url: move _wraphttpresponse() from httpeer

This is a generally useful function. Having it on the url module
will make it more accessible outside of the HTTP peers.

Differential Revision: https://phab.mercurial-scm.org/D4770

mercurial/httppeer.py

@@ -1,1006 +1,970 b''
 # httppeer.py - HTTP repository proxy classes for mercurial
 #
 # Copyright 2005, 2006 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.
 
 from __future__ import absolute_import
 
 import errno
 import io
 import os
 import socket
 import struct
 import weakref
 
 from .i18n import _
 from . import (
     bundle2,
     error,
     httpconnection,
     pycompat,
     repository,
     statichttprepo,
     url as urlmod,
     util,
     wireprotoframing,
     wireprototypes,
     wireprotov1peer,
     wireprotov2peer,
     wireprotov2server,
 )
 from .utils import (
     cborutil,
     interfaceutil,
     stringutil,
 )
 
 httplib = util.httplib
 urlerr = util.urlerr
 urlreq = util.urlreq
 
 def encodevalueinheaders(value, header, limit):
     """Encode a string value into multiple HTTP headers.
 
     ``value`` will be encoded into 1 or more HTTP headers with the names
     ``header-<N>`` where ``<N>`` is an integer starting at 1. Each header
     name + value will be at most ``limit`` bytes long.
 
     Returns an iterable of 2-tuples consisting of header names and
     values as native strings.
     """
     # HTTP Headers are ASCII. Python 3 requires them to be unicodes,
     # not bytes. This function always takes bytes in as arguments.
     fmt = pycompat.strurl(header) + r'-%s'
     # Note: it is *NOT* a bug that the last bit here is a bytestring
     # and not a unicode: we're just getting the encoded length anyway,
     # and using an r-string to make it portable between Python 2 and 3
     # doesn't work because then the \r is a literal backslash-r
     # instead of a carriage return.
     valuelen = limit - len(fmt % r'000') - len(': \r\n')
     result = []
 
     n = 0
     for i in pycompat.xrange(0, len(value), valuelen):
         n += 1
         result.append((fmt % str(n), pycompat.strurl(value[i:i + valuelen])))
 
     return result
 
-def _wraphttpresponse(resp):
-    """Wrap an HTTPResponse with common error handlers.
-
-    This ensures that any I/O from any consumer raises the appropriate
-    error and messaging.
-    """
-    origread = resp.read
-
-    class readerproxy(resp.__class__):
-        def read(self, size=None):
-            try:
-                return origread(size)
-            except httplib.IncompleteRead as e:
-                # e.expected is an integer if length known or None otherwise.
-                if e.expected:
-                    got = len(e.partial)
-                    total = e.expected + got
-                    msg = _('HTTP request error (incomplete response; '
-                            'expected %d bytes got %d)') % (total, got)
-                else:
-                    msg = _('HTTP request error (incomplete response)')
-
-                raise error.PeerTransportError(
-                    msg,
-                    hint=_('this may be an intermittent network failure; '
-                           'if the error persists, consider contacting the '
-                           'network or server operator'))
-            except httplib.HTTPException as e:
-                raise error.PeerTransportError(
-                    _('HTTP request error (%s)') % e,
-                    hint=_('this may be an intermittent network failure; '
-                           'if the error persists, consider contacting the '
-                           'network or server operator'))
-
-    resp.__class__ = readerproxy
-
 class _multifile(object):
     def __init__(self, *fileobjs):
         for f in fileobjs:
             if not util.safehasattr(f, 'length'):
                 raise ValueError(
                     '_multifile only supports file objects that '
                     'have a length but this one does not:', type(f), f)
         self._fileobjs = fileobjs
         self._index = 0
 
     @property
     def length(self):
         return sum(f.length for f in self._fileobjs)
 
     def read(self, amt=None):
         if amt <= 0:
             return ''.join(f.read() for f in self._fileobjs)
         parts = []
         while amt and self._index < len(self._fileobjs):
             parts.append(self._fileobjs[self._index].read(amt))
             got = len(parts[-1])
             if got < amt:
                 self._index += 1
             amt -= got
         return ''.join(parts)
 
     def seek(self, offset, whence=os.SEEK_SET):
         if whence != os.SEEK_SET:
             raise NotImplementedError(
                 '_multifile does not support anything other'
                 ' than os.SEEK_SET for whence on seek()')
         if offset != 0:
             raise NotImplementedError(
                 '_multifile only supports seeking to start, but that '
                 'could be fixed if you need it')
         for f in self._fileobjs:
             f.seek(0)
         self._index = 0
 
 def makev1commandrequest(ui, requestbuilder, caps, capablefn,
                          repobaseurl, cmd, args):
     """Make an HTTP request to run a command for a version 1 client.
 
     ``caps`` is a set of known server capabilities. The value may be
     None if capabilities are not yet known.
 
     ``capablefn`` is a function to evaluate a capability.
 
     ``cmd``, ``args``, and ``data`` define the command, its arguments, and
     raw data to pass to it.
     """
     if cmd == 'pushkey':
         args['data'] = ''
     data = args.pop('data', None)
     headers = args.pop('headers', {})
 
     ui.debug("sending %s command\n" % cmd)
     q = [('cmd', cmd)]
     headersize = 0
     # Important: don't use self.capable() here or else you end up
     # with infinite recursion when trying to look up capabilities
     # for the first time.
     postargsok = caps is not None and 'httppostargs' in caps
 
     # Send arguments via POST.
     if postargsok and args:
         strargs = urlreq.urlencode(sorted(args.items()))
         if not data:
             data = strargs
         else:
             if isinstance(data, bytes):
                 i = io.BytesIO(data)
                 i.length = len(data)
                 data = i
             argsio = io.BytesIO(strargs)
             argsio.length = len(strargs)
             data = _multifile(argsio, data)
         headers[r'X-HgArgs-Post'] = len(strargs)
     elif args:
         # Calling self.capable() can infinite loop if we are calling
         # "capabilities". But that command should never accept wire
         # protocol arguments. So this should never happen.
         assert cmd != 'capabilities'
         httpheader = capablefn('httpheader')
         if httpheader:
             headersize = int(httpheader.split(',', 1)[0])
 
         # Send arguments via HTTP headers.
         if headersize > 0:
             # The headers can typically carry more data than the URL.
             encargs = urlreq.urlencode(sorted(args.items()))
             for header, value in encodevalueinheaders(encargs, 'X-HgArg',
                                                       headersize):
                 headers[header] = value
         # Send arguments via query string (Mercurial <1.9).
         else:
             q += sorted(args.items())
 
     qs = '?%s' % urlreq.urlencode(q)
     cu = "%s%s" % (repobaseurl, qs)
     size = 0
     if util.safehasattr(data, 'length'):
         size = data.length
     elif data is not None:
         size = len(data)
     if data is not None and r'Content-Type' not in headers:
         headers[r'Content-Type'] = r'application/mercurial-0.1'
 
     # Tell the server we accept application/mercurial-0.2 and multiple
     # compression formats if the server is capable of emitting those
     # payloads.
     # Note: Keep this set empty by default, as client advertisement of
     # protocol parameters should only occur after the handshake.
     protoparams = set()
 
     mediatypes = set()
     if caps is not None:
         mt = capablefn('httpmediatype')
         if mt:
             protoparams.add('0.1')
             mediatypes = set(mt.split(','))
 
         protoparams.add('partial-pull')
 
     if '0.2tx' in mediatypes:
         protoparams.add('0.2')
 
     if '0.2tx' in mediatypes and capablefn('compression'):
         # We /could/ compare supported compression formats and prune
         # non-mutually supported or error if nothing is mutually supported.
         # For now, send the full list to the server and have it error.
         comps = [e.wireprotosupport().name for e in
                  util.compengines.supportedwireengines(util.CLIENTROLE)]
         protoparams.add('comp=%s' % ','.join(comps))
 
     if protoparams:
         protoheaders = encodevalueinheaders(' '.join(sorted(protoparams)),
                                             'X-HgProto',
                                             headersize or 1024)
         for header, value in protoheaders:
             headers[header] = value
 
     varyheaders = []
     for header in headers:
         if header.lower().startswith(r'x-hg'):
             varyheaders.append(header)
 
     if varyheaders:
         headers[r'Vary'] = r','.join(sorted(varyheaders))
 
     req = requestbuilder(pycompat.strurl(cu), data, headers)
 
     if data is not None:
         ui.debug("sending %d bytes\n" % size)
         req.add_unredirected_header(r'Content-Length', r'%d' % size)
 
     return req, cu, qs
 
 def _reqdata(req):
     """Get request data, if any. If no data, returns None."""
     if pycompat.ispy3:
         return req.data
     if not req.has_data():
         return None
     return req.get_data()
 
 def sendrequest(ui, opener, req):
     """Send a prepared HTTP request.
 
     Returns the response object.
     """
     dbg = ui.debug
     if (ui.debugflag
         and ui.configbool('devel', 'debug.peer-request')):
         line = 'devel-peer-request: %s\n'
         dbg(line % '%s %s' % (pycompat.bytesurl(req.get_method()),
                               pycompat.bytesurl(req.get_full_url())))
         hgargssize = None
 
         for header, value in sorted(req.header_items()):
             header = pycompat.bytesurl(header)
             value = pycompat.bytesurl(value)
             if header.startswith('X-hgarg-'):
                 if hgargssize is None:
                     hgargssize = 0
                 hgargssize += len(value)
             else:
                 dbg(line % '  %s %s' % (header, value))
 
         if hgargssize is not None:
             dbg(line % '  %d bytes of commands arguments in headers'
                 % hgargssize)
         data = _reqdata(req)
         if data is not None:
             length = getattr(data, 'length', None)
             if length is None:
                 length = len(data)
             dbg(line % '  %d bytes of data' % length)
 
         start = util.timer()
 
     res = None
     try:
         res = opener.open(req)
     except urlerr.httperror as inst:
         if inst.code == 401:
             raise error.Abort(_('authorization failed'))
         raise
     except httplib.HTTPException as inst:
         ui.debug('http error requesting %s\n' %
                  util.hidepassword(req.get_full_url()))
         ui.traceback()
         raise IOError(None, inst)
     finally:
         if ui.debugflag and ui.configbool('devel', 'debug.peer-request'):
             code = res.code if res else -1
             dbg(line % '  finished in %.4f seconds (%d)'
                 % (util.timer() - start, code))
 
     # Insert error handlers for common I/O failures.
-    _wraphttpresponse(res)
+    urlmod.wrapresponse(res)
 
     return res
 
 class RedirectedRepoError(error.RepoError):
     def __init__(self, msg, respurl):
         super(RedirectedRepoError, self).__init__(msg)
         self.respurl = respurl
 
 def parsev1commandresponse(ui, baseurl, requrl, qs, resp, compressible,
                            allowcbor=False):
     # record the url we got redirected to
     redirected = False
     respurl = pycompat.bytesurl(resp.geturl())
     if respurl.endswith(qs):
         respurl = respurl[:-len(qs)]
         qsdropped = False
     else:
         qsdropped = True
 
     if baseurl.rstrip('/') != respurl.rstrip('/'):
         redirected = True
         if not ui.quiet:
             ui.warn(_('real URL is %s\n') % respurl)
 
     try:
         proto = pycompat.bytesurl(resp.getheader(r'content-type', r''))
     except AttributeError:
         proto = pycompat.bytesurl(resp.headers.get(r'content-type', r''))
 
     safeurl = util.hidepassword(baseurl)
     if proto.startswith('application/hg-error'):
         raise error.OutOfBandError(resp.read())
 
     # Pre 1.0 versions of Mercurial used text/plain and
     # application/hg-changegroup. We don't support such old servers.
     if not proto.startswith('application/mercurial-'):
         ui.debug("requested URL: '%s'\n" % util.hidepassword(requrl))
         msg = _("'%s' does not appear to be an hg repository:\n"
                 "---%%<--- (%s)\n%s\n---%%<---\n") % (
             safeurl, proto or 'no content-type', resp.read(1024))
 
         # Some servers may strip the query string from the redirect. We
         # raise a special error type so callers can react to this specially.
         if redirected and qsdropped:
             raise RedirectedRepoError(msg, respurl)
         else:
             raise error.RepoError(msg)
 
     try:
         subtype = proto.split('-', 1)[1]
 
         # Unless we end up supporting CBOR in the legacy wire protocol,
         # this should ONLY be encountered for the initial capabilities
         # request during handshake.
         if subtype == 'cbor':
             if allowcbor:
                 return respurl, proto, resp
             else:
                 raise error.RepoError(_('unexpected CBOR response from '
                                         'server'))
 
         version_info = tuple([int(n) for n in subtype.split('.')])
     except ValueError:
         raise error.RepoError(_("'%s' sent a broken Content-Type "
                                 "header (%s)") % (safeurl, proto))
 
     # TODO consider switching to a decompression reader that uses
     # generators.
     if version_info == (0, 1):
         if compressible:
             resp = util.compengines['zlib'].decompressorreader(resp)
 
     elif version_info == (0, 2):
         # application/mercurial-0.2 always identifies the compression
         # engine in the payload header.
         elen = struct.unpack('B', util.readexactly(resp, 1))[0]
         ename = util.readexactly(resp, elen)
         engine = util.compengines.forwiretype(ename)
 
         resp = engine.decompressorreader(resp)
     else:
         raise error.RepoError(_("'%s' uses newer protocol %s") %
                               (safeurl, subtype))
 
     return respurl, proto, resp
 
 class httppeer(wireprotov1peer.wirepeer):
     def __init__(self, ui, path, url, opener, requestbuilder, caps):
         self.ui = ui
         self._path = path
         self._url = url
         self._caps = caps
         self._urlopener = opener
         self._requestbuilder = requestbuilder
 
     def __del__(self):
         for h in self._urlopener.handlers:
             h.close()
             getattr(h, "close_all", lambda: None)()
 
     # Begin of ipeerconnection interface.
 
     def url(self):
         return self._path
 
     def local(self):
         return None
 
     def peer(self):
         return self
 
     def canpush(self):
         return True
 
     def close(self):
         pass
 
     # End of ipeerconnection interface.
 
     # Begin of ipeercommands interface.
 
     def capabilities(self):
         return self._caps
 
     # End of ipeercommands interface.
 
     def _callstream(self, cmd, _compressible=False, **args):
         args = pycompat.byteskwargs(args)
 
         req, cu, qs = makev1commandrequest(self.ui, self._requestbuilder,
                                            self._caps, self.capable,
                                            self._url, cmd, args)
 
         resp = sendrequest(self.ui, self._urlopener, req)
 
         self._url, ct, resp = parsev1commandresponse(self.ui, self._url, cu, qs,
                                                      resp, _compressible)
 
         return resp
 
     def _call(self, cmd, **args):
         fp = self._callstream(cmd, **args)
         try:
             return fp.read()
         finally:
             # if using keepalive, allow connection to be reused
             fp.close()
 
     def _callpush(self, cmd, cg, **args):
         # have to stream bundle to a temp file because we do not have
         # http 1.1 chunked transfer.
 
         types = self.capable('unbundle')
         try:
             types = types.split(',')
         except AttributeError:
             # servers older than d1b16a746db6 will send 'unbundle' as a
             # boolean capability. They only support headerless/uncompressed
             # bundles.
             types = [""]
         for x in types:
             if x in bundle2.bundletypes:
                 type = x
                 break
 
         tempname = bundle2.writebundle(self.ui, cg, None, type)
         fp = httpconnection.httpsendfile(self.ui, tempname, "rb")
         headers = {r'Content-Type': r'application/mercurial-0.1'}
 
         try:
             r = self._call(cmd, data=fp, headers=headers, **args)
             vals = r.split('\n', 1)
             if len(vals) < 2:
                 raise error.ResponseError(_("unexpected response:"), r)
             return vals
         except urlerr.httperror:
             # Catch and re-raise these so we don't try and treat them
             # like generic socket errors. They lack any values in
             # .args on Python 3 which breaks our socket.error block.
             raise
         except socket.error as err:
             if err.args[0] in (errno.ECONNRESET, errno.EPIPE):
                 raise error.Abort(_('push failed: %s') % err.args[1])
             raise error.Abort(err.args[1])
         finally:
             fp.close()
             os.unlink(tempname)
 
     def _calltwowaystream(self, cmd, fp, **args):
         fh = None
         fp_ = None
         filename = None
         try:
             # dump bundle to disk
             fd, filename = pycompat.mkstemp(prefix="hg-bundle-", suffix=".hg")
             fh = os.fdopen(fd, r"wb")
             d = fp.read(4096)
             while d:
                 fh.write(d)
                 d = fp.read(4096)
             fh.close()
             # start http push
             fp_ = httpconnection.httpsendfile(self.ui, filename, "rb")
             headers = {r'Content-Type': r'application/mercurial-0.1'}
             return self._callstream(cmd, data=fp_, headers=headers, **args)
         finally:
             if fp_ is not None:
                 fp_.close()
             if fh is not None:
                 fh.close()
                 os.unlink(filename)
 
     def _callcompressable(self, cmd, **args):
         return self._callstream(cmd, _compressible=True, **args)
 
     def _abort(self, exception):
         raise exception
 
 def sendv2request(ui, opener, requestbuilder, apiurl, permission, requests):
     reactor = wireprotoframing.clientreactor(hasmultiplesend=False,
                                              buffersends=True)
 
     handler = wireprotov2peer.clienthandler(ui, reactor)
 
     url = '%s/%s' % (apiurl, permission)
 
     if len(requests) > 1:
         url += '/multirequest'
     else:
         url += '/%s' % requests[0][0]
 
     ui.debug('sending %d commands\n' % len(requests))
     for command, args, f in requests:
         ui.debug('sending command %s: %s\n' % (
             command, stringutil.pprint(args, indent=2)))
         assert not list(handler.callcommand(command, args, f))
 
     # TODO stream this.
     body = b''.join(map(bytes, handler.flushcommands()))
 
     # TODO modify user-agent to reflect v2
     headers = {
         r'Accept': wireprotov2server.FRAMINGTYPE,
         r'Content-Type': wireprotov2server.FRAMINGTYPE,
     }
 
     req = requestbuilder(pycompat.strurl(url), body, headers)
     req.add_unredirected_header(r'Content-Length', r'%d' % len(body))
 
     try:
         res = opener.open(req)
     except urlerr.httperror as e:
         if e.code == 401:
             raise error.Abort(_('authorization failed'))
 
         raise
     except httplib.HTTPException as e:
         ui.traceback()
         raise IOError(None, e)
 
     return handler, res
 
 class queuedcommandfuture(pycompat.futures.Future):
     """Wraps result() on command futures to trigger submission on call."""
 
     def result(self, timeout=None):
         if self.done():
             return pycompat.futures.Future.result(self, timeout)
 
         self._peerexecutor.sendcommands()
 
         # sendcommands() will restore the original __class__ and self.result
         # will resolve to Future.result.
         return self.result(timeout)
 
 @interfaceutil.implementer(repository.ipeercommandexecutor)
 class httpv2executor(object):
     def __init__(self, ui, opener, requestbuilder, apiurl, descriptor):
         self._ui = ui
         self._opener = opener
         self._requestbuilder = requestbuilder
         self._apiurl = apiurl
         self._descriptor = descriptor
         self._sent = False
         self._closed = False
         self._neededpermissions = set()
         self._calls = []
         self._futures = weakref.WeakSet()
         self._responseexecutor = None
         self._responsef = None
 
     def __enter__(self):
         return self
 
     def __exit__(self, exctype, excvalue, exctb):
         self.close()
 
     def callcommand(self, command, args):
         if self._sent:
             raise error.ProgrammingError('callcommand() cannot be used after '
                                          'commands are sent')
 
         if self._closed:
             raise error.ProgrammingError('callcommand() cannot be used after '
                                          'close()')
 
         # The service advertises which commands are available. So if we attempt
         # to call an unknown command or pass an unknown argument, we can screen
         # for this.
         if command not in self._descriptor['commands']:
             raise error.ProgrammingError(
                 'wire protocol command %s is not available' % command)
 
         cmdinfo = self._descriptor['commands'][command]
         unknownargs = set(args.keys()) - set(cmdinfo.get('args', {}))
 
         if unknownargs:
             raise error.ProgrammingError(
                 'wire protocol command %s does not accept argument: %s' % (
                     command, ', '.join(sorted(unknownargs))))
 
         self._neededpermissions |= set(cmdinfo['permissions'])
 
         # TODO we /could/ also validate types here, since the API descriptor
         # includes types...
 
         f = pycompat.futures.Future()
 
         # Monkeypatch it so result() triggers sendcommands(), otherwise result()
         # could deadlock.
         f.__class__ = queuedcommandfuture
         f._peerexecutor = self
 
         self._futures.add(f)
         self._calls.append((command, args, f))
 
         return f
 
     def sendcommands(self):
         if self._sent:
             return
 
         if not self._calls:
             return
 
         self._sent = True
 
         # Unhack any future types so caller sees a clean type and so we
         # break reference cycle.
         for f in self._futures:
             if isinstance(f, queuedcommandfuture):
                 f.__class__ = pycompat.futures.Future
                 f._peerexecutor = None
 
         # Mark the future as running and filter out cancelled futures.
         calls = [(command, args, f)
                  for command, args, f in self._calls
                  if f.set_running_or_notify_cancel()]
 
         # Clear out references, prevent improper object usage.
         self._calls = None
 
         if not calls:
             return
 
         permissions = set(self._neededpermissions)
 
         if 'push' in permissions and 'pull' in permissions:
             permissions.remove('pull')
 
         if len(permissions) > 1:
             raise error.RepoError(_('cannot make request requiring multiple '
                                     'permissions: %s') %
                                   _(', ').join(sorted(permissions)))
 
         permission = {
             'push': 'rw',
             'pull': 'ro',
         }[permissions.pop()]
 
         handler, resp = sendv2request(
             self._ui, self._opener, self._requestbuilder, self._apiurl,
             permission, calls)
 
         # TODO we probably want to validate the HTTP code, media type, etc.
 
         self._responseexecutor = pycompat.futures.ThreadPoolExecutor(1)
         self._responsef = self._responseexecutor.submit(self._handleresponse,
                                                         handler, resp)
 
     def close(self):
         if self._closed:
             return
 
         self.sendcommands()
 
         self._closed = True
 
         if not self._responsef:
             return
 
         # TODO ^C here may not result in immediate program termination.
 
         try:
             self._responsef.result()
         finally:
             self._responseexecutor.shutdown(wait=True)
             self._responsef = None
             self._responseexecutor = None
 
             # If any of our futures are still in progress, mark them as
             # errored, otherwise a result() could wait indefinitely.
             for f in self._futures:
                 if not f.done():
                     f.set_exception(error.ResponseError(
                         _('unfulfilled command response')))
 
             self._futures = None
 
     def _handleresponse(self, handler, resp):
         # Called in a thread to read the response.
 
         while handler.readframe(resp):
             pass
 
 # TODO implement interface for version 2 peers
 @interfaceutil.implementer(repository.ipeerconnection,
                            repository.ipeercapabilities,
                            repository.ipeerrequests)
 class httpv2peer(object):
     def __init__(self, ui, repourl, apipath, opener, requestbuilder,
                  apidescriptor):
         self.ui = ui
 
         if repourl.endswith('/'):
             repourl = repourl[:-1]
 
         self._url = repourl
         self._apipath = apipath
         self._apiurl = '%s/%s' % (repourl, apipath)
         self._opener = opener
         self._requestbuilder = requestbuilder
         self._descriptor = apidescriptor
 
     # Start of ipeerconnection.
 
     def url(self):
         return self._url
 
     def local(self):
         return None
 
     def peer(self):
         return self
 
     def canpush(self):
         # TODO change once implemented.
         return False
 
     def close(self):
         pass
 
     # End of ipeerconnection.
 
     # Start of ipeercapabilities.
 
     def capable(self, name):
         # The capabilities used internally historically map to capabilities
         # advertised from the "capabilities" wire protocol command. However,
         # version 2 of that command works differently.
 
         # Maps to commands that are available.
         if name in ('branchmap', 'getbundle', 'known', 'lookup', 'pushkey'):
             return True
 
         # Other concepts.
         if name in ('bundle2'):
             return True
 
         # Alias command-* to presence of command of that name.
         if name.startswith('command-'):
             return name[len('command-'):] in self._descriptor['commands']
 
         return False
 
     def requirecap(self, name, purpose):
         if self.capable(name):
             return
 
         raise error.CapabilityError(
             _('cannot %s; client or remote repository does not support the %r '
               'capability') % (purpose, name))
 
     # End of ipeercapabilities.
 
     def _call(self, name, **args):
         with self.commandexecutor() as e:
             return e.callcommand(name, args).result()
 
     def commandexecutor(self):
         return httpv2executor(self.ui, self._opener, self._requestbuilder,
                               self._apiurl, self._descriptor)
 
 # Registry of API service names to metadata about peers that handle it.
 #
 # The following keys are meaningful:
 #
 # init
 #    Callable receiving (ui, repourl, servicepath, opener, requestbuilder,
 #                        apidescriptor) to create a peer.
 #
 # priority
 #    Integer priority for the service. If we could choose from multiple
 #    services, we choose the one with the highest priority.
 API_PEERS = {
     wireprototypes.HTTP_WIREPROTO_V2: {
         'init': httpv2peer,
         'priority': 50,
     },
 }
 
 def performhandshake(ui, url, opener, requestbuilder):
     # The handshake is a request to the capabilities command.
 
     caps = None
     def capable(x):
         raise error.ProgrammingError('should not be called')
 
     args = {}
 
     # The client advertises support for newer protocols by adding an
     # X-HgUpgrade-* header with a list of supported APIs and an
     # X-HgProto-* header advertising which serializing formats it supports.
     # We only support the HTTP version 2 transport and CBOR responses for
     # now.
     advertisev2 = ui.configbool('experimental', 'httppeer.advertise-v2')
 
     if advertisev2:
         args['headers'] = {
             r'X-HgProto-1': r'cbor',
         }
 
         args['headers'].update(
             encodevalueinheaders(' '.join(sorted(API_PEERS)),
                                  'X-HgUpgrade',
                                  # We don't know the header limit this early.
                                  # So make it small.
                                  1024))
 
     req, requrl, qs = makev1commandrequest(ui, requestbuilder, caps,
                                            capable, url, 'capabilities',
                                            args)
     resp = sendrequest(ui, opener, req)
 
     # The server may redirect us to the repo root, stripping the
     # ?cmd=capabilities query string from the URL. The server would likely
     # return HTML in this case and ``parsev1commandresponse()`` would raise.
     # We catch this special case and re-issue the capabilities request against
     # the new URL.
     #
     # We should ideally not do this, as a redirect that drops the query
     # string from the URL is arguably a server bug. (Garbage in, garbage out).
     # However,  Mercurial clients for several years appeared to handle this
     # issue without behavior degradation. And according to issue 5860, it may
     # be a longstanding bug in some server implementations. So we allow a
     # redirect that drops the query string to "just work."
     try:
         respurl, ct, resp = parsev1commandresponse(ui, url, requrl, qs, resp,
                                                    compressible=False,
                                                    allowcbor=advertisev2)
     except RedirectedRepoError as e:
         req, requrl, qs = makev1commandrequest(ui, requestbuilder, caps,
                                                capable, e.respurl,
                                                'capabilities', args)
         resp = sendrequest(ui, opener, req)
         respurl, ct, resp = parsev1commandresponse(ui, url, requrl, qs, resp,
                                                    compressible=False,
                                                    allowcbor=advertisev2)
 
     try:
         rawdata = resp.read()
     finally:
         resp.close()
 
     if not ct.startswith('application/mercurial-'):
         raise error.ProgrammingError('unexpected content-type: %s' % ct)
 
     if advertisev2:
         if ct == 'application/mercurial-cbor':
             try:
                 info = cborutil.decodeall(rawdata)[0]
             except cborutil.CBORDecodeError:
                 raise error.Abort(_('error decoding CBOR from remote server'),
                                   hint=_('try again and consider contacting '
                                          'the server operator'))
 
         # We got a legacy response. That's fine.
         elif ct in ('application/mercurial-0.1', 'application/mercurial-0.2'):
             info = {
                 'v1capabilities': set(rawdata.split())
             }
 
         else:
             raise error.RepoError(
                 _('unexpected response type from server: %s') % ct)
     else:
         info = {
             'v1capabilities': set(rawdata.split())
         }
 
     return respurl, info
 
 def makepeer(ui, path, opener=None, requestbuilder=urlreq.request):
     """Construct an appropriate HTTP peer instance.
 
     ``opener`` is an ``url.opener`` that should be used to establish
     connections, perform HTTP requests.
 
     ``requestbuilder`` is the type used for constructing HTTP requests.
     It exists as an argument so extensions can override the default.
     """
     u = util.url(path)
     if u.query or u.fragment:
         raise error.Abort(_('unsupported URL component: "%s"') %
                           (u.query or u.fragment))
 
     # urllib cannot handle URLs with embedded user or passwd.
     url, authinfo = u.authinfo()
     ui.debug('using %s\n' % url)
 
     opener = opener or urlmod.opener(ui, authinfo)
 
     respurl, info = performhandshake(ui, url, opener, requestbuilder)
 
     # Given the intersection of APIs that both we and the server support,
     # sort by their advertised priority and pick the first one.
     #
     # TODO consider making this request-based and interface driven. For
     # example, the caller could say "I want a peer that does X." It's quite
     # possible that not all peers would do that. Since we know the service
     # capabilities, we could filter out services not meeting the
     # requirements. Possibly by consulting the interfaces defined by the
     # peer type.
     apipeerchoices = set(info.get('apis', {}).keys()) & set(API_PEERS.keys())
 
     preferredchoices = sorted(apipeerchoices,
                               key=lambda x: API_PEERS[x]['priority'],
                               reverse=True)
 
     for service in preferredchoices:
         apipath = '%s/%s' % (info['apibase'].rstrip('/'), service)
 
         return API_PEERS[service]['init'](ui, respurl, apipath, opener,
                                           requestbuilder,
                                           info['apis'][service])
 
     # Failed to construct an API peer. Fall back to legacy.
     return httppeer(ui, path, respurl, opener, requestbuilder,
                     info['v1capabilities'])
 
 def instance(ui, path, create, intents=None, createopts=None):
     if create:
         raise error.Abort(_('cannot create new http repository'))
     try:
         if path.startswith('https:') and not urlmod.has_https:
             raise error.Abort(_('Python support for SSL and HTTPS '
                                 'is not installed'))
 
         inst = makepeer(ui, path)
 
         return inst
     except error.RepoError as httpexception:
         try:
             r = statichttprepo.instance(ui, "static-" + path, create)
             ui.note(_('(falling back to static-http)\n'))
             return r
         except error.RepoError:
             raise httpexception # use the original http RepoError instead

mercurial/url.py

@@ -1,597 +1,633 b''
 # url.py - HTTP handling for mercurial
 #
 # Copyright 2005, 2006, 2007, 2008 Matt Mackall <[e-mail]>
 # Copyright 2006, 2007 Alexis S. L. Carvalho <[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.
 
 from __future__ import absolute_import
 
 import base64
 import os
 import socket
 import sys
 
 from .i18n import _
 from . import (
     encoding,
     error,
     httpconnection as httpconnectionmod,
     keepalive,
     pycompat,
     sslutil,
     urllibcompat,
     util,
 )
 from .utils import (
     stringutil,
 )
 
 httplib = util.httplib
 stringio = util.stringio
 urlerr = util.urlerr
 urlreq = util.urlreq
 
 def escape(s, quote=None):
     '''Replace special characters "&", "<" and ">" to HTML-safe sequences.
     If the optional flag quote is true, the quotation mark character (")
     is also translated.
 
     This is the same as cgi.escape in Python, but always operates on
     bytes, whereas cgi.escape in Python 3 only works on unicodes.
     '''
     s = s.replace(b"&", b"&amp;")
     s = s.replace(b"<", b"&lt;")
     s = s.replace(b">", b"&gt;")
     if quote:
         s = s.replace(b'"', b"&quot;")
     return s
 
 class passwordmgr(object):
     def __init__(self, ui, passwddb):
         self.ui = ui
         self.passwddb = passwddb
 
     def add_password(self, realm, uri, user, passwd):
         return self.passwddb.add_password(realm, uri, user, passwd)
 
     def find_user_password(self, realm, authuri):
         authinfo = self.passwddb.find_user_password(realm, authuri)
         user, passwd = authinfo
         if user and passwd:
             self._writedebug(user, passwd)
             return (user, passwd)
 
         if not user or not passwd:
             res = httpconnectionmod.readauthforuri(self.ui, authuri, user)
             if res:
                 group, auth = res
                 user, passwd = auth.get('username'), auth.get('password')
                 self.ui.debug("using auth.%s.* for authentication\n" % group)
         if not user or not passwd:
             u = util.url(pycompat.bytesurl(authuri))
             u.query = None
             if not self.ui.interactive():
                 raise error.Abort(_('http authorization required for %s') %
                                   util.hidepassword(bytes(u)))
 
             self.ui.write(_("http authorization required for %s\n") %
                           util.hidepassword(bytes(u)))
             self.ui.write(_("realm: %s\n") % pycompat.bytesurl(realm))
             if user:
                 self.ui.write(_("user: %s\n") % user)
             else:
                 user = self.ui.prompt(_("user:"), default=None)
 
             if not passwd:
                 passwd = self.ui.getpass()
 
         self.passwddb.add_password(realm, authuri, user, passwd)
         self._writedebug(user, passwd)
         return (user, passwd)
 
     def _writedebug(self, user, passwd):
         msg = _('http auth: user %s, password %s\n')
         self.ui.debug(msg % (user, passwd and '*' * len(passwd) or 'not set'))
 
     def find_stored_password(self, authuri):
         return self.passwddb.find_user_password(None, authuri)
 
 class proxyhandler(urlreq.proxyhandler):
     def __init__(self, ui):
         proxyurl = (ui.config("http_proxy", "host") or
                         encoding.environ.get('http_proxy'))
         # XXX proxyauthinfo = None
 
         if proxyurl:
             # proxy can be proper url or host[:port]
             if not (proxyurl.startswith('http:') or
                     proxyurl.startswith('https:')):
                 proxyurl = 'http://' + proxyurl + '/'
             proxy = util.url(proxyurl)
             if not proxy.user:
                 proxy.user = ui.config("http_proxy", "user")
                 proxy.passwd = ui.config("http_proxy", "passwd")
 
             # see if we should use a proxy for this url
             no_list = ["localhost", "127.0.0.1"]
             no_list.extend([p.lower() for
                             p in ui.configlist("http_proxy", "no")])
             no_list.extend([p.strip().lower() for
                             p in encoding.environ.get("no_proxy", '').split(',')
                             if p.strip()])
             # "http_proxy.always" config is for running tests on localhost
             if ui.configbool("http_proxy", "always"):
                 self.no_list = []
             else:
                 self.no_list = no_list
 
             proxyurl = bytes(proxy)
             proxies = {'http': proxyurl, 'https': proxyurl}
             ui.debug('proxying through %s\n' % util.hidepassword(proxyurl))
         else:
             proxies = {}
 
         urlreq.proxyhandler.__init__(self, proxies)
         self.ui = ui
 
     def proxy_open(self, req, proxy, type_):
         host = urllibcompat.gethost(req).split(':')[0]
         for e in self.no_list:
             if host == e:
                 return None
             if e.startswith('*.') and host.endswith(e[2:]):
                 return None
             if e.startswith('.') and host.endswith(e[1:]):
                 return None
 
         return urlreq.proxyhandler.proxy_open(self, req, proxy, type_)
 
 def _gen_sendfile(orgsend):
     def _sendfile(self, data):
         # send a file
         if isinstance(data, httpconnectionmod.httpsendfile):
             # if auth required, some data sent twice, so rewind here
             data.seek(0)
             for chunk in util.filechunkiter(data):
                 orgsend(self, chunk)
         else:
             orgsend(self, data)
     return _sendfile
 
 has_https = util.safehasattr(urlreq, 'httpshandler')
 
 class httpconnection(keepalive.HTTPConnection):
     # must be able to send big bundle as stream.
     send = _gen_sendfile(keepalive.HTTPConnection.send)
 
     def getresponse(self):
         proxyres = getattr(self, 'proxyres', None)
         if proxyres:
             if proxyres.will_close:
                 self.close()
             self.proxyres = None
             return proxyres
         return keepalive.HTTPConnection.getresponse(self)
 
 # general transaction handler to support different ways to handle
 # HTTPS proxying before and after Python 2.6.3.
 def _generic_start_transaction(handler, h, req):
     tunnel_host = getattr(req, '_tunnel_host', None)
     if tunnel_host:
         if tunnel_host[:7] not in ['http://', 'https:/']:
             tunnel_host = 'https://' + tunnel_host
         new_tunnel = True
     else:
         tunnel_host = urllibcompat.getselector(req)
         new_tunnel = False
 
     if new_tunnel or tunnel_host == urllibcompat.getfullurl(req): # has proxy
         u = util.url(tunnel_host)
         if new_tunnel or u.scheme == 'https': # only use CONNECT for HTTPS
             h.realhostport = ':'.join([u.host, (u.port or '443')])
             h.headers = req.headers.copy()
             h.headers.update(handler.parent.addheaders)
             return
 
     h.realhostport = None
     h.headers = None
 
 def _generic_proxytunnel(self):
     proxyheaders = dict(
             [(x, self.headers[x]) for x in self.headers
              if x.lower().startswith('proxy-')])
     self.send('CONNECT %s HTTP/1.0\r\n' % self.realhostport)
     for header in proxyheaders.iteritems():
         self.send('%s: %s\r\n' % header)
     self.send('\r\n')
 
     # majority of the following code is duplicated from
     # httplib.HTTPConnection as there are no adequate places to
     # override functions to provide the needed functionality
     res = self.response_class(self.sock,
                               strict=self.strict,
                               method=self._method)
 
     while True:
         version, status, reason = res._read_status()
         if status != httplib.CONTINUE:
             break
         # skip lines that are all whitespace
         list(iter(lambda: res.fp.readline().strip(), ''))
     res.status = status
     res.reason = reason.strip()
 
     if res.status == 200:
         # skip lines until we find a blank line
         list(iter(res.fp.readline, '\r\n'))
         return True
 
     if version == 'HTTP/1.0':
         res.version = 10
     elif version.startswith('HTTP/1.'):
         res.version = 11
     elif version == 'HTTP/0.9':
         res.version = 9
     else:
         raise httplib.UnknownProtocol(version)
 
     if res.version == 9:
         res.length = None
         res.chunked = 0
         res.will_close = 1
         res.msg = httplib.HTTPMessage(stringio())
         return False
 
     res.msg = httplib.HTTPMessage(res.fp)
     res.msg.fp = None
 
     # are we using the chunked-style of transfer encoding?
     trenc = res.msg.getheader('transfer-encoding')
     if trenc and trenc.lower() == "chunked":
         res.chunked = 1
         res.chunk_left = None
     else:
         res.chunked = 0
 
     # will the connection close at the end of the response?
     res.will_close = res._check_close()
 
     # do we have a Content-Length?
     # NOTE: RFC 2616, section 4.4, #3 says we ignore this if
     # transfer-encoding is "chunked"
     length = res.msg.getheader('content-length')
     if length and not res.chunked:
         try:
             res.length = int(length)
         except ValueError:
             res.length = None
         else:
             if res.length < 0:  # ignore nonsensical negative lengths
                 res.length = None
     else:
         res.length = None
 
     # does the body have a fixed length? (of zero)
     if (status == httplib.NO_CONTENT or status == httplib.NOT_MODIFIED or
         100 <= status < 200 or # 1xx codes
         res._method == 'HEAD'):
         res.length = 0
 
     # if the connection remains open, and we aren't using chunked, and
     # a content-length was not provided, then assume that the connection
     # WILL close.
     if (not res.will_close and
        not res.chunked and
        res.length is None):
         res.will_close = 1
 
     self.proxyres = res
 
     return False
 
 class httphandler(keepalive.HTTPHandler):
     def http_open(self, req):
         return self.do_open(httpconnection, req)
 
     def _start_transaction(self, h, req):
         _generic_start_transaction(self, h, req)
         return keepalive.HTTPHandler._start_transaction(self, h, req)
 
 class logginghttpconnection(keepalive.HTTPConnection):
     def __init__(self, createconn, *args, **kwargs):
         keepalive.HTTPConnection.__init__(self, *args, **kwargs)
         self._create_connection = createconn
 
     if sys.version_info < (2, 7, 7):
         # copied from 2.7.14, since old implementations directly call
         # socket.create_connection()
         def connect(self):
             self.sock = self._create_connection((self.host, self.port),
                                                 self.timeout,
                                                 self.source_address)
             if self._tunnel_host:
                 self._tunnel()
 
 class logginghttphandler(httphandler):
     """HTTP handler that logs socket I/O."""
     def __init__(self, logfh, name, observeropts):
         super(logginghttphandler, self).__init__()
 
         self._logfh = logfh
         self._logname = name
         self._observeropts = observeropts
 
     # do_open() calls the passed class to instantiate an HTTPConnection. We
     # pass in a callable method that creates a custom HTTPConnection instance
     # whose callback to create the socket knows how to proxy the socket.
     def http_open(self, req):
         return self.do_open(self._makeconnection, req)
 
     def _makeconnection(self, *args, **kwargs):
         def createconnection(*args, **kwargs):
             sock = socket.create_connection(*args, **kwargs)
             return util.makeloggingsocket(self._logfh, sock, self._logname,
                                           **self._observeropts)
 
         return logginghttpconnection(createconnection, *args, **kwargs)
 
 if has_https:
     class httpsconnection(httplib.HTTPConnection):
         response_class = keepalive.HTTPResponse
         default_port = httplib.HTTPS_PORT
         # must be able to send big bundle as stream.
         send = _gen_sendfile(keepalive.safesend)
         getresponse = keepalive.wrapgetresponse(httplib.HTTPConnection)
 
         def __init__(self, host, port=None, key_file=None, cert_file=None,
                      *args, **kwargs):
             httplib.HTTPConnection.__init__(self, host, port, *args, **kwargs)
             self.key_file = key_file
             self.cert_file = cert_file
 
         def connect(self):
             self.sock = socket.create_connection((self.host, self.port))
 
             host = self.host
             if self.realhostport: # use CONNECT proxy
                 _generic_proxytunnel(self)
                 host = self.realhostport.rsplit(':', 1)[0]
             self.sock = sslutil.wrapsocket(
                 self.sock, self.key_file, self.cert_file, ui=self.ui,
                 serverhostname=host)
             sslutil.validatesocket(self.sock)
 
     class httpshandler(keepalive.KeepAliveHandler, urlreq.httpshandler):
         def __init__(self, ui):
             keepalive.KeepAliveHandler.__init__(self)
             urlreq.httpshandler.__init__(self)
             self.ui = ui
             self.pwmgr = passwordmgr(self.ui,
                                      self.ui.httppasswordmgrdb)
 
         def _start_transaction(self, h, req):
             _generic_start_transaction(self, h, req)
             return keepalive.KeepAliveHandler._start_transaction(self, h, req)
 
         def https_open(self, req):
             # urllibcompat.getfullurl() does not contain credentials
             # and we may need them to match the certificates.
             url = urllibcompat.getfullurl(req)
             user, password = self.pwmgr.find_stored_password(url)
             res = httpconnectionmod.readauthforuri(self.ui, url, user)
             if res:
                 group, auth = res
                 self.auth = auth
                 self.ui.debug("using auth.%s.* for authentication\n" % group)
             else:
                 self.auth = None
             return self.do_open(self._makeconnection, req)
 
         def _makeconnection(self, host, port=None, *args, **kwargs):
             keyfile = None
             certfile = None
 
             if len(args) >= 1: # key_file
                 keyfile = args[0]
             if len(args) >= 2: # cert_file
                 certfile = args[1]
             args = args[2:]
 
             # if the user has specified different key/cert files in
             # hgrc, we prefer these
             if self.auth and 'key' in self.auth and 'cert' in self.auth:
                 keyfile = self.auth['key']
                 certfile = self.auth['cert']
 
             conn = httpsconnection(host, port, keyfile, certfile, *args,
                                    **kwargs)
             conn.ui = self.ui
             return conn
 
 class httpdigestauthhandler(urlreq.httpdigestauthhandler):
     def __init__(self, *args, **kwargs):
         urlreq.httpdigestauthhandler.__init__(self, *args, **kwargs)
         self.retried_req = None
 
     def reset_retry_count(self):
         # Python 2.6.5 will call this on 401 or 407 errors and thus loop
         # forever. We disable reset_retry_count completely and reset in
         # http_error_auth_reqed instead.
         pass
 
     def http_error_auth_reqed(self, auth_header, host, req, headers):
         # Reset the retry counter once for each request.
         if req is not self.retried_req:
             self.retried_req = req
             self.retried = 0
         return urlreq.httpdigestauthhandler.http_error_auth_reqed(
                     self, auth_header, host, req, headers)
 
 class httpbasicauthhandler(urlreq.httpbasicauthhandler):
     def __init__(self, *args, **kwargs):
         self.auth = None
         urlreq.httpbasicauthhandler.__init__(self, *args, **kwargs)
         self.retried_req = None
 
     def http_request(self, request):
         if self.auth:
             request.add_unredirected_header(self.auth_header, self.auth)
 
         return request
 
     def https_request(self, request):
         if self.auth:
             request.add_unredirected_header(self.auth_header, self.auth)
 
         return request
 
     def reset_retry_count(self):
         # Python 2.6.5 will call this on 401 or 407 errors and thus loop
         # forever. We disable reset_retry_count completely and reset in
         # http_error_auth_reqed instead.
         pass
 
     def http_error_auth_reqed(self, auth_header, host, req, headers):
         # Reset the retry counter once for each request.
         if req is not self.retried_req:
             self.retried_req = req
             self.retried = 0
         return urlreq.httpbasicauthhandler.http_error_auth_reqed(
                         self, auth_header, host, req, headers)
 
     def retry_http_basic_auth(self, host, req, realm):
         user, pw = self.passwd.find_user_password(
             realm, urllibcompat.getfullurl(req))
         if pw is not None:
             raw = "%s:%s" % (pycompat.bytesurl(user), pycompat.bytesurl(pw))
             auth = r'Basic %s' % pycompat.strurl(base64.b64encode(raw).strip())
             if req.get_header(self.auth_header, None) == auth:
                 return None
             self.auth = auth
             req.add_unredirected_header(self.auth_header, auth)
             return self.parent.open(req)
         else:
             return None
 
 class cookiehandler(urlreq.basehandler):
     def __init__(self, ui):
         self.cookiejar = None
 
         cookiefile = ui.config('auth', 'cookiefile')
         if not cookiefile:
             return
 
         cookiefile = util.expandpath(cookiefile)
         try:
             cookiejar = util.cookielib.MozillaCookieJar(
                 pycompat.fsdecode(cookiefile))
             cookiejar.load()
             self.cookiejar = cookiejar
         except util.cookielib.LoadError as e:
             ui.warn(_('(error loading cookie file %s: %s; continuing without '
                       'cookies)\n') % (cookiefile, stringutil.forcebytestr(e)))
 
     def http_request(self, request):
         if self.cookiejar:
             self.cookiejar.add_cookie_header(request)
 
         return request
 
     def https_request(self, request):
         if self.cookiejar:
             self.cookiejar.add_cookie_header(request)
 
         return request
 
 handlerfuncs = []
 
 def opener(ui, authinfo=None, useragent=None, loggingfh=None,
            loggingname=b's', loggingopts=None, sendaccept=True):
     '''
     construct an opener suitable for urllib2
     authinfo will be added to the password manager
 
     The opener can be configured to log socket events if the various
     ``logging*`` arguments are specified.
 
     ``loggingfh`` denotes a file object to log events to.
     ``loggingname`` denotes the name of the to print when logging.
     ``loggingopts`` is a dict of keyword arguments to pass to the constructed
     ``util.socketobserver`` instance.
 
     ``sendaccept`` allows controlling whether the ``Accept`` request header
     is sent. The header is sent by default.
     '''
     handlers = []
 
     if loggingfh:
         handlers.append(logginghttphandler(loggingfh, loggingname,
                                            loggingopts or {}))
         # We don't yet support HTTPS when logging I/O. If we attempt to open
         # an HTTPS URL, we'll likely fail due to unknown protocol.
 
     else:
         handlers.append(httphandler())
         if has_https:
             handlers.append(httpshandler(ui))
 
     handlers.append(proxyhandler(ui))
 
     passmgr = passwordmgr(ui, ui.httppasswordmgrdb)
     if authinfo is not None:
         realm, uris, user, passwd = authinfo
         saveduser, savedpass = passmgr.find_stored_password(uris[0])
         if user != saveduser or passwd:
             passmgr.add_password(realm, uris, user, passwd)
         ui.debug('http auth: user %s, password %s\n' %
                  (user, passwd and '*' * len(passwd) or 'not set'))
 
     handlers.extend((httpbasicauthhandler(passmgr),
                      httpdigestauthhandler(passmgr)))
     handlers.extend([h(ui, passmgr) for h in handlerfuncs])
     handlers.append(cookiehandler(ui))
     opener = urlreq.buildopener(*handlers)
 
     # The user agent should should *NOT* be used by servers for e.g.
     # protocol detection or feature negotiation: there are other
     # facilities for that.
     #
     # "mercurial/proto-1.0" was the original user agent string and
     # exists for backwards compatibility reasons.
     #
     # The "(Mercurial %s)" string contains the distribution
     # name and version. Other client implementations should choose their
     # own distribution name. Since servers should not be using the user
     # agent string for anything, clients should be able to define whatever
     # user agent they deem appropriate.
     #
     # The custom user agent is for lfs, because unfortunately some servers
     # do look at this value.
     if not useragent:
         agent = 'mercurial/proto-1.0 (Mercurial %s)' % util.version()
         opener.addheaders = [(r'User-agent', pycompat.sysstr(agent))]
     else:
         opener.addheaders = [(r'User-agent', pycompat.sysstr(useragent))]
 
     # This header should only be needed by wire protocol requests. But it has
     # been sent on all requests since forever. We keep sending it for backwards
     # compatibility reasons. Modern versions of the wire protocol use
     # X-HgProto-<N> for advertising client support.
     if sendaccept:
         opener.addheaders.append((r'Accept', r'application/mercurial-0.1'))
 
     return opener
 
 def open(ui, url_, data=None):
     u = util.url(url_)
     if u.scheme:
         u.scheme = u.scheme.lower()
         url_, authinfo = u.authinfo()
     else:
         path = util.normpath(os.path.abspath(url_))
         url_ = 'file://' + pycompat.bytesurl(urlreq.pathname2url(path))
         authinfo = None
     return opener(ui, authinfo).open(pycompat.strurl(url_), data)
+
+def wrapresponse(resp):
+    """Wrap a response object with common error handlers.
+
+    This ensures that any I/O from any consumer raises the appropriate
+    error and messaging.
+    """
+    origread = resp.read
+
+    class readerproxy(resp.__class__):
+        def read(self, size=None):
+            try:
+                return origread(size)
+            except httplib.IncompleteRead as e:
+                # e.expected is an integer if length known or None otherwise.
+                if e.expected:
+                    got = len(e.partial)
+                    total = e.expected + got
+                    msg = _('HTTP request error (incomplete response; '
+                            'expected %d bytes got %d)') % (total, got)
+                else:
+                    msg = _('HTTP request error (incomplete response)')
+
+                raise error.PeerTransportError(
+                    msg,
+                    hint=_('this may be an intermittent network failure; '
+                           'if the error persists, consider contacting the '
+                           'network or server operator'))
+            except httplib.HTTPException as e:
+                raise error.PeerTransportError(
+                    _('HTTP request error (%s)') % e,
+                    hint=_('this may be an intermittent network failure; '
+                           'if the error persists, consider contacting the '
+                           'network or server operator'))
+
+    resp.__class__ = readerproxy