##// END OF EJS Templates
wireproto: add streams to frame-based protocol...
wireproto: add streams to frame-based protocol Previously, the frame-based protocol was just a series of frames, with each frame associated with a request ID. In order to scale the protocol, we'll want to enable the use of compression. While it is possible to enable compression at the socket/pipe level, this has its disadvantages. The big one is it undermines the point of frames being standalone, atomic units that can be read and written: if you add compression above the framing protocol, you are back to having a stream-based protocol as opposed to something frame-based. So in order to preserve frames, compression needs to occur at the frame payload level. Compressing each frame's payload individually will limit compression ratios because the window size of the compressor will be limited by the max frame size, which is 32-64kb as currently defined. It will also add CPU overhead, as it is more efficient for compressors to operate on fewer, larger blocks of data than more, smaller blocks. So compressing each frame independently is out. This means we need to compress each frame's payload as if it is part of a larger stream. The simplest approach is to have 1 stream per connection. This could certainly work. However, it has disadvantages (documented below). We could also have 1 stream per RPC/command invocation. (This is the model HTTP/2 goes with.) This also has disadvantages. The main disadvantage to one global stream is that it has the very real potential to create CPU bottlenecks doing compression. Networks are only getting faster and the performance of single CPU cores has been relatively flat. Newer compression formats like zstandard offer better CPU cycle efficiency than predecessors like zlib. But it still all too common to saturate your CPU with compression overhead long before you saturate the network pipe. The main disadvantage with streams per request is that you can't reap the benefits of the compression context for multiple requests. For example, if you send 1000 RPC requests (or HTTP/2 requests for that matter), the response to each would have its own compression context. The overall size of the raw responses would be larger because compression contexts wouldn't be able to reference data from another request or response. The approach for streams as implemented in this commit is to support N streams per connection and for streams to potentially span requests and responses. As explained by the added internals docs, this facilitates servers and clients delegating independent streams and compression to independent threads / CPU cores. This helps alleviate the CPU bottleneck of compression. This design also allows compression contexts to be reused across requests/responses. This can result in improved compression ratios and less overhead for compressors and decompressors having to build new contexts. Another feature that was defined was the ability for individual frames within a stream to declare whether that individual frame's payload uses the content encoding (read: compression) defined by the stream. The idea here is that some servers may serve data from a combination of caches and dynamic resolution. Data coming from caches may be pre-compressed. We want to facilitate servers being able to essentially stream bytes from caches to the wire with minimal overhead. Being able to mix and match with frames are compressed within a stream enables these types of advanced server functionality. This commit defines the new streams mechanism. Basic code for supporting streams in frames has been added. But that code is seriously lacking and doesn't fully conform to the defined protocol. For example, we don't close any streams. And support for content encoding within streams is not yet implemented. The change was rather invasive and I didn't think it would be reasonable to implement the entire feature in a single commit. For the record, I would have loved to reuse an existing multiplexing protocol to build the new wire protocol on top of. However, I couldn't find a protocol that offers the performance and scaling characteristics that I desired. Namely, it should support multiple compression contexts to facilitate scaling out to multiple CPU cores and compression contexts should be able to live longer than single RPC requests. HTTP/2 *almost* fits the bill. But the semantics of HTTP message exchange state that streams can only live for a single request-response. We /could/ tunnel on top of HTTP/2 streams and frames with HEADER and DATA frames. But there's no guarantee that HTTP/2 libraries and proxies would allow us to use HTTP/2 streams and frames without the HTTP message exchange semantics defined in RFC 7540 Section 8. Other RPC protocols like gRPC tunnel are built on top of HTTP/2 and thus preserve its semantics of stream per RPC invocation. Even QUIC does this. We could attempt to invent a higher-level stream that spans HTTP/2 streams. But this would be violating HTTP/2 because there is no guarantee that HTTP/2 streams are routed to the same server. The best we can do - which is what this protocol does - is shoehorn all request and response data into a single HTTP message and create streams within. At that point, we've defined a Content-Type in HTTP parlance. It just so happens our media type can also work as a standalone, stream-based protocol, without leaning on HTTP or similar protocol. Differential Revision: https://phab.mercurial-scm.org/D2907

File last commit:

r37113:e24802ea default
r37304:9bfcbe4f default
Show More
windows.py
495 lines | 14.7 KiB | text/x-python | PythonLexer
# windows.py - Windows utility function implementations for Mercurial
#
# Copyright 2005-2009 Matt Mackall <mpm@selenic.com> 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 errno
import msvcrt
import os
import re
import stat
import sys
from .i18n import _
from . import (
encoding,
error,
policy,
pycompat,
win32,
)
try:
import _winreg as winreg
winreg.CloseKey
except ImportError:
import winreg
osutil = policy.importmod(r'osutil')
getfsmountpoint = win32.getvolumename
getfstype = win32.getfstype
getuser = win32.getuser
hidewindow = win32.hidewindow
makedir = win32.makedir
nlinks = win32.nlinks
oslink = win32.oslink
samedevice = win32.samedevice
samefile = win32.samefile
setsignalhandler = win32.setsignalhandler
spawndetached = win32.spawndetached
split = os.path.split
testpid = win32.testpid
unlink = win32.unlink
umask = 0o022
class mixedfilemodewrapper(object):
"""Wraps a file handle when it is opened in read/write mode.
fopen() and fdopen() on Windows have a specific-to-Windows requirement
that files opened with mode r+, w+, or a+ make a call to a file positioning
function when switching between reads and writes. Without this extra call,
Python will raise a not very intuitive "IOError: [Errno 0] Error."
This class wraps posixfile instances when the file is opened in read/write
mode and automatically adds checks or inserts appropriate file positioning
calls when necessary.
"""
OPNONE = 0
OPREAD = 1
OPWRITE = 2
def __init__(self, fp):
object.__setattr__(self, r'_fp', fp)
object.__setattr__(self, r'_lastop', 0)
def __enter__(self):
return self._fp.__enter__()
def __exit__(self, exc_type, exc_val, exc_tb):
self._fp.__exit__(exc_type, exc_val, exc_tb)
def __getattr__(self, name):
return getattr(self._fp, name)
def __setattr__(self, name, value):
return self._fp.__setattr__(name, value)
def _noopseek(self):
self._fp.seek(0, os.SEEK_CUR)
def seek(self, *args, **kwargs):
object.__setattr__(self, r'_lastop', self.OPNONE)
return self._fp.seek(*args, **kwargs)
def write(self, d):
if self._lastop == self.OPREAD:
self._noopseek()
object.__setattr__(self, r'_lastop', self.OPWRITE)
return self._fp.write(d)
def writelines(self, *args, **kwargs):
if self._lastop == self.OPREAD:
self._noopeseek()
object.__setattr__(self, r'_lastop', self.OPWRITE)
return self._fp.writelines(*args, **kwargs)
def read(self, *args, **kwargs):
if self._lastop == self.OPWRITE:
self._noopseek()
object.__setattr__(self, r'_lastop', self.OPREAD)
return self._fp.read(*args, **kwargs)
def readline(self, *args, **kwargs):
if self._lastop == self.OPWRITE:
self._noopseek()
object.__setattr__(self, r'_lastop', self.OPREAD)
return self._fp.readline(*args, **kwargs)
def readlines(self, *args, **kwargs):
if self._lastop == self.OPWRITE:
self._noopseek()
object.__setattr__(self, r'_lastop', self.OPREAD)
return self._fp.readlines(*args, **kwargs)
def posixfile(name, mode='r', buffering=-1):
'''Open a file with even more POSIX-like semantics'''
try:
fp = osutil.posixfile(name, mode, buffering) # may raise WindowsError
# The position when opening in append mode is implementation defined, so
# make it consistent with other platforms, which position at EOF.
if 'a' in mode:
fp.seek(0, os.SEEK_END)
if '+' in mode:
return mixedfilemodewrapper(fp)
return fp
except WindowsError as err:
# convert to a friendlier exception
raise IOError(err.errno, '%s: %s' % (
name, encoding.strtolocal(err.strerror)))
# may be wrapped by win32mbcs extension
listdir = osutil.listdir
class winstdout(object):
'''stdout on windows misbehaves if sent through a pipe'''
def __init__(self, fp):
self.fp = fp
def __getattr__(self, key):
return getattr(self.fp, key)
def close(self):
try:
self.fp.close()
except IOError:
pass
def write(self, s):
try:
# This is workaround for "Not enough space" error on
# writing large size of data to console.
limit = 16000
l = len(s)
start = 0
self.softspace = 0
while start < l:
end = start + limit
self.fp.write(s[start:end])
start = end
except IOError as inst:
if inst.errno != 0:
raise
self.close()
raise IOError(errno.EPIPE, 'Broken pipe')
def flush(self):
try:
return self.fp.flush()
except IOError as inst:
if inst.errno != errno.EINVAL:
raise
raise IOError(errno.EPIPE, 'Broken pipe')
def _is_win_9x():
'''return true if run on windows 95, 98 or me.'''
try:
return sys.getwindowsversion()[3] == 1
except AttributeError:
return 'command' in encoding.environ.get('comspec', '')
def openhardlinks():
return not _is_win_9x()
def parsepatchoutput(output_line):
"""parses the output produced by patch and returns the filename"""
pf = output_line[14:]
if pf[0] == '`':
pf = pf[1:-1] # Remove the quotes
return pf
def sshargs(sshcmd, host, user, port):
'''Build argument list for ssh or Plink'''
pflag = 'plink' in sshcmd.lower() and '-P' or '-p'
args = user and ("%s@%s" % (user, host)) or host
if args.startswith('-') or args.startswith('/'):
raise error.Abort(
_('illegal ssh hostname or username starting with - or /: %s') %
args)
args = shellquote(args)
if port:
args = '%s %s %s' % (pflag, shellquote(port), args)
return args
def setflags(f, l, x):
pass
def copymode(src, dst, mode=None):
pass
def checkexec(path):
return False
def checklink(path):
return False
def setbinary(fd):
# When run without console, pipes may expose invalid
# fileno(), usually set to -1.
fno = getattr(fd, 'fileno', None)
if fno is not None and fno() >= 0:
msvcrt.setmode(fno(), os.O_BINARY)
def pconvert(path):
return path.replace(pycompat.ossep, '/')
def localpath(path):
return path.replace('/', '\\')
def normpath(path):
return pconvert(os.path.normpath(path))
def normcase(path):
return encoding.upper(path) # NTFS compares via upper()
# see posix.py for definitions
normcasespec = encoding.normcasespecs.upper
normcasefallback = encoding.upperfallback
def samestat(s1, s2):
return False
# A sequence of backslashes is special iff it precedes a double quote:
# - if there's an even number of backslashes, the double quote is not
# quoted (i.e. it ends the quoted region)
# - if there's an odd number of backslashes, the double quote is quoted
# - in both cases, every pair of backslashes is unquoted into a single
# backslash
# (See http://msdn2.microsoft.com/en-us/library/a1y7w461.aspx )
# So, to quote a string, we must surround it in double quotes, double
# the number of backslashes that precede double quotes and add another
# backslash before every double quote (being careful with the double
# quote we've appended to the end)
_quotere = None
_needsshellquote = None
def shellquote(s):
r"""
>>> shellquote(br'C:\Users\xyz')
'"C:\\Users\\xyz"'
>>> shellquote(br'C:\Users\xyz/mixed')
'"C:\\Users\\xyz/mixed"'
>>> # Would be safe not to quote too, since it is all double backslashes
>>> shellquote(br'C:\\Users\\xyz')
'"C:\\\\Users\\\\xyz"'
>>> # But this must be quoted
>>> shellquote(br'C:\\Users\\xyz/abc')
'"C:\\\\Users\\\\xyz/abc"'
"""
global _quotere
if _quotere is None:
_quotere = re.compile(r'(\\*)("|\\$)')
global _needsshellquote
if _needsshellquote is None:
# ":" is also treated as "safe character", because it is used as a part
# of path name on Windows. "\" is also part of a path name, but isn't
# safe because shlex.split() (kind of) treats it as an escape char and
# drops it. It will leave the next character, even if it is another
# "\".
_needsshellquote = re.compile(r'[^a-zA-Z0-9._:/-]').search
if s and not _needsshellquote(s) and not _quotere.search(s):
# "s" shouldn't have to be quoted
return s
return '"%s"' % _quotere.sub(r'\1\1\\\2', s)
def _unquote(s):
if s.startswith(b'"') and s.endswith(b'"'):
return s[1:-1]
return s
def shellsplit(s):
"""Parse a command string in cmd.exe way (best-effort)"""
return pycompat.maplist(_unquote, pycompat.shlexsplit(s, posix=False))
def quotecommand(cmd):
"""Build a command string suitable for os.popen* calls."""
if sys.version_info < (2, 7, 1):
# Python versions since 2.7.1 do this extra quoting themselves
return '"' + cmd + '"'
return cmd
def popen(command, mode='r'):
# Work around "popen spawned process may not write to stdout
# under windows"
# http://bugs.python.org/issue1366
command += " 2> %s" % pycompat.bytestr(os.devnull)
return os.popen(quotecommand(command), mode)
def explainexit(code):
return _("exited with status %d") % code, code
# if you change this stub into a real check, please try to implement the
# username and groupname functions above, too.
def isowner(st):
return True
def findexe(command):
'''Find executable for command searching like cmd.exe does.
If command is a basename then PATH is searched for command.
PATH isn't searched if command is an absolute or relative path.
An extension from PATHEXT is found and added if not present.
If command isn't found None is returned.'''
pathext = encoding.environ.get('PATHEXT', '.COM;.EXE;.BAT;.CMD')
pathexts = [ext for ext in pathext.lower().split(pycompat.ospathsep)]
if os.path.splitext(command)[1].lower() in pathexts:
pathexts = ['']
def findexisting(pathcommand):
'Will append extension (if needed) and return existing file'
for ext in pathexts:
executable = pathcommand + ext
if os.path.exists(executable):
return executable
return None
if pycompat.ossep in command:
return findexisting(command)
for path in encoding.environ.get('PATH', '').split(pycompat.ospathsep):
executable = findexisting(os.path.join(path, command))
if executable is not None:
return executable
return findexisting(os.path.expanduser(os.path.expandvars(command)))
_wantedkinds = {stat.S_IFREG, stat.S_IFLNK}
def statfiles(files):
'''Stat each file in files. Yield each stat, or None if a file
does not exist or has a type we don't care about.
Cluster and cache stat per directory to minimize number of OS stat calls.'''
dircache = {} # dirname -> filename -> status | None if file does not exist
getkind = stat.S_IFMT
for nf in files:
nf = normcase(nf)
dir, base = os.path.split(nf)
if not dir:
dir = '.'
cache = dircache.get(dir, None)
if cache is None:
try:
dmap = dict([(normcase(n), s)
for n, k, s in listdir(dir, True)
if getkind(s.st_mode) in _wantedkinds])
except OSError as err:
# Python >= 2.5 returns ENOENT and adds winerror field
# EINVAL is raised if dir is not a directory.
if err.errno not in (errno.ENOENT, errno.EINVAL,
errno.ENOTDIR):
raise
dmap = {}
cache = dircache.setdefault(dir, dmap)
yield cache.get(base, None)
def username(uid=None):
"""Return the name of the user with the given uid.
If uid is None, return the name of the current user."""
return None
def groupname(gid=None):
"""Return the name of the group with the given gid.
If gid is None, return the name of the current group."""
return None
def removedirs(name):
"""special version of os.removedirs that does not remove symlinked
directories or junction points if they actually contain files"""
if listdir(name):
return
os.rmdir(name)
head, tail = os.path.split(name)
if not tail:
head, tail = os.path.split(head)
while head and tail:
try:
if listdir(head):
return
os.rmdir(head)
except (ValueError, OSError):
break
head, tail = os.path.split(head)
def rename(src, dst):
'''atomically rename file src to dst, replacing dst if it exists'''
try:
os.rename(src, dst)
except OSError as e:
if e.errno != errno.EEXIST:
raise
unlink(dst)
os.rename(src, dst)
def gethgcmd():
return [sys.executable] + sys.argv[:1]
def groupmembers(name):
# Don't support groups on Windows for now
raise KeyError
def isexec(f):
return False
class cachestat(object):
def __init__(self, path):
pass
def cacheable(self):
return False
def lookupreg(key, valname=None, scope=None):
''' Look up a key/value name in the Windows registry.
valname: value name. If unspecified, the default value for the key
is used.
scope: optionally specify scope for registry lookup, this can be
a sequence of scopes to look up in order. Default (CURRENT_USER,
LOCAL_MACHINE).
'''
if scope is None:
scope = (winreg.HKEY_CURRENT_USER, winreg.HKEY_LOCAL_MACHINE)
elif not isinstance(scope, (list, tuple)):
scope = (scope,)
for s in scope:
try:
val = winreg.QueryValueEx(winreg.OpenKey(s, key), valname)[0]
# never let a Unicode string escape into the wild
return encoding.unitolocal(val)
except EnvironmentError:
pass
expandglobs = True
def statislink(st):
'''check whether a stat result is a symlink'''
return False
def statisexec(st):
'''check whether a stat result is an executable file'''
return False
def poll(fds):
# see posix.py for description
raise NotImplementedError()
def readpipe(pipe):
"""Read all available data from a pipe."""
chunks = []
while True:
size = win32.peekpipe(pipe)
if not size:
break
s = pipe.read(size)
if not s:
break
chunks.append(s)
return ''.join(chunks)
def bindunixsocket(sock, path):
raise NotImplementedError('unsupported platform')