins no longer found in lfdirstate have been deleted
                     for standin in ctx1.walk(lfutil.getstandinmatcher(self)):
                         lfile = lfutil.splitstandin(standin)
                         if not match(lfile):
                             continue
                         if lfile not in lfdirstate:
                             deleted.append(lfile)
                             # Sync "largefile has been removed" back to the
                             # standin. Removing a file as a side effect of
                             # running status is gross, but the alternatives (if
                             # any) are worse.
                             self.wvfs.unlinkpath(standin, ignoremissing=True)
 
                     # 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 they are still in
                     # lfdirstate. This happens in merges where files
                     # change type.
                     removed = [f for f in removed if f not in self.dirstate]
                     result[2] = [f for f in result[2] if f not in lfdirstate]
 
                     lfiles = set(lfdirstate)
                     # Unknown files
                     result[4] = set(result[4]).difference(lfiles)
                     # Ignored files
                     result[5] = set(result[5]).difference(lfiles)
                     # combine normal files and largefiles
                     normals = [
                         [fn for fn in filelist if not lfutil.isstandin(fn)]
                         for filelist in result
                     ]
                     lfstatus = (
                         modified,
                         added,
                         removed,
                         deleted,
                         [],
                         [],
                         clean,
                     )
                     result = [
                         sorted(list1 + list2)
                         for (list1, list2) in zip(normals, lfstatus)
                     ]
                 else:  # not against working directory
                     result = [
                         [lfutil.splitstandin(f) or f for f in items]
                         for items in result
                     ]
 
                 if gotlock:
                     lfdirstate.write(self.currenttransaction())
                 else:
                     lfdirstate.invalidate()
 
             self.lfstatus = True
             return scmutil.status(*result)
 
         def commitctx(self, ctx, *args, **kwargs):
             node = super(lfilesrepo, self).commitctx(ctx, *args, **kwargs)
 
             class lfilesctx(ctx.__class__):
                 def markcommitted(self, node):
                     orig = super(lfilesctx, self).markcommitted
                     return lfutil.markcommitted(orig, self, node)
 
             ctx.__class__ = lfilesctx
             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=b"",
             user=None,
             date=None,
             match=None,
             force=False,
             editor=False,
             extra=None,
         ):
             if extra is None:
                 extra = {}
             orig = super(lfilesrepo, self).commit
 
             with self.wlock():
                 lfcommithook = self._lfcommithooks[-1]
                 match = lfcommithook(self, match)
                 result = orig(
                     text=text,
                     user=user,
                     date=date,
                     match=match,
                     force=force,
                     editor=editor,
                     extra=extra,
                 )
                 return result
 
         # TODO: _subdirlfs should be moved into "lfutil.py", because
         # it is referred only from "lfutil.updatestandinsbymatch"
         def _subdirlfs(self, files, lfiles):
             """
             Adjust matched file list
             If we pass a directory to commit whose only committable files
             are largefiles, the core commit code aborts before finding
             the largefiles.
             So we do the following:
             For directories that only have largefiles as matches,
             we explicitly add the largefiles to the match list and remove
             the directory.
             In other cases, we leave the match list unmodified.
             """
             actualfiles = []
             dirs = []
             regulars = []
 
             for f in files:
                 if lfutil.isstandin(f + b'/'):
                     raise error.Abort(
                         _(b'file "%s" is a largefile standin') % f,
                         hint=b'commit the largefile itself instead',
                     )
                 # Scan directories
                 if self.wvfs.isdir(f):
                     dirs.append(f)
                 else:
                     regulars.append(f)
 
             for f in dirs:
                 matcheddir = False
                 d = self.dirstate.normalize(f) + b'/'
                 # Check for matched normal files
                 for mf in regulars:
                     if self.dirstate.normalize(mf).startswith(d):
                         actualfiles.append(f)
                         matcheddir = True
                         break
                 if not matcheddir:
                     # If no normal match, manually append
                     # any matching largefiles
                     for lf in lfiles:
                         if self.dirstate.normalize(lf).startswith(d):
                             actualfiles.append(lf)
                             if not matcheddir:
                                 # There may still be normal files in the dir, so
                                 # add a directory to the list, which
                                 # forces status/dirstate to walk all files and
                                 # call the match function on the matcher, even
                                 # on case sensitive filesystems.
                                 actualfiles.append(b'.')
                                 matcheddir = True
                 # Nothing in dir, so readd it
                 # and let commit reject it
                 if not matcheddir:
                     actualfiles.append(f)
 
             # Always add normal files
             actualfiles += regulars
             return actualfiles
 
     repo.__class__ = lfilesrepo
 
     # stack of hooks being executed before committing.
     # only last element ("_lfcommithooks[-1]") is used for each committing.
     repo._lfcommithooks = [lfutil.updatestandinsbymatch]
 
     # Stack of status writer functions taking "*msg, **opts" arguments
     # like "ui.status()". Only last element ("_lfstatuswriters[-1]")
     # is used to write status out.
     repo._lfstatuswriters = [ui.status]
 
     def prepushoutgoinghook(pushop):
         """Push largefiles for pushop before pushing revisions."""
         lfrevs = pushop.lfrevs
         if lfrevs is None:
             lfrevs = pushop.outgoing.missing
         if lfrevs:
             toupload = set()
             addfunc = lambda fn, lfhash: toupload.add(lfhash)
             lfutil.getlfilestoupload(pushop.repo, lfrevs, addfunc)
             lfcommands.uploadlfiles(ui, pushop.repo, pushop.remote, toupload)
 
     repo.prepushoutgoinghooks.add(b"largefiles", prepushoutgoinghook)
 
     def checkrequireslfiles(ui, repo, **kwargs):
         with repo.lock():
             if b'largefiles' in repo.requirements:
                 return
             marker = lfutil.shortnameslash
             for entry in repo.store.data_entries():
                 # XXX note that this match is not rooted and can wrongly match
                 # directory ending with ".hglf"
                 if entry.is_revlog and marker in entry.target_id:
                     repo.requirements.add(b'largefiles')
                     scmutil.writereporequirements(repo)
                     break
 
     ui.setconfig(
         b'hooks', b'changegroup.lfiles', checkrequireslfiles, b'largefiles'
     )
     ui.setconfig(b'hooks', b'commit.lfiles', checkrequireslfiles, b'largefiles')

hgext/rebase.py

@@ -1,2281 +1,2281 b''
 # rebase.py - rebasing feature for mercurial
 #
 # Copyright 2008 Stefano Tortarolo <stefano.tortarolo at gmail dot com>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''command to move sets of revisions to a different ancestor
 
 This extension lets you rebase changesets in an existing Mercurial
 repository.
 
 For more information:
 https://mercurial-scm.org/wiki/RebaseExtension
 '''
 
 
 import os
 
 from mercurial.i18n import _
 from mercurial.node import (
     nullrev,
     short,
     wdirrev,
 )
 from mercurial.pycompat import open
 from mercurial import (
     bookmarks,
     cmdutil,
     commands,
     copies,
     destutil,
     error,
     extensions,
     logcmdutil,
     merge as mergemod,
     mergestate as mergestatemod,
     mergeutil,
     obsolete,
     obsutil,
     patch,
     phases,
     pycompat,
     registrar,
     repair,
     revset,
     revsetlang,
     rewriteutil,
     scmutil,
     smartset,
     state as statemod,
     util,
 )
 
 
 # The following constants are used throughout the rebase module. The ordering of
 # their values must be maintained.
 
 # Indicates that a revision needs to be rebased
 revtodo = -1
 revtodostr = b'-1'
 
 # legacy revstates no longer needed in current code
 # -2: nullmerge, -3: revignored, -4: revprecursor, -5: revpruned
 legacystates = {b'-2', b'-3', b'-4', b'-5'}
 
 cmdtable = {}
 command = registrar.command(cmdtable)
 
 configtable = {}
 configitem = registrar.configitem(configtable)
 configitem(
     b'devel',
     b'rebase.force-in-memory-merge',
     default=False,
 )
 # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
 # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
 # be specifying the version(s) of Mercurial they are tested with, or
 # leave the attribute unspecified.
 testedwith = b'ships-with-hg-core'
 
 
 def _nothingtorebase():
     return 1
 
 
 def _savebranch(ctx, extra):
     extra[b'branch'] = ctx.branch()
 
 
 def _destrebase(repo, sourceset, destspace=None):
     """small wrapper around destmerge to pass the right extra args
 
     Please wrap destutil.destmerge instead."""
     return destutil.destmerge(
         repo,
         action=b'rebase',
         sourceset=sourceset,
         onheadcheck=False,
         destspace=destspace,
     )
 
 
 revsetpredicate = registrar.revsetpredicate()
 
 
 @revsetpredicate(b'_destrebase')
 def _revsetdestrebase(repo, subset, x):
     # ``_rebasedefaultdest()``
 
     # default destination for rebase.
     # # XXX: Currently private because I expect the signature to change.
     # # XXX: - bailing out in case of ambiguity vs returning all data.
     # i18n: "_rebasedefaultdest" is a keyword
     sourceset = None
     if x is not None:
         sourceset = revset.getset(repo, smartset.fullreposet(repo), x)
     return subset & smartset.baseset([_destrebase(repo, sourceset)])
 
 
 @revsetpredicate(b'_destautoorphanrebase')
 def _revsetdestautoorphanrebase(repo, subset, x):
     # ``_destautoorphanrebase()``
 
     # automatic rebase destination for a single orphan revision.
     unfi = repo.unfiltered()
     obsoleted = unfi.revs(b'obsolete()')
 
     src = revset.getset(repo, subset, x).first()
 
     # Empty src or already obsoleted - Do not return a destination
     if not src or src in obsoleted:
         return smartset.baseset()
     dests = destutil.orphanpossibledestination(repo, src)
     if len(dests) > 1:
         raise error.StateError(
             _(b"ambiguous automatic rebase: %r could end up on any of %r")
             % (src, dests)
         )
     # We have zero or one destination, so we can just return here.
     return smartset.baseset(dests)
 
 
 def _ctxdesc(ctx):
     """short description for a context"""
     return cmdutil.format_changeset_summary(
         ctx.repo().ui, ctx, command=b'rebase'
     )
 
 
 class rebaseruntime:
     """This class is a container for rebase runtime state"""
 
     def __init__(self, repo, ui, inmemory=False, dryrun=False, opts=None):
         if opts is None:
             opts = {}
 
         # prepared: whether we have rebasestate prepared or not. Currently it
         # decides whether "self.repo" is unfiltered or not.
         # The rebasestate has explicit hash to hash instructions not depending
         # on visibility. If rebasestate exists (in-memory or on-disk), use
         # unfiltered repo to avoid visibility issues.
         # Before knowing rebasestate (i.e. when starting a new rebase (not
         # --continue or --abort)), the original repo should be used so
         # visibility-dependent revsets are correct.
         self.prepared = False
         self.resume = False
         self._repo = repo
 
         self.ui = ui
         self.opts = opts
         self.originalwd = None
         self.external = nullrev
         # Mapping between the old revision id and either what is the new rebased
         # revision or what needs to be done with the old revision. The state
         # dict will be what contains most of the rebase progress state.
         self.state = {}
         self.activebookmark = None
         self.destmap = {}
         self.skipped = set()
 
         self.collapsef = opts.get('collapse', False)
         self.collapsemsg = cmdutil.logmessage(ui, pycompat.byteskwargs(opts))
         self.date = opts.get('date', None)
 
         e = opts.get('extrafn')  # internal, used by e.g. hgsubversion
         self.extrafns = [rewriteutil.preserve_extras_on_rebase]
         if e:
             self.extrafns = [e]
 
         self.backupf = ui.configbool(b'rewrite', b'backup-bundle')
         self.keepf = opts.get('keep', False)
         self.keepbranchesf = opts.get('keepbranches', False)
         self.skipemptysuccessorf = rewriteutil.skip_empty_successor(
             repo.ui, b'rebase'
         )
         self.obsolete_with_successor_in_destination = {}
         self.obsolete_with_successor_in_rebase_set = set()
         self.inmemory = inmemory
         self.dryrun = dryrun
         self.stateobj = statemod.cmdstate(repo, b'rebasestate')
 
     @property
     def repo(self):
         if self.prepared:
             return self._repo.unfiltered()
         else:
             return self._repo
 
     def storestatus(self, tr=None):
         """Store the current status to allow recovery"""
         if tr:
             tr.addfilegenerator(
                 b'rebasestate',
                 (b'rebasestate',),
                 self._writestatus,
                 location=b'plain',
             )
         else:
             with self.repo.vfs(b"rebasestate", b"w") as f:
                 self._writestatus(f)
 
     def _writestatus(self, f):
         repo = self.repo
         assert repo.filtername is None
         f.write(repo[self.originalwd].hex() + b'\n')
         # was "dest". we now write dest per src root below.
         f.write(b'\n')
         f.write(repo[self.external].hex() + b'\n')
         f.write(b'%d\n' % int(self.collapsef))
         f.write(b'%d\n' % int(self.keepf))
         f.write(b'%d\n' % int(self.keepbranchesf))
         f.write(b'%s\n' % (self.activebookmark or b''))
         destmap = self.destmap
         for d, v in self.state.items():
             oldrev = repo[d].hex()
             if v >= 0:
                 newrev = repo[v].hex()
             else:
                 newrev = b"%d" % v
             destnode = repo[destmap[d]].hex()
             f.write(b"%s:%s:%s\n" % (oldrev, newrev, destnode))
         repo.ui.debug(b'rebase status stored\n')
 
     def restorestatus(self):
         """Restore a previously stored status"""
         if not self.stateobj.exists():
             cmdutil.wrongtooltocontinue(self.repo, _(b'rebase'))
 
         data = self._read()
         self.repo.ui.debug(b'rebase status resumed\n')
 
         self.originalwd = data[b'originalwd']
         self.destmap = data[b'destmap']
         self.state = data[b'state']
         self.skipped = data[b'skipped']
         self.collapsef = data[b'collapse']
         self.keepf = data[b'keep']
         self.keepbranchesf = data[b'keepbranches']
         self.external = data[b'external']
         self.activebookmark = data[b'activebookmark']
 
     def _read(self):
         self.prepared = True
         repo = self.repo
         assert repo.filtername is None
         data = {
             b'keepbranches': None,
             b'collapse': None,
             b'activebookmark': None,
             b'external': nullrev,
             b'keep': None,
             b'originalwd': None,
         }
         legacydest = None
         state = {}
         destmap = {}
 
         if True:
             f = repo.vfs(b"rebasestate")
             for i, l in enumerate(f.read().splitlines()):
                 if i == 0:
                     data[b'originalwd'] = repo[l].rev()
                 elif i == 1:
                     # this line should be empty in newer version. but legacy
                     # clients may still use it
                     if l:
                         legacydest = repo[l].rev()
                 elif i == 2:
                     data[b'external'] = repo[l].rev()
                 elif i == 3:
                     data[b'collapse'] = bool(int(l))
                 elif i == 4:
                     data[b'keep'] = bool(int(l))
                 elif i == 5:
                     data[b'keepbranches'] = bool(int(l))
                 elif i == 6 and not (len(l) == 81 and b':' in l):
                     # line 6 is a recent addition, so for backwards
                     # compatibility check that the line doesn't look like the
                     # oldrev:newrev lines
                     data[b'activebookmark'] = l
                 else:
                     args = l.split(b':')
                     oldrev = repo[args[0]].rev()
                     newrev = args[1]
                     if newrev in legacystates:
                         continue
                     if len(args) > 2:
                         destrev = repo[args[2]].rev()
                     else:
                         destrev = legacydest
                     destmap[oldrev] = destrev
                     if newrev == revtodostr:
                         state[oldrev] = revtodo
                         # Legacy compat special case
                     else:
                         state[oldrev] = repo[newrev].rev()
 
         if data[b'keepbranches'] is None:
             raise error.Abort(_(b'.hg/rebasestate is incomplete'))
 
         data[b'destmap'] = destmap
         data[b'state'] = state
         skipped = set()
         # recompute the set of skipped revs
         if not data[b'collapse']:
             seen = set(destmap.values())
             for old, new in sorted(state.items()):
                 if new != revtodo and new in seen:
                     skipped.add(old)
                 seen.add(new)
         data[b'skipped'] = skipped
         repo.ui.debug(
             b'computed skipped revs: %s\n'
             % (b' '.join(b'%d' % r for r in sorted(skipped)) or b'')
         )
 
         return data
 
     def _handleskippingobsolete(self):
         """Compute structures necessary for skipping obsolete revisions"""
         if self.keepf:
             return
         if not self.ui.configbool(b'experimental', b'rebaseskipobsolete'):
             return
         obsoleteset = {r for r in self.state if self.repo[r].obsolete()}
         (
             self.obsolete_with_successor_in_destination,
             self.obsolete_with_successor_in_rebase_set,
         ) = _compute_obsolete_sets(self.repo, obsoleteset, self.destmap)
         skippedset = set(self.obsolete_with_successor_in_destination)
         skippedset.update(self.obsolete_with_successor_in_rebase_set)
         _checkobsrebase(self.repo, self.ui, obsoleteset, skippedset)
         if obsolete.isenabled(self.repo, obsolete.allowdivergenceopt):
             self.obsolete_with_successor_in_rebase_set = set()
         else:
             for rev in self.repo.revs(
                 b'descendants(%ld) and not %ld',
                 self.obsolete_with_successor_in_rebase_set,
                 self.obsolete_with_successor_in_rebase_set,
             ):
                 self.state.pop(rev, None)
                 self.destmap.pop(rev, None)
 
     def _prepareabortorcontinue(
         self, isabort, backup=True, suppwarns=False, dryrun=False, confirm=False
     ):
         self.resume = True
         try:
             self.restorestatus()
             # Calculate self.obsolete_* sets
             self._handleskippingobsolete()
             self.collapsemsg = restorecollapsemsg(self.repo, isabort)
         except error.RepoLookupError:
             if isabort:
                 clearstatus(self.repo)
                 clearcollapsemsg(self.repo)
                 self.repo.ui.warn(
                     _(
                         b'rebase aborted (no revision is removed,'
                         b' only broken state is cleared)\n'
                     )
                 )
                 return 0
             else:
                 msg = _(b'cannot continue inconsistent rebase')
                 hint = _(b'use "hg rebase --abort" to clear broken state')
                 raise error.Abort(msg, hint=hint)
 
         if isabort:
             backup = backup and self.backupf
             return self._abort(
                 backup=backup,
                 suppwarns=suppwarns,
                 dryrun=dryrun,
                 confirm=confirm,
             )
 
     def _preparenewrebase(self, destmap):
         if not destmap:
             return _nothingtorebase()
 
         result = buildstate(self.repo, destmap, self.collapsef)
 
         if not result:
             # Empty state built, nothing to rebase
             self.ui.status(_(b'nothing to rebase\n'))
             return _nothingtorebase()
 
         (self.originalwd, self.destmap, self.state) = result
         if self.collapsef:
             dests = set(self.destmap.values())
             if len(dests) != 1:
                 raise error.InputError(
                     _(b'--collapse does not work with multiple destinations')
                 )
             destrev = next(iter(dests))
             destancestors = self.repo.changelog.ancestors(
                 [destrev], inclusive=True
             )
             self.external = externalparent(self.repo, self.state, destancestors)
 
         for destrev in sorted(set(destmap.values())):
             dest = self.repo[destrev]
             if dest.closesbranch() and not self.keepbranchesf:
                 self.ui.status(_(b'reopening closed branch head %s\n') % dest)
 
         # Calculate self.obsolete_* sets
         self._handleskippingobsolete()
 
         if not self.keepf:
             rebaseset = set(destmap.keys())
             rebaseset -= set(self.obsolete_with_successor_in_destination)
             rebaseset -= self.obsolete_with_successor_in_rebase_set
             # We have our own divergence-checking in the rebase extension
             overrides = {}
             if obsolete.isenabled(self.repo, obsolete.createmarkersopt):
                 overrides = {
                     (b'experimental', b'evolution.allowdivergence'): b'true'
                 }
             try:
                 with self.ui.configoverride(overrides):
                     rewriteutil.precheck(self.repo, rebaseset, action=b'rebase')
             except error.Abort as e:
                 if e.hint is None:
                     e.hint = _(b'use --keep to keep original changesets')
                 raise e
 
         self.prepared = True
 
     def _assignworkingcopy(self):
         if self.inmemory:
             from mercurial.context import overlayworkingctx
 
             self.wctx = overlayworkingctx(self.repo)
             self.repo.ui.debug(b"rebasing in memory\n")
         else:
             self.wctx = self.repo[None]
             self.repo.ui.debug(b"rebasing on disk\n")
         self.repo.ui.log(
             b"rebase",
             b"using in-memory rebase: %r\n",
             self.inmemory,
             rebase_imm_used=self.inmemory,
         )
 
     def _performrebase(self, tr):
         self._assignworkingcopy()
         repo, ui = self.repo, self.ui
         if self.keepbranchesf:
             # insert _savebranch at the start of extrafns so if
             # there's a user-provided extrafn it can clobber branch if
             # desired
             self.extrafns.insert(0, _savebranch)
             if self.collapsef:
                 branches = set()
                 for rev in self.state:
                     branches.add(repo[rev].branch())
                     if len(branches) > 1:
                         raise error.InputError(
                             _(b'cannot collapse multiple named branches')
                         )
 
         # Keep track of the active bookmarks in order to reset them later
         self.activebookmark = self.activebookmark or repo._activebookmark
         if self.activebookmark:
             bookmarks.deactivate(repo)
 
         # Store the state before we begin so users can run 'hg rebase --abort'
         # if we fail before the transaction closes.
         self.storestatus()
         if tr:
             # When using single transaction, store state when transaction
             # commits.
             self.storestatus(tr)
 
         cands = [k for k, v in self.state.items() if v == revtodo]
         p = repo.ui.makeprogress(
             _(b"rebasing"), unit=_(b'changesets'), total=len(cands)
         )
 
         def progress(ctx):
             p.increment(item=(b"%d:%s" % (ctx.rev(), ctx)))
 
         for subset in sortsource(self.destmap):
             sortedrevs = self.repo.revs(b'sort(%ld, -topo)', subset)
             for rev in sortedrevs:
                 self._rebasenode(tr, rev, progress)
         p.complete()
         ui.note(_(b'rebase merging completed\n'))
 
     def _concludenode(self, rev, editor, commitmsg=None):
         """Commit the wd changes with parents p1 and p2.
 
         Reuse commit info from rev but also store useful information in extra.
         Return node of committed revision."""
         repo = self.repo
         ctx = repo[rev]
         if commitmsg is None:
             commitmsg = ctx.description()
 
         # Skip replacement if collapsing, as that degenerates to p1 for all
         # nodes.
         if not self.collapsef:
             cl = repo.changelog
             commitmsg = rewriteutil.update_hash_refs(
                 repo,
                 commitmsg,
                 {
                     cl.node(oldrev): [cl.node(newrev)]
                     for oldrev, newrev in self.state.items()
                     if newrev != revtodo
                 },
             )
 
         date = self.date
         if date is None:
             date = ctx.date()
         extra = {}
         if repo.ui.configbool(b'rebase', b'store-source'):
             extra = {b'rebase_source': ctx.hex()}
         for c in self.extrafns:
             c(ctx, extra)
         destphase = max(ctx.phase(), phases.draft)
         overrides = {
             (b'phases', b'new-commit'): destphase,
             (b'ui', b'allowemptycommit'): not self.skipemptysuccessorf,
         }
         with repo.ui.configoverride(overrides, b'rebase'):
             if self.inmemory:
                 newnode = commitmemorynode(
                     repo,
                     wctx=self.wctx,
                     extra=extra,
                     commitmsg=commitmsg,
                     editor=editor,
                     user=ctx.user(),
                     date=date,
                 )
             else:
                 newnode = commitnode(
                     repo,
                     extra=extra,
                     commitmsg=commitmsg,
                     editor=editor,
                     user=ctx.user(),
                     date=date,
                 )
 
             return newnode
 
     def _rebasenode(self, tr, rev, progressfn):
         repo, ui, opts = self.repo, self.ui, self.opts
         ctx = repo[rev]
         desc = _ctxdesc(ctx)
         if self.state[rev] == rev:
             ui.status(_(b'already rebased %s\n') % desc)
         elif rev in self.obsolete_with_successor_in_rebase_set:
             msg = (
                 _(
                     b'note: not rebasing %s and its descendants as '
                     b'this would cause divergence\n'
                 )
                 % desc
             )
             repo.ui.status(msg)
             self.skipped.add(rev)
         elif rev in self.obsolete_with_successor_in_destination:
             succ = self.obsolete_with_successor_in_destination[rev]
             if succ is None:
                 msg = _(b'note: not rebasing %s, it has no successor\n') % desc
             else:
                 succdesc = _ctxdesc(repo[succ])
                 msg = _(
                     b'note: not rebasing %s, already in destination as %s\n'
                 ) % (desc, succdesc)
             repo.ui.status(msg)
             # Make clearrebased aware state[rev] is not a true successor
             self.skipped.add(rev)
             # Record rev as moved to its desired destination in self.state.
             # This helps bookmark and working parent movement.
             dest = max(
                 adjustdest(repo, rev, self.destmap, self.state, self.skipped)
             )
             self.state[rev] = dest
         elif self.state[rev] == revtodo:
             ui.status(_(b'rebasing %s\n') % desc)
             progressfn(ctx)
             p1, p2, base = defineparents(
                 repo,
                 rev,
                 self.destmap,
                 self.state,
                 self.skipped,
                 self.obsolete_with_successor_in_destination,
             )
             if self.resume and self.wctx.p1().rev() == p1:
                 repo.ui.debug(b'resuming interrupted rebase\n')
                 self.resume = False
             else:
                 overrides = {(b'ui', b'forcemerge'): opts.get('tool', b'')}
                 with ui.configoverride(overrides, b'rebase'):
                     try:
                         rebasenode(
                             repo,
                             rev,
                             p1,
                             p2,
                             base,
                             self.collapsef,
                             wctx=self.wctx,
                         )
                     except error.InMemoryMergeConflictsError:
                         if self.dryrun:
                             raise error.ConflictResolutionRequired(b'rebase')
                         if self.collapsef:
                             # TODO: Make the overlayworkingctx reflected
                             # in the working copy here instead of re-raising
                             # so the entire rebase operation is retried.
                             raise
                         ui.status(
                             _(
                                 b"hit merge conflicts; rebasing that "
                                 b"commit again in the working copy\n"
                             )
                         )
                         try:
                             cmdutil.bailifchanged(repo)
                         except error.Abort:
                             clearstatus(repo)
                             clearcollapsemsg(repo)
                             raise
                         self.inmemory = False
                         self._assignworkingcopy()
                         mergemod.update(repo[p1], wc=self.wctx)
                         rebasenode(
                             repo,
                             rev,
                             p1,
                             p2,
                             base,
                             self.collapsef,
                             wctx=self.wctx,
                         )
             if not self.collapsef:
                 merging = p2 != nullrev
                 editform = cmdutil.mergeeditform(merging, b'rebase')
                 editor = cmdutil.getcommiteditor(editform=editform, **opts)
                 # We need to set parents again here just in case we're continuing
                 # a rebase started with an old hg version (before 9c9cfecd4600),
                 # because those old versions would have left us with two dirstate
                 # parents, and we don't want to create a merge commit here (unless
                 # we're rebasing a merge commit).
                 self.wctx.setparents(repo[p1].node(), repo[p2].node())
                 newnode = self._concludenode(rev, editor)
             else:
                 # Skip commit if we are collapsing
                 newnode = None
             # Update the state
             if newnode is not None:
                 self.state[rev] = repo[newnode].rev()
                 ui.debug(b'rebased as %s\n' % short(newnode))
                 if repo[newnode].isempty():
                     ui.warn(
                         _(
                             b'note: created empty successor for %s, its '
                             b'destination already has all its changes\n'
                         )
                         % desc
                     )
             else:
                 if not self.collapsef:
                     ui.warn(
                         _(
                             b'note: not rebasing %s, its destination already '
                             b'has all its changes\n'
                         )
                         % desc
                     )
                     self.skipped.add(rev)
                 self.state[rev] = p1
                 ui.debug(b'next revision set to %d\n' % p1)
         else:
             ui.status(
                 _(b'already rebased %s as %s\n') % (desc, repo[self.state[rev]])
             )
         if not tr:
             # When not using single transaction, store state after each
             # commit is completely done. On InterventionRequired, we thus
             # won't store the status. Instead, we'll hit the "len(parents) == 2"
             # case and realize that the commit was in progress.
             self.storestatus()
 
     def _finishrebase(self):
         repo, ui, opts = self.repo, self.ui, self.opts
         fm = ui.formatter(b'rebase', pycompat.byteskwargs(opts))
         fm.startitem()
         if self.collapsef:
             p1, p2, _base = defineparents(
                 repo,
                 min(self.state),
                 self.destmap,
                 self.state,
                 self.skipped,
                 self.obsolete_with_successor_in_destination,
             )
             editopt = opts.get('edit')
             editform = b'rebase.collapse'
             if self.collapsemsg:
                 commitmsg = self.collapsemsg
             else:
                 commitmsg = b'Collapsed revision'
                 for rebased in sorted(self.state):
                     if rebased not in self.skipped:
                         commitmsg += b'\n* %s' % repo[rebased].description()
                 editopt = True
             editor = cmdutil.getcommiteditor(edit=editopt, editform=editform)
             revtoreuse = max(self.state)
 
             self.wctx.setparents(repo[p1].node(), repo[self.external].node())
             newnode = self._concludenode(
                 revtoreuse, editor, commitmsg=commitmsg
             )
 
             if newnode is not None:
                 newrev = repo[newnode].rev()
                 for oldrev in self.state:
                     self.state[oldrev] = newrev
 
         if b'qtip' in repo.tags():
             updatemq(repo, self.state, self.skipped, **opts)
 
         # restore original working directory
         # (we do this before stripping)
         newwd = self.state.get(self.originalwd, self.originalwd)
         if newwd < 0:
             # original directory is a parent of rebase set root or ignored
             newwd = self.originalwd
         if newwd not in [c.rev() for c in repo[None].parents()]:
             ui.note(_(b"update back to initial working directory parent\n"))
             mergemod.update(repo[newwd])
 
         collapsedas = None
         if self.collapsef and not self.keepf:
             collapsedas = newnode
         clearrebased(
             ui,
             repo,
             self.destmap,
             self.state,
             self.skipped,
             collapsedas,
             self.keepf,
             fm=fm,
             backup=self.backupf,
         )
 
         clearstatus(repo)
         clearcollapsemsg(repo)
 
         ui.note(_(b"rebase completed\n"))
         util.unlinkpath(repo.sjoin(b'undo'), ignoremissing=True)
         if self.skipped:
             skippedlen = len(self.skipped)
             ui.note(_(b"%d revisions have been skipped\n") % skippedlen)
         fm.end()
 
         if (
             self.activebookmark
             and self.activebookmark in repo._bookmarks
             and repo[b'.'].node() == repo._bookmarks[self.activebookmark]
         ):
             bookmarks.activate(repo, self.activebookmark)
 
     def _abort(self, backup=True, suppwarns=False, dryrun=False, confirm=False):
         '''Restore the repository to its original state.'''
 
         repo = self.repo
         try:
             # If the first commits in the rebased set get skipped during the
             # rebase, their values within the state mapping will be the dest
             # rev id. The rebased list must must not contain the dest rev
             # (issue4896)
             rebased = [
                 s
                 for r, s in self.state.items()
                 if s >= 0 and s != r and s != self.destmap[r]
             ]
             immutable = [d for d in rebased if not repo[d].mutable()]
             cleanup = True
             if immutable:
                 repo.ui.warn(
                     _(b"warning: can't clean up public changesets %s\n")
                     % b', '.join(bytes(repo[r]) for r in immutable),
                     hint=_(b"see 'hg help phases' for details"),
                 )
                 cleanup = False
 
             descendants = set()
             if rebased:
                 descendants = set(repo.changelog.descendants(rebased))
             if descendants - set(rebased):
                 repo.ui.warn(
                     _(
                         b"warning: new changesets detected on "
                         b"destination branch, can't strip\n"
                     )
                 )
                 cleanup = False
 
             if cleanup:
 
                 if rebased:
                     strippoints = [
                         c.node() for c in repo.set(b'roots(%ld)', rebased)
                     ]
 
                 updateifonnodes = set(rebased)
                 updateifonnodes.update(self.destmap.values())
 
                 if not confirm:
                     # note: when dry run is set the `rebased` and `destmap`
                     # variables seem to contain "bad" contents, so do not
                     # rely on them. As dryrun does not need this part of
                     # the cleanup, this is "fine"
                     updateifonnodes.add(self.originalwd)
 
                 shouldupdate = repo[b'.'].rev() in updateifonnodes
 
                 # Update away from the rebase if necessary
                 if not dryrun and shouldupdate:
                     mergemod.clean_update(repo[self.originalwd])
 
                 # Strip from the first rebased revision
                 if rebased:
                     repair.strip(repo.ui, repo, strippoints, backup=backup)
 
             if self.activebookmark and self.activebookmark in repo._bookmarks:
                 bookmarks.activate(repo, self.activebookmark)
 
         finally:
             clearstatus(repo)
             clearcollapsemsg(repo)
             if not suppwarns:
                 repo.ui.warn(_(b'rebase aborted\n'))
         return 0
 
 
 @command(
     b'rebase',
     [
         (
             b's',
             b'source',
             [],
             _(b'rebase the specified changesets and their descendants'),
             _(b'REV'),
         ),
         (
             b'b',
             b'base',
             [],
             _(b'rebase everything from branching point of specified changeset'),
             _(b'REV'),
         ),
         (b'r', b'rev', [], _(b'rebase these revisions'), _(b'REV')),
         (
             b'd',
             b'dest',
             b'',
             _(b'rebase onto the specified changeset'),
             _(b'REV'),
         ),
         (b'', b'collapse', False, _(b'collapse the rebased changesets')),
         (
             b'm',
             b'message',
             b'',
             _(b'use text as collapse commit message'),
             _(b'TEXT'),
         ),
         (b'e', b'edit', False, _(b'invoke editor on commit messages')),
         (
             b'l',
             b'logfile',
             b'',
             _(b'read collapse commit message from file'),
             _(b'FILE'),
         ),
         (b'k', b'keep', False, _(b'keep original changesets')),
         (b'', b'keepbranches', False, _(b'keep original branch names')),
         (b'D', b'detach', False, _(b'(DEPRECATED)')),
         (b'i', b'interactive', False, _(b'(DEPRECATED)')),
         (b't', b'tool', b'', _(b'specify merge tool')),
         (b'', b'stop', False, _(b'stop interrupted rebase')),
         (b'c', b'continue', False, _(b'continue an interrupted rebase')),
         (b'a', b'abort', False, _(b'abort an interrupted rebase')),
         (
             b'',
             b'auto-orphans',
             b'',
             _(
                 b'automatically rebase orphan revisions '
                 b'in the specified revset (EXPERIMENTAL)'
             ),
         ),
     ]
     + cmdutil.dryrunopts
     + cmdutil.formatteropts
     + cmdutil.confirmopts,
     _(b'[[-s REV]... | [-b REV]... | [-r REV]...] [-d REV] [OPTION]...'),
     helpcategory=command.CATEGORY_CHANGE_MANAGEMENT,
 )
 def rebase(ui, repo, **opts):
     """move changeset (and descendants) to a different branch
 
     Rebase uses repeated merging to graft changesets from one part of
     history (the source) onto another (the destination). This can be
     useful for linearizing *local* changes relative to a master
     development tree.
 
     Published commits cannot be rebased (see :hg:`help phases`).
     To copy commits, see :hg:`help graft`.
 
     If you don't specify a destination changeset (``-d/--dest``), rebase
     will use the same logic as :hg:`merge` to pick a destination.  if
     the current branch contains exactly one other head, the other head
     is merged with by default.  Otherwise, an explicit revision with
     which to merge with must be provided.  (destination changeset is not
     modified by rebasing, but new changesets are added as its
     descendants.)
 
     Here are the ways to select changesets:
 
       1. Explicitly select them using ``--rev``.
 
       2. Use ``--source`` to select a root changeset and include all of its
          descendants.
 
       3. Use ``--base`` to select a changeset; rebase will find ancestors
          and their descendants which are not also ancestors of the destination.
 
       4. If you do not specify any of ``--rev``, ``--source``, or ``--base``,
          rebase will use ``--base .`` as above.
 
     If ``--source`` or ``--rev`` is used, special names ``SRC`` and ``ALLSRC``
     can be used in ``--dest``. Destination would be calculated per source
     revision with ``SRC`` substituted by that single source revision and
     ``ALLSRC`` substituted by all source revisions.
 
     Rebase will destroy original changesets unless you use ``--keep``.
     It will also move your bookmarks (even if you do).
 
     Some changesets may be dropped if they do not contribute changes
     (e.g. merges from the destination branch).
 
     Unlike ``merge``, rebase will do nothing if you are at the branch tip of
     a named branch with two heads. You will need to explicitly specify source
     and/or destination.
 
     If you need to use a tool to automate merge/conflict decisions, you
     can specify one with ``--tool``, see :hg:`help merge-tools`.
     As a caveat: the tool will not be used to mediate when a file was
     deleted, there is no hook presently available for this.
 
     If a rebase is interrupted to manually resolve a conflict, it can be
     continued with --continue/-c, aborted with --abort/-a, or stopped with
     --stop.
 
     .. container:: verbose
 
       Examples:
 
       - move "local changes" (current commit back to branching point)
         to the current branch tip after a pull::
 
           hg rebase
 
       - move a single changeset to the stable branch::
 
           hg rebase -r 5f493448 -d stable
 
       - splice a commit and all its descendants onto another part of history::
 
           hg rebase --source c0c3 --dest 4cf9
 
       - rebase everything on a branch marked by a bookmark onto the
         default branch::
 
           hg rebase --base myfeature --dest default
 
       - collapse a sequence of changes into a single commit::
 
           hg rebase --collapse -r 1520:1525 -d .
 
       - move a named branch while preserving its name::
 
           hg rebase -r "branch(featureX)" -d 1.3 --keepbranches
 
       - stabilize orphaned changesets so history looks linear::
 
           hg rebase -r 'orphan()-obsolete()'\
  -d 'first(max((successors(max(roots(ALLSRC) & ::SRC)^)-obsolete())::) +\
  max(::((roots(ALLSRC) & ::SRC)^)-obsolete()))'
 
     Configuration Options:
 
     You can make rebase require a destination if you set the following config
     option::
 
       [commands]
       rebase.requiredest = True
 
     By default, rebase will close the transaction after each commit. For
     performance purposes, you can configure rebase to use a single transaction
     across the entire rebase. WARNING: This setting introduces a significant
     risk of losing the work you've done in a rebase if the rebase aborts
     unexpectedly::
 
       [rebase]
       singletransaction = True
 
     By default, rebase writes to the working copy, but you can configure it to
     run in-memory for better performance. When the rebase is not moving the
     parent(s) of the working copy (AKA the "currently checked out changesets"),
     this may also allow it to run even if the working copy is dirty::
 
       [rebase]
       experimental.inmemory = True
 
     Return Values:
 
     Returns 0 on success, 1 if nothing to rebase or there are
     unresolved conflicts.
 
     """
     inmemory = ui.configbool(b'rebase', b'experimental.inmemory')
     action = cmdutil.check_at_most_one_arg(opts, 'abort', 'stop', 'continue')
     if action:
         cmdutil.check_incompatible_arguments(
             opts, action, ['confirm', 'dry_run']
         )
         cmdutil.check_incompatible_arguments(
             opts, action, ['rev', 'source', 'base', 'dest']
         )
     cmdutil.check_at_most_one_arg(opts, 'confirm', 'dry_run')
     cmdutil.check_at_most_one_arg(opts, 'rev', 'source', 'base')
 
     if action or repo.currenttransaction() is not None:
         # in-memory rebase is not compatible with resuming rebases.
         # (Or if it is run within a transaction, since the restart logic can
         # fail the entire transaction.)
         inmemory = False
 
     if opts.get('auto_orphans'):
         disallowed_opts = set(opts) - {'auto_orphans'}
         cmdutil.check_incompatible_arguments(
             opts, 'auto_orphans', disallowed_opts
         )
 
         userrevs = list(repo.revs(opts.get('auto_orphans')))
         opts['rev'] = [revsetlang.formatspec(b'%ld and orphan()', userrevs)]
         opts['dest'] = b'_destautoorphanrebase(SRC)'
 
     if opts.get('dry_run') or opts.get('confirm'):
         return _dryrunrebase(ui, repo, action, opts)
     elif action == 'stop':
         rbsrt = rebaseruntime(repo, ui)
         with repo.wlock(), repo.lock():
             rbsrt.restorestatus()
             if rbsrt.collapsef:
                 raise error.StateError(_(b"cannot stop in --collapse session"))
             allowunstable = obsolete.isenabled(repo, obsolete.allowunstableopt)
             if not (rbsrt.keepf or allowunstable):
                 raise error.StateError(
                     _(
                         b"cannot remove original changesets with"
                         b" unrebased descendants"
                     ),
                     hint=_(
                         b'either enable obsmarkers to allow unstable '
                         b'revisions or use --keep to keep original '
                         b'changesets'
                     ),
                 )
             # update to the current working revision
             # to clear interrupted merge
             mergemod.clean_update(repo[rbsrt.originalwd])
             rbsrt._finishrebase()
             return 0
     elif inmemory:
         try:
             # in-memory merge doesn't support conflicts, so if we hit any, abort
             # and re-run as an on-disk merge.
             overrides = {(b'rebase', b'singletransaction'): True}
             with ui.configoverride(overrides, b'rebase'):
                 return _dorebase(ui, repo, action, opts, inmemory=inmemory)
         except error.InMemoryMergeConflictsError:
             if ui.configbool(b'devel', b'rebase.force-in-memory-merge'):
                 raise
             ui.warn(
                 _(
                     b'hit merge conflicts; re-running rebase without in-memory'
                     b' merge\n'
                 )
             )
             clearstatus(repo)
             clearcollapsemsg(repo)
             return _dorebase(ui, repo, action, opts, inmemory=False)
     else:
         return _dorebase(ui, repo, action, opts)
 
 
 def _dryrunrebase(ui, repo, action, opts):
     rbsrt = rebaseruntime(repo, ui, inmemory=True, dryrun=True, opts=opts)
     confirm = opts.get('confirm')
     if confirm:
         ui.status(_(b'starting in-memory rebase\n'))
     else:
         ui.status(
             _(b'starting dry-run rebase; repository will not be changed\n')
         )
     with repo.wlock(), repo.lock():
         needsabort = True
         try:
             overrides = {(b'rebase', b'singletransaction'): True}
             with ui.configoverride(overrides, b'rebase'):
                 res = _origrebase(
                     ui,
                     repo,
                     action,
                     opts,
                     rbsrt,
                 )
                 if res == _nothingtorebase():
                     needsabort = False
                     return res
         except error.ConflictResolutionRequired:
             ui.status(_(b'hit a merge conflict\n'))
             return 1
         except error.Abort:
             needsabort = False
             raise
         else:
             if confirm:
                 ui.status(_(b'rebase completed successfully\n'))
                 if not ui.promptchoice(_(b'apply changes (yn)?$$ &Yes $$ &No')):
                     # finish unfinished rebase
                     rbsrt._finishrebase()
                 else:
                     rbsrt._prepareabortorcontinue(
                         isabort=True,
                         backup=False,
                         suppwarns=True,
                         confirm=confirm,
                     )
                 needsabort = False
             else:
                 ui.status(
                     _(
                         b'dry-run rebase completed successfully; run without'
                         b' -n/--dry-run to perform this rebase\n'
                     )
                 )
             return 0
         finally:
             if needsabort:
                 # no need to store backup in case of dryrun
                 rbsrt._prepareabortorcontinue(
                     isabort=True,
                     backup=False,
                     suppwarns=True,
                     dryrun=opts.get('dry_run'),
                 )
 
 
 def _dorebase(ui, repo, action, opts, inmemory=False):
     rbsrt = rebaseruntime(repo, ui, inmemory, opts=opts)
     return _origrebase(ui, repo, action, opts, rbsrt)
 
 
 def _origrebase(ui, repo, action, opts, rbsrt):
     assert action != 'stop'
     with repo.wlock(), repo.lock():
         if opts.get('interactive'):
             try:
                 if extensions.find(b'histedit'):
                     enablehistedit = b''
             except KeyError:
                 enablehistedit = b" --config extensions.histedit="
             help = b"hg%s help -e histedit" % enablehistedit
             msg = (
                 _(
                     b"interactive history editing is supported by the "
                     b"'histedit' extension (see \"%s\")"
                 )
                 % help
             )
             raise error.InputError(msg)
 
         if rbsrt.collapsemsg and not rbsrt.collapsef:
             raise error.InputError(
                 _(b'message can only be specified with collapse')
             )
 
         if action:
             if rbsrt.collapsef:
                 raise error.InputError(
                     _(b'cannot use collapse with continue or abort')
                 )
             if action == 'abort' and opts.get('tool', False):
                 ui.warn(_(b'tool option will be ignored\n'))
             if action == 'continue':
                 ms = mergestatemod.mergestate.read(repo)
                 mergeutil.checkunresolved(ms)
 
             retcode = rbsrt._prepareabortorcontinue(isabort=(action == 'abort'))
             if retcode is not None:
                 return retcode
         else:
             # search default destination in this space
             # used in the 'hg pull --rebase' case, see issue 5214.
             destspace = opts.get('_destspace')
             destmap = _definedestmap(
                 ui,
                 repo,
                 rbsrt.inmemory,
                 opts.get('dest', None),
                 opts.get('source', []),
                 opts.get('base', []),
                 opts.get('rev', []),
                 destspace=destspace,
             )
             retcode = rbsrt._preparenewrebase(destmap)
             if retcode is not None:
                 return retcode
             storecollapsemsg(repo, rbsrt.collapsemsg)
 
         tr = None
 
         singletr = ui.configbool(b'rebase', b'singletransaction')
         if singletr:
             tr = repo.transaction(b'rebase')
 
         # If `rebase.singletransaction` is enabled, wrap the entire operation in
         # one transaction here. Otherwise, transactions are obtained when
         # committing each node, which is slower but allows partial success.
         with util.acceptintervention(tr):
             rbsrt._performrebase(tr)
             if not rbsrt.dryrun:
                 rbsrt._finishrebase()
 
 
 def _definedestmap(ui, repo, inmemory, destf, srcf, basef, revf, destspace):
     """use revisions argument to define destmap {srcrev: destrev}"""
     if revf is None:
         revf = []
 
     # destspace is here to work around issues with `hg pull --rebase` see
     # issue5214 for details
 
     cmdutil.checkunfinished(repo)
     if not inmemory:
         cmdutil.bailifchanged(repo)
 
     if ui.configbool(b'commands', b'rebase.requiredest') and not destf:
         raise error.InputError(
             _(b'you must specify a destination'),
             hint=_(b'use: hg rebase -d REV'),
         )
 
     dest = None
 
     if revf:
         rebaseset = logcmdutil.revrange(repo, revf)
         if not rebaseset:
             ui.status(_(b'empty "rev" revision set - nothing to rebase\n'))
             return None
     elif srcf:
         src = logcmdutil.revrange(repo, srcf)
         if not src:
             ui.status(_(b'empty "source" revision set - nothing to rebase\n'))
             return None
         # `+  (%ld)` to work around `wdir()::` being empty
         rebaseset = repo.revs(b'(%ld):: + (%ld)', src, src)
     else:
         base = logcmdutil.revrange(repo, basef or [b'.'])
         if not base:
             ui.status(
                 _(b'empty "base" revision set - ' b"can't compute rebase set\n")
             )
             return None
         if destf:
             # --base does not support multiple destinations
             dest = logcmdutil.revsingle(repo, destf)
         else:
             dest = repo[_destrebase(repo, base, destspace=destspace)]
             destf = bytes(dest)
 
         roots = []  # selected children of branching points
         bpbase = {}  # {branchingpoint: [origbase]}
         for b in base:  # group bases by branching points
             bp = repo.revs(b'ancestor(%d, %d)', b, dest.rev()).first()
             bpbase[bp] = bpbase.get(bp, []) + [b]
         if None in bpbase:
             # emulate the old behavior, showing "nothing to rebase" (a better
             # behavior may be abort with "cannot find branching point" error)
             bpbase.clear()
         for bp, bs in bpbase.items():  # calculate roots
             roots += list(repo.revs(b'children(%d) & ancestors(%ld)', bp, bs))
 
         rebaseset = repo.revs(b'%ld::', roots)
 
         if not rebaseset:
             # transform to list because smartsets are not comparable to
             # lists. This should be improved to honor laziness of
             # smartset.
             if list(base) == [dest.rev()]:
                 if basef:
                     ui.status(
                         _(
                             b'nothing to rebase - %s is both "base"'
                             b' and destination\n'
                         )
                         % dest
                     )
                 else:
                     ui.status(
                         _(
                             b'nothing to rebase - working directory '
                             b'parent is also destination\n'
                         )
                     )
             elif not repo.revs(b'%ld - ::%d', base, dest.rev()):
                 if basef:
                     ui.status(
                         _(
                             b'nothing to rebase - "base" %s is '
                             b'already an ancestor of destination '
                             b'%s\n'
                         )
                         % (b'+'.join(bytes(repo[r]) for r in base), dest)
                     )
                 else:
                     ui.status(
                         _(
                             b'nothing to rebase - working '
                             b'directory parent is already an '
                             b'ancestor of destination %s\n'
                         )
                         % dest
                     )
             else:  # can it happen?
                 ui.status(
                     _(b'nothing to rebase from %s to %s\n')
                     % (b'+'.join(bytes(repo[r]) for r in base), dest)
                 )
             return None
 
     if wdirrev in rebaseset:
         raise error.InputError(_(b'cannot rebase the working copy'))
     rebasingwcp = repo[b'.'].rev() in rebaseset
     ui.log(
         b"rebase",
         b"rebasing working copy parent: %r\n",
         rebasingwcp,
         rebase_rebasing_wcp=rebasingwcp,
     )
     if inmemory and rebasingwcp:
         # Check these since we did not before.
         cmdutil.checkunfinished(repo)
         cmdutil.bailifchanged(repo)
 
     if not destf:
         dest = repo[_destrebase(repo, rebaseset, destspace=destspace)]
         destf = bytes(dest)
 
     allsrc = revsetlang.formatspec(b'%ld', rebaseset)
     alias = {b'ALLSRC': allsrc}
 
     if dest is None:
         try:
             # fast path: try to resolve dest without SRC alias
             dest = scmutil.revsingle(repo, destf, localalias=alias)
         except error.RepoLookupError:
             # multi-dest path: resolve dest for each SRC separately
             destmap = {}
             for r in rebaseset:
                 alias[b'SRC'] = revsetlang.formatspec(b'%d', r)
                 # use repo.anyrevs instead of scmutil.revsingle because we
                 # don't want to abort if destset is empty.
                 destset = repo.anyrevs([destf], user=True, localalias=alias)
                 size = len(destset)
                 if size == 1:
                     destmap[r] = destset.first()
                 elif size == 0:
                     ui.note(_(b'skipping %s - empty destination\n') % repo[r])
                 else:
                     raise error.InputError(
                         _(b'rebase destination for %s is not unique') % repo[r]
                     )
 
     if dest is not None:
         # single-dest case: assign dest to each rev in rebaseset
         destrev = dest.rev()
         destmap = {r: destrev for r in rebaseset}  # {srcrev: destrev}
 
     if not destmap:
         ui.status(_(b'nothing to rebase - empty destination\n'))
         return None
 
     return destmap
 
 
 def externalparent(repo, state, destancestors):
     """Return the revision that should be used as the second parent
     when the revisions in state is collapsed on top of destancestors.
     Abort if there is more than one parent.
     """
     parents = set()
     source = min(state)
     for rev in state:
         if rev == source:
             continue
         for p in repo[rev].parents():
             if p.rev() not in state and p.rev() not in destancestors:
                 parents.add(p.rev())
     if not parents:
         return nullrev
     if len(parents) == 1:
         return parents.pop()
     raise error.StateError(
         _(
             b'unable to collapse on top of %d, there is more '
             b'than one external parent: %s'
         )
         % (max(destancestors), b', '.join(b"%d" % p for p in sorted(parents)))
     )
 
 
 def commitmemorynode(repo, wctx, editor, extra, user, date, commitmsg):
     """Commit the memory changes with parents p1 and p2.
     Return node of committed revision."""
     # By convention, ``extra['branch']`` (set by extrafn) clobbers
     # ``branch`` (used when passing ``--keepbranches``).
     branch = None
     if b'branch' in extra:
         branch = extra[b'branch']
 
     # FIXME: We call _compact() because it's required to correctly detect
     # changed files. This was added to fix a regression shortly before the 5.5
     # release. A proper fix will be done in the default branch.
     wctx._compact()
     memctx = wctx.tomemctx(
         commitmsg,
         date=date,
         extra=extra,
         user=user,
         branch=branch,
         editor=editor,
     )
     if memctx.isempty() and not repo.ui.configbool(b'ui', b'allowemptycommit'):
         return None
     commitres = repo.commitctx(memctx)
     wctx.clean()  # Might be reused
     return commitres
 
 
 def commitnode(repo, editor, extra, user, date, commitmsg):
     """Commit the wd changes with parents p1 and p2.
     Return node of committed revision."""
     tr = util.nullcontextmanager
     if not repo.ui.configbool(b'rebase', b'singletransaction'):
         tr = lambda: repo.transaction(b'rebase')
     with tr():
         # Commit might fail if unresolved files exist
         newnode = repo.commit(
             text=commitmsg, user=user, date=date, extra=extra, editor=editor
         )
 
         repo.dirstate.setbranch(
             repo[newnode].branch(), repo.currenttransaction()
         )
         return newnode
 
 
 def rebasenode(repo, rev, p1, p2, base, collapse, wctx):
     """Rebase a single revision rev on top of p1 using base as merge ancestor"""
     # Merge phase
     # Update to destination and merge it with local
     p1ctx = repo[p1]
     if wctx.isinmemory():
         wctx.setbase(p1ctx)
         scope = util.nullcontextmanager
     else:
         if repo[b'.'].rev() != p1:
             repo.ui.debug(b" update to %d:%s\n" % (p1, p1ctx))
             mergemod.clean_update(p1ctx)
         else:
             repo.ui.debug(b" already in destination\n")
         scope = lambda: repo.dirstate.changing_parents(repo)
         # This is, alas, necessary to invalidate workingctx's manifest cache,
         # as well as other data we litter on it in other places.
         wctx = repo[None]
         repo.dirstate.write(repo.currenttransaction())
     ctx = repo[rev]
     repo.ui.debug(b" merge against %d:%s\n" % (rev, ctx))
     if base is not None:
         repo.ui.debug(b"   detach base %d:%s\n" % (base, repo[base]))
 
     with scope():
         # See explanation in merge.graft()
         mergeancestor = repo.changelog.isancestor(p1ctx.node(), ctx.node())
         stats = mergemod._update(
             repo,
             rev,
             branchmerge=True,
             force=True,
             ancestor=base,
             mergeancestor=mergeancestor,
             labels=[b'dest', b'source', b'parent of source'],
             wc=wctx,
         )
         wctx.setparents(p1ctx.node(), repo[p2].node())
         if collapse:
             copies.graftcopies(wctx, ctx, p1ctx)
         else:
             # If we're not using --collapse, we need to
             # duplicate copies between the revision we're
             # rebasing and its first parent.
             copies.graftcopies(wctx, ctx, ctx.p1())
 
     if stats.unresolvedcount > 0:
         if wctx.isinmemory():
             raise error.InMemoryMergeConflictsError()
         else:
             raise error.ConflictResolutionRequired(b'rebase')
 
 
 def adjustdest(repo, rev, destmap, state, skipped):
     r"""adjust rebase destination given the current rebase state
 
     rev is what is being rebased. Return a list of two revs, which are the
     adjusted destinations for rev's p1 and p2, respectively. If a parent is
     nullrev, return dest without adjustment for it.
 
     For example, when doing rebasing B+E to F, C to G, rebase will first move B
     to B1, and E's destination will be adjusted from F to B1.
 
         B1 <- written during rebasing B
         |
         F <- original destination of B, E
         |
         | E <- rev, which is being rebased
         | |
         | D <- prev, one parent of rev being checked
         | |
         | x <- skipped, ex. no successor or successor in (::dest)
         | |
         | C <- rebased as C', different destination
         | |
         | B <- rebased as B1     C'
         |/                       |
         A                        G <- destination of C, different
 
     Another example about merge changeset, rebase -r C+G+H -d K, rebase will
     first move C to C1, G to G1, and when it's checking H, the adjusted
     destinations will be [C1, G1].
 
             H       C1 G1
            /|       | /
           F G       |/
         K | |  ->   K
         | C D       |
         | |/        |
         | B         | ...
         |/          |/
         A           A
 
     Besides, adjust dest according to existing rebase information. For example,
 
       B C D    B needs to be rebased on top of C, C needs to be rebased on top
        \|/     of D. We will rebase C first.
         A
 
           C'   After rebasing C, when considering B's destination, use C'
           |    instead of the original C.
       B   D
        \ /
         A
     """
     # pick already rebased revs with same dest from state as interesting source
     dest = destmap[rev]
     source = [
         s
         for s, d in state.items()
         if d > 0 and destmap[s] == dest and s not in skipped
     ]
 
     result = []
     for prev in repo.changelog.parentrevs(rev):
         adjusted = dest
         if prev != nullrev:
             candidate = repo.revs(b'max(%ld and (::%d))', source, prev).first()
             if candidate is not None:
                 adjusted = state[candidate]
         if adjusted == dest and dest in state:
             adjusted = state[dest]
             if adjusted == revtodo:
                 # sortsource should produce an order that makes this impossible
                 raise error.ProgrammingError(
                     b'rev %d should be rebased already at this time' % dest
                 )
         result.append(adjusted)
     return result
 
 
 def _checkobsrebase(repo, ui, rebaseobsrevs, rebaseobsskipped):
     """
     Abort if rebase will create divergence or rebase is noop because of markers
 
     `rebaseobsrevs`: set of obsolete revision in source
     `rebaseobsskipped`: set of revisions from source skipped because they have
     successors in destination or no non-obsolete successor.
     """
     # Obsolete node with successors not in dest leads to divergence
     divergenceok = obsolete.isenabled(repo, obsolete.allowdivergenceopt)
     divergencebasecandidates = rebaseobsrevs - rebaseobsskipped
 
     if divergencebasecandidates and not divergenceok:
         divhashes = (bytes(repo[r]) for r in divergencebasecandidates)
         msg = _(b"this rebase will cause divergences from: %s")
         h = _(
             b"to force the rebase please set "
             b"experimental.evolution.allowdivergence=True"
         )
         raise error.StateError(msg % (b",".join(divhashes),), hint=h)
 
 
 def successorrevs(unfi, rev):
     """yield revision numbers for successors of rev"""
     assert unfi.filtername is None
     get_rev = unfi.changelog.index.get_rev
     for s in obsutil.allsuccessors(unfi.obsstore, [unfi[rev].node()]):
         r = get_rev(s)
         if r is not None:
             yield r
 
 
 def defineparents(repo, rev, destmap, state, skipped, obsskipped):
     """Return new parents and optionally a merge base for rev being rebased
 
     The destination specified by "dest" cannot always be used directly because
     previously rebase result could affect destination. For example,
 
           D E    rebase -r C+D+E -d B
           |/     C will be rebased to C'
         B C      D's new destination will be C' instead of B
         |/       E's new destination will be C' instead of B
         A
 
     The new parents of a merge is slightly more complicated. See the comment
     block below.
     """
     # use unfiltered changelog since successorrevs may return filtered nodes
     assert repo.filtername is None
     cl = repo.changelog
     isancestor = cl.isancestorrev
 
     dest = destmap[rev]
     oldps = repo.changelog.parentrevs(rev)  # old parents
     newps = [nullrev, nullrev]  # new parents
     dests = adjustdest(repo, rev, destmap, state, skipped)
     bases = list(oldps)  # merge base candidates, initially just old parents
 
     if all(r == nullrev for r in oldps[1:]):
         # For non-merge changeset, just move p to adjusted dest as requested.
         newps[0] = dests[0]
     else:
         # For merge changeset, if we move p to dests[i] unconditionally, both
         # parents may change and the end result looks like "the merge loses a
         # parent", which is a surprise. This is a limit because "--dest" only
         # accepts one dest per src.
         #
         # Therefore, only move p with reasonable conditions (in this order):
         #   1. use dest, if dest is a descendent of (p or one of p's successors)
         #   2. use p's rebased result, if p is rebased (state[p] > 0)
         #
         # Comparing with adjustdest, the logic here does some additional work:
         #   1. decide which parents will not be moved towards dest
         #   2. if the above decision is "no", should a parent still be moved
         #      because it was rebased?
         #
         # For example:
         #
         #     C    # "rebase -r C -d D" is an error since none of the parents
         #    /|    # can be moved. "rebase -r B+C -d D" will move C's parent
         #   A B D  # B (using rule "2."), since B will be rebased.
         #
         # The loop tries to be not rely on the fact that a Mercurial node has
         # at most 2 parents.
         for i, p in enumerate(oldps):
             np = p  # new parent
             if any(isancestor(x, dests[i]) for x in successorrevs(repo, p)):
                 np = dests[i]
             elif p in state and state[p] > 0:
                 np = state[p]
 
             # If one parent becomes an ancestor of the other, drop the ancestor
             for j, x in enumerate(newps[:i]):
                 if x == nullrev:
                     continue
                 if isancestor(np, x):  # CASE-1
                     np = nullrev
                 elif isancestor(x, np):  # CASE-2
                     newps[j] = np
                     np = nullrev
                     # New parents forming an ancestor relationship does not
                     # mean the old parents have a similar relationship. Do not
                     # set bases[x] to nullrev.
                     bases[j], bases[i] = bases[i], bases[j]
 
             newps[i] = np
 
         # "rebasenode" updates to new p1, and the old p1 will be used as merge
         # base. If only p2 changes, merging using unchanged p1 as merge base is
         # suboptimal. Therefore swap parents to make the merge sane.
         if newps[1] != nullrev and oldps[0] == newps[0]:
             assert len(newps) == 2 and len(oldps) == 2
             newps.reverse()
             bases.reverse()
 
         # No parent change might be an error because we fail to make rev a
         # descendent of requested dest. This can happen, for example:
         #
         #     C    # rebase -r C -d D
         #    /|    # None of A and B will be changed to D and rebase fails.
         #   A B D
         if set(newps) == set(oldps) and dest not in newps:
             raise error.InputError(
                 _(
                     b'cannot rebase %d:%s without '
                     b'moving at least one of its parents'
                 )
                 % (rev, repo[rev])
             )
 
     # Source should not be ancestor of dest. The check here guarantees it's
     # impossible. With multi-dest, the initial check does not cover complex
     # cases since we don't have abstractions to dry-run rebase cheaply.
     if any(p != nullrev and isancestor(rev, p) for p in newps):
         raise error.InputError(_(b'source is ancestor of destination'))
 
     # Check if the merge will contain unwanted changes. That may happen if
     # there are multiple special (non-changelog ancestor) merge bases, which
     # cannot be handled well by the 3-way merge algorithm. For example:
     #
     #     F
     #    /|
     #   D E  # "rebase -r D+E+F -d Z", when rebasing F, if "D" was chosen
     #   | |  # as merge base, the difference between D and F will include
     #   B C  # C, so the rebased F will contain C surprisingly. If "E" was
     #   |/   #  chosen, the rebased F will contain B.
     #   A Z
     #
     # But our merge base candidates (D and E in above case) could still be
     # better than the default (ancestor(F, Z) == null). Therefore still
     # pick one (so choose p1 above).
     if sum(1 for b in set(bases) if b != nullrev and b not in newps) > 1:
         unwanted = [None, None]  # unwanted[i]: unwanted revs if choose bases[i]
         for i, base in enumerate(bases):
             if base == nullrev or base in newps:
                 continue
             # Revisions in the side (not chosen as merge base) branch that
             # might contain "surprising" contents
             other_bases = set(bases) - {base}
             siderevs = list(
                 repo.revs(b'(%ld %% (%d+%d))', other_bases, base, dest)
             )
 
             # If those revisions are covered by rebaseset, the result is good.
             # A merge in rebaseset would be considered to cover its ancestors.
             if siderevs:
                 rebaseset = [
                     r for r, d in state.items() if d > 0 and r not in obsskipped
                 ]
                 merges = [
                     r for r in rebaseset if cl.parentrevs(r)[1] != nullrev
                 ]
                 unwanted[i] = list(
                     repo.revs(
                         b'%ld - (::%ld) - %ld', siderevs, merges, rebaseset
                     )
                 )
 
         if any(revs is not None for revs in unwanted):
             # Choose a merge base that has a minimal number of unwanted revs.
             l, i = min(
                 (len(revs), i)
                 for i, revs in enumerate(unwanted)
                 if revs is not None
             )
 
             # The merge will include unwanted revisions. Abort now. Revisit this if
             # we have a more advanced merge algorithm that handles multiple bases.
             if l > 0:
                 unwanteddesc = _(b' or ').join(
                     (
                         b', '.join(b'%d:%s' % (r, repo[r]) for r in revs)
                         for revs in unwanted
                         if revs is not None
                     )
                 )
                 raise error.InputError(
                     _(b'rebasing %d:%s will include unwanted changes from %s')
                     % (rev, repo[rev], unwanteddesc)
                 )
 
             # newps[0] should match merge base if possible. Currently, if newps[i]
             # is nullrev, the only case is newps[i] and newps[j] (j < i), one is
             # the other's ancestor. In that case, it's fine to not swap newps here.
             # (see CASE-1 and CASE-2 above)
             if i != 0:
                 if newps[i] != nullrev:
                     newps[0], newps[i] = newps[i], newps[0]
                 bases[0], bases[i] = bases[i], bases[0]
 
     # "rebasenode" updates to new p1, use the corresponding merge base.
     base = bases[0]
 
     repo.ui.debug(b" future parents are %d and %d\n" % tuple(newps))
 
     return newps[0], newps[1], base
 
 
 def isagitpatch(repo, patchname):
     """Return true if the given patch is in git format"""
     mqpatch = os.path.join(repo.mq.path, patchname)
     for line in patch.linereader(open(mqpatch, b'rb')):
         if line.startswith(b'diff --git'):
             return True
     return False
 
 
 def updatemq(repo, state, skipped, **opts):
     """Update rebased mq patches - finalize and then import them"""
     mqrebase = {}
     mq = repo.mq
     original_series = mq.fullseries[:]
     skippedpatches = set()
 
     for p in mq.applied:
         rev = repo[p.node].rev()
         if rev in state:
             repo.ui.debug(
                 b'revision %d is an mq patch (%s), finalize it.\n'
                 % (rev, p.name)
             )
             mqrebase[rev] = (p.name, isagitpatch(repo, p.name))
         else:
             # Applied but not rebased, not sure this should happen
             skippedpatches.add(p.name)
 
     if mqrebase:
         mq.finish(repo, mqrebase.keys())
 
         # We must start import from the newest revision
         for rev in sorted(mqrebase, reverse=True):
             if rev not in skipped:
                 name, isgit = mqrebase[rev]
                 repo.ui.note(
                     _(b'updating mq patch %s to %d:%s\n')
                     % (name, state[rev], repo[state[rev]])
                 )
                 mq.qimport(
                     repo,
                     (),
                     patchname=name,
                     git=isgit,
                     rev=[b"%d" % state[rev]],
                 )
             else:
                 # Rebased and skipped
                 skippedpatches.add(mqrebase[rev][0])
 
         # Patches were either applied and rebased and imported in
         # order, applied and removed or unapplied. Discard the removed
         # ones while preserving the original series order and guards.
         newseries = [
             s
             for s in original_series
             if mq.guard_re.split(s, 1)[0] not in skippedpatches
         ]
         mq.fullseries[:] = newseries
         mq.seriesdirty = True
         mq.savedirty()
 
 
 def storecollapsemsg(repo, collapsemsg):
     """Store the collapse message to allow recovery"""
     collapsemsg = collapsemsg or b''
     f = repo.vfs(b"last-message.txt", b"w")
     f.write(b"%s\n" % collapsemsg)
     f.close()
 
 
 def clearcollapsemsg(repo):
     """Remove collapse message file"""
     repo.vfs.unlinkpath(b"last-message.txt", ignoremissing=True)
 
 
 def restorecollapsemsg(repo, isabort):
     """Restore previously stored collapse message"""
     try:
         f = repo.vfs(b"last-message.txt")
         collapsemsg = f.readline().strip()
         f.close()
     except FileNotFoundError:
         if isabort:
             # Oh well, just abort like normal
             collapsemsg = b''
         else:
             raise error.Abort(_(b'missing .hg/last-message.txt for rebase'))
     return collapsemsg
 
 
 def clearstatus(repo):
     """Remove the status files"""
     # Make sure the active transaction won't write the state file
     tr = repo.currenttransaction()
     if tr:
         tr.removefilegenerator(b'rebasestate')
     repo.vfs.unlinkpath(b"rebasestate", ignoremissing=True)
 
 
 def sortsource(destmap):
     """yield source revisions in an order that we only rebase things once
 
     If source and destination overlaps, we should filter out revisions
     depending on other revisions which hasn't been rebased yet.
 
     Yield a sorted list of revisions each time.
 
     For example, when rebasing A to B, B to C. This function yields [B], then
     [A], indicating B needs to be rebased first.
 
     Raise if there is a cycle so the rebase is impossible.
     """
     srcset = set(destmap)
     while srcset:
         srclist = sorted(srcset)
         result = []
         for r in srclist:
             if destmap[r] not in srcset:
                 result.append(r)
         if not result:
             raise error.InputError(_(b'source and destination form a cycle'))
         srcset -= set(result)
         yield result
 
 
 def buildstate(repo, destmap, collapse):
     """Define which revisions are going to be rebased and where
 
     repo: repo
     destmap: {srcrev: destrev}
     """
     rebaseset = destmap.keys()
     originalwd = repo[b'.'].rev()
 
     # This check isn't strictly necessary, since mq detects commits over an
     # applied patch. But it prevents messing up the working directory when
     # a partially completed rebase is blocked by mq.
     if b'qtip' in repo.tags():
         mqapplied = {repo[s.node].rev() for s in repo.mq.applied}
         if set(destmap.values()) & mqapplied:
             raise error.StateError(_(b'cannot rebase onto an applied mq patch'))
 
     # Get "cycle" error early by exhausting the generator.
     sortedsrc = list(sortsource(destmap))  # a list of sorted revs
     if not sortedsrc:
         raise error.InputError(_(b'no matching revisions'))
 
     # Only check the first batch of revisions to rebase not depending on other
     # rebaseset. This means "source is ancestor of destination" for the second
     # (and following) batches of revisions are not checked here. We rely on
     # "defineparents" to do that check.
     roots = list(repo.set(b'roots(%ld)', sortedsrc[0]))
     if not roots:
         raise error.InputError(_(b'no matching revisions'))
 
     def revof(r):
         return r.rev()
 
     roots = sorted(roots, key=revof)
     state = dict.fromkeys(rebaseset, revtodo)
     emptyrebase = len(sortedsrc) == 1
     for root in roots:
         dest = repo[destmap[root.rev()]]
         commonbase = root.ancestor(dest)
         if commonbase == root:
             raise error.InputError(_(b'source is ancestor of destination'))
         if commonbase == dest:
             wctx = repo[None]
             if dest == wctx.p1():
                 # when rebasing to '.', it will use the current wd branch name
                 samebranch = root.branch() == wctx.branch()
             else:
                 samebranch = root.branch() == dest.branch()
             if not collapse and samebranch and dest in root.parents():
                 # mark the revision as done by setting its new revision
                 # equal to its old (current) revisions
                 state[root.rev()] = root.rev()
                 repo.ui.debug(b'source is a child of destination\n')
                 continue
 
         emptyrebase = False
         repo.ui.debug(b'rebase onto %s starting from %s\n' % (dest, root))
     if emptyrebase:
         return None
     for rev in sorted(state):
         parents = [p for p in repo.changelog.parentrevs(rev) if p != nullrev]
         # if all parents of this revision are done, then so is this revision
         if parents and all((state.get(p) == p for p in parents)):
             state[rev] = rev
     return originalwd, destmap, state
 
 
 def clearrebased(
     ui,
     repo,
     destmap,
     state,
     skipped,
     collapsedas=None,
     keepf=False,
     fm=None,
     backup=True,
 ):
     """dispose of rebased revision at the end of the rebase
 
     If `collapsedas` is not None, the rebase was a collapse whose result if the
     `collapsedas` node.
 
     If `keepf` is not True, the rebase has --keep set and no nodes should be
     removed (but bookmarks still need to be moved).
 
     If `backup` is False, no backup will be stored when stripping rebased
     revisions.
     """
     tonode = repo.changelog.node
     replacements = {}
     moves = {}
     stripcleanup = not obsolete.isenabled(repo, obsolete.createmarkersopt)
 
     collapsednodes = []
     for rev, newrev in sorted(state.items()):
         if newrev >= 0 and newrev != rev:
             oldnode = tonode(rev)
             newnode = collapsedas or tonode(newrev)
             moves[oldnode] = newnode
             succs = None
             if rev in skipped:
                 if stripcleanup or not repo[rev].obsolete():
                     succs = ()
             elif collapsedas:
                 collapsednodes.append(oldnode)
             else:
                 succs = (newnode,)
             if succs is not None:
                 replacements[(oldnode,)] = succs
     if collapsednodes:
         replacements[tuple(collapsednodes)] = (collapsedas,)
     if fm:
         hf = fm.hexfunc
         fl = fm.formatlist
         fd = fm.formatdict
         changes = {}
         for oldns, newn in replacements.items():
             for oldn in oldns:
                 changes[hf(oldn)] = fl([hf(n) for n in newn], name=b'node')
         nodechanges = fd(changes, key=b"oldnode", value=b"newnodes")
         fm.data(nodechanges=nodechanges)
     if keepf:
         replacements = {}
     scmutil.cleanupnodes(repo, replacements, b'rebase', moves, backup=backup)
 
 
 def pullrebase(orig, ui, repo, *args, **opts):
     """Call rebase after pull if the latter has been invoked with --rebase"""
     if opts.get('rebase'):
         if ui.configbool(b'commands', b'rebase.requiredest'):
             msg = _(b'rebase destination required by configuration')
             hint = _(b'use hg pull followed by hg rebase -d DEST')
             raise error.InputError(msg, hint=hint)
 
         with repo.wlock(), repo.lock():
             if opts.get('update'):
                 del opts['update']
                 ui.debug(
                     b'--update and --rebase are not compatible, ignoring '
                     b'the update flag\n'
                 )
 
             cmdutil.checkunfinished(repo, skipmerge=True)
             cmdutil.bailifchanged(
                 repo,
                 hint=_(
                     b'cannot pull with rebase: '
                     b'please commit or shelve your changes first'
                 ),
             )
 
             revsprepull = len(repo)
-            origpostincoming = commands.postincoming
+            origpostincoming = cmdutil.postincoming
 
             def _dummy(*args, **kwargs):
                 pass
 
-            commands.postincoming = _dummy
+            cmdutil.postincoming = _dummy
             try:
                 ret = orig(ui, repo, *args, **opts)
             finally:
-                commands.postincoming = origpostincoming
+                cmdutil.postincoming = origpostincoming
             revspostpull = len(repo)
             if revspostpull > revsprepull:
                 # --rev option from pull conflict with rebase own --rev
                 # dropping it
                 if 'rev' in opts:
                     del opts['rev']
                 # positional argument from pull conflicts with rebase's own
                 # --source.
                 if 'source' in opts:
                     del opts['source']
                 # revsprepull is the len of the repo, not revnum of tip.
                 destspace = list(repo.changelog.revs(start=revsprepull))
                 opts['_destspace'] = destspace
                 try:
                     rebase(ui, repo, **opts)
                 except error.NoMergeDestAbort:
                     # we can maybe update instead
                     rev, _a, _b = destutil.destupdate(repo)
                     if rev == repo[b'.'].rev():
                         ui.status(_(b'nothing to rebase\n'))
                     else:
                         ui.status(_(b'nothing to rebase - updating instead\n'))
                         # not passing argument to get the bare update behavior
                         # with warning and trumpets
                         commands.update(ui, repo)
     else:
         if opts.get('tool'):
             raise error.InputError(_(b'--tool can only be used with --rebase'))
         ret = orig(ui, repo, *args, **opts)
 
     return ret
 
 
 def _compute_obsolete_sets(repo, rebaseobsrevs, destmap):
     """Figure out what to do about about obsolete revisions
 
     `obsolete_with_successor_in_destination` is a mapping mapping obsolete => successor for all
     obsolete nodes to be rebased given in `rebaseobsrevs`.
 
     `obsolete_with_successor_in_rebase_set` is a set with obsolete revisions,
     without a successor in destination, that would cause divergence.
     """
     obsolete_with_successor_in_destination = {}
     obsolete_with_successor_in_rebase_set = set()
 
     cl = repo.changelog
     get_rev = cl.index.get_rev
     extinctrevs = set(repo.revs(b'extinct()'))
     for srcrev in rebaseobsrevs:
         srcnode = cl.node(srcrev)
         # XXX: more advanced APIs are required to handle split correctly
         successors = set(obsutil.allsuccessors(repo.obsstore, [srcnode]))
         # obsutil.allsuccessors includes node itself
         successors.remove(srcnode)
         succrevs = {get_rev(s) for s in successors}
         succrevs.discard(None)
         if not successors or succrevs.issubset(extinctrevs):
             # no successor, or all successors are extinct
             obsolete_with_successor_in_destination[srcrev] = None
         else:
             dstrev = destmap[srcrev]
             for succrev in succrevs:
                 if cl.isancestorrev(succrev, dstrev):
                     obsolete_with_successor_in_destination[srcrev] = succrev
                     break
             else:
                 # If 'srcrev' has a successor in rebase set but none in
                 # destination (which would be catched above), we shall skip it
                 # and its descendants to avoid divergence.
                 if srcrev in extinctrevs or any(s in destmap for s in succrevs):
                     obsolete_with_successor_in_rebase_set.add(srcrev)
 
     return (
         obsolete_with_successor_in_destination,
         obsolete_with_successor_in_rebase_set,
     )
 
 
 def abortrebase(ui, repo):
     with repo.wlock(), repo.lock():
         rbsrt = rebaseruntime(repo, ui)
         rbsrt._prepareabortorcontinue(isabort=True)
 
 
 def continuerebase(ui, repo):
     with repo.wlock(), repo.lock():
         rbsrt = rebaseruntime(repo, ui)
         ms = mergestatemod.mergestate.read(repo)
         mergeutil.checkunresolved(ms)
         retcode = rbsrt._prepareabortorcontinue(isabort=False)
         if retcode is not None:
             return retcode
         rbsrt._performrebase(None)
         rbsrt._finishrebase()
 
 
 def summaryhook(ui, repo):
     if not repo.vfs.exists(b'rebasestate'):
         return
     try:
         rbsrt = rebaseruntime(repo, ui, {})
         rbsrt.restorestatus()
         state = rbsrt.state
     except error.RepoLookupError:
         # i18n: column positioning for "hg summary"
         msg = _(b'rebase: (use "hg rebase --abort" to clear broken state)\n')
         ui.write(msg)
         return
     numrebased = len([i for i in state.values() if i >= 0])
     # i18n: column positioning for "hg summary"
     ui.write(
         _(b'rebase: %s, %s (rebase --continue)\n')
         % (
             ui.label(_(b'%d rebased'), b'rebase.rebased') % numrebased,
             ui.label(_(b'%d remaining'), b'rebase.remaining')
             % (len(state) - numrebased),
         )
     )
 
 
 def uisetup(ui):
     # Replace pull with a decorator to provide --rebase option
     entry = extensions.wrapcommand(commands.table, b'pull', pullrebase)
     entry[1].append(
         (b'', b'rebase', None, _(b"rebase working directory to branch head"))
     )
     entry[1].append((b't', b'tool', b'', _(b"specify merge tool for rebase")))
     cmdutil.summaryhooks.add(b'rebase', summaryhook)
     statemod.addunfinished(
         b'rebase',
         fname=b'rebasestate',
         stopflag=True,
         continueflag=True,
         abortfunc=abortrebase,
         continuefunc=continuerebase,
     )