diff mercurial/debugcommands.py @ 36528:72e487851a53

debugcommands: add debugwireproto command We currently don't have a low-level mechanism for sending arbitrary wire protocol commands. Having a generic and robust mechanism for sending wire protocol commands, examining wire data, etc would make it vastly easier to test the wire protocol and debug server operation. This is a problem I've wanted a solution for numerous times, especially recently as I've been hacking on a new version of the wire protocol. This commit establishes a `hg debugwireproto` command for sending data to a peer. The command invents a mini language for specifying actions to take. This will enable a lot of flexibility for issuing commands and testing variations for how commands are sent. Right now, we only support low-level raw sends and receives. These are probably the least valuable commands to intended users of this command. But they are the most useful commands to implement to bootstrap the feature (I've chosen to reimplement test-ssh-proto.t using this command to prove its usefulness). My eventual goal of `hg debugwireproto` is to allow calling wire protocol commands with a human-friendly interface. Essentially, people can type in a command name and arguments and `hg debugwireproto` will figure out how to send that on the wire. I'd love to eventually be able to save the server's raw response to a file. This would allow us to e.g. call "getbundle" wire protocol commands easily. test-ssh-proto.t has been updated to use the new command in lieu of piping directly to a server process. As part of the transition, test behavior improved. Before, we piped all request data to the server at once. Now, we have explicit control over the ordering of operations. e.g. we can send one command, receive its response, then send another command. This will allow us to more robustly test race conditions, buffering behavior, etc. There were some subtle changes in test behavior. For example, previous behavior would often send trailing newlines to the server. The new mechanism doesn't treat literal newlines specially and requires newlines be escaped in the payload. Because the new logging code is very low level, it is easy to introduce race conditions in tests. For example, the number of bytes returned by a read() may vary depending on load. This is why tests make heavy use of "readline" for consuming data: the result of that operation should be deterministic and not subject to race conditions. There are still some uses of "readavailable." However, those are only for reading from stderr. I was able to reproduce timing issues with my system under load when using "readavailable" globally. But if I "readline" to grab stdout, "readavailable" appears to work deterministically for stderr. I think this is because the server writes to stderr first. As long as the OS delivers writes to pipes in the same order they were made, this should work. If there are timing issues, we can introduce a mechanism to readline from stderr. Differential Revision: https://phab.mercurial-scm.org/D2392
author Gregory Szorc <gregory.szorc@gmail.com>
date Thu, 01 Mar 2018 08:24:54 -0800
parents 44dc34b8d17b
children bde0bd50f368
line wrap: on
line diff
--- a/mercurial/debugcommands.py	Tue Feb 27 15:47:44 2018 -0800
+++ b/mercurial/debugcommands.py	Thu Mar 01 08:24:54 2018 -0800
@@ -17,6 +17,7 @@
 import socket
 import ssl
 import string
+import subprocess
 import sys
 import tempfile
 import time
@@ -65,6 +66,7 @@
     setdiscovery,
     simplemerge,
     smartset,
+    sshpeer,
     sslutil,
     streamclone,
     templater,
@@ -2529,3 +2531,204 @@
     ui.write("%s\n" % res1)
     if res1 != res2:
         ui.warn("%s\n" % res2)
+
+def _parsewirelangblocks(fh):
+    activeaction = None
+    blocklines = []
+
+    for line in fh:
+        line = line.rstrip()
+        if not line:
+            continue
+
+        if line.startswith(b'#'):
+            continue
+
+        if not line.startswith(' '):
+            # New block. Flush previous one.
+            if activeaction:
+                yield activeaction, blocklines
+
+            activeaction = line
+            blocklines = []
+            continue
+
+        # Else we start with an indent.
+
+        if not activeaction:
+            raise error.Abort(_('indented line outside of block'))
+
+        blocklines.append(line)
+
+    # Flush last block.
+    if activeaction:
+        yield activeaction, blocklines
+
+@command('debugwireproto',
+    [
+        ('', 'localssh', False, _('start an SSH server for this repo')),
+        ('', 'peer', '', _('construct a specific version of the peer')),
+    ] + cmdutil.remoteopts,
+    _('[REPO]'),
+    optionalrepo=True)
+def debugwireproto(ui, repo, **opts):
+    """send wire protocol commands to a server
+
+    This command can be used to issue wire protocol commands to remote
+    peers and to debug the raw data being exchanged.
+
+    ``--localssh`` will start an SSH server against the current repository
+    and connect to that. By default, the connection will perform a handshake
+    and establish an appropriate peer instance.
+
+    ``--peer`` can be used to bypass the handshake protocol and construct a
+    peer instance using the specified class type. Valid values are ``raw``,
+    ``ssh1``, and ``ssh2``. ``raw`` instances only allow sending raw data
+    payloads and don't support higher-level command actions.
+
+    Commands are issued via a mini language which is specified via stdin.
+    The language consists of individual actions to perform. An action is
+    defined by a block. A block is defined as a line with no leading
+    space followed by 0 or more lines with leading space. Blocks are
+    effectively a high-level command with additional metadata.
+
+    Lines beginning with ``#`` are ignored.
+
+    The following sections denote available actions.
+
+    raw
+    ---
+
+    Send raw data to the server.
+
+    The block payload contains the raw data to send as one atomic send
+    operation. The data may not actually be delivered in a single system
+    call: it depends on the abilities of the transport being used.
+
+    Each line in the block is de-indented and concatenated. Then, that
+    value is evaluated as a Python b'' literal. This allows the use of
+    backslash escaping, etc.
+
+    raw+
+    ----
+
+    Behaves like ``raw`` except flushes output afterwards.
+
+    close
+    -----
+
+    Close the connection to the server.
+
+    flush
+    -----
+
+    Flush data written to the server.
+
+    readavailable
+    -------------
+
+    Read all available data from the server.
+
+    If the connection to the server encompasses multiple pipes, we poll both
+    pipes and read available data.
+
+    readline
+    --------
+
+    Read a line of output from the server. If there are multiple output
+    pipes, reads only the main pipe.
+    """
+    opts = pycompat.byteskwargs(opts)
+
+    if opts['localssh'] and not repo:
+        raise error.Abort(_('--localssh requires a repository'))
+
+    if opts['peer'] and opts['peer'] not in ('raw', 'ssh1', 'ssh2'):
+        raise error.Abort(_('invalid value for --peer'),
+                          hint=_('valid values are "raw", "ssh1", and "ssh2"'))
+
+    if ui.interactive():
+        ui.write(_('(waiting for commands on stdin)\n'))
+
+    blocks = list(_parsewirelangblocks(ui.fin))
+
+    proc = None
+
+    if opts['localssh']:
+        # We start the SSH server in its own process so there is process
+        # separation. This prevents a whole class of potential bugs around
+        # shared state from interfering with server operation.
+        args = util.hgcmd() + [
+            '-R', repo.root,
+            'debugserve', '--sshstdio',
+        ]
+        proc = subprocess.Popen(args, stdin=subprocess.PIPE,
+                                stdout=subprocess.PIPE, stderr=subprocess.PIPE,
+                                bufsize=0)
+
+        stdin = proc.stdin
+        stdout = proc.stdout
+        stderr = proc.stderr
+
+        # We turn the pipes into observers so we can log I/O.
+        if ui.verbose or opts['peer'] == 'raw':
+            stdin = util.makeloggingfileobject(ui, proc.stdin, b'i',
+                                               logdata=True)
+            stdout = util.makeloggingfileobject(ui, proc.stdout, b'o',
+                                                logdata=True)
+            stderr = util.makeloggingfileobject(ui, proc.stderr, b'e',
+                                                logdata=True)
+
+        # --localssh also implies the peer connection settings.
+
+        url = 'ssh://localserver'
+
+        if opts['peer'] == 'ssh1':
+            ui.write(_('creating ssh peer for wire protocol version 1\n'))
+            peer = sshpeer.sshv1peer(ui, url, proc, stdin, stdout, stderr,
+                                     None)
+        elif opts['peer'] == 'ssh2':
+            ui.write(_('creating ssh peer for wire protocol version 2\n'))
+            peer = sshpeer.sshv2peer(ui, url, proc, stdin, stdout, stderr,
+                                     None)
+        elif opts['peer'] == 'raw':
+            ui.write(_('using raw connection to peer\n'))
+            peer = None
+        else:
+            ui.write(_('creating ssh peer from handshake results\n'))
+            peer = sshpeer.makepeer(ui, url, proc, stdin, stdout, stderr)
+
+    else:
+        raise error.Abort(_('only --localssh is currently supported'))
+
+    # Now perform actions based on the parsed wire language instructions.
+    for action, lines in blocks:
+        if action in ('raw', 'raw+'):
+            # Concatenate the data together.
+            data = ''.join(l.lstrip() for l in lines)
+            data = util.unescapestr(data)
+            stdin.write(data)
+
+            if action == 'raw+':
+                stdin.flush()
+        elif action == 'flush':
+            stdin.flush()
+        elif action == 'close':
+            peer.close()
+        elif action == 'readavailable':
+            fds = util.poll([stdout.fileno(), stderr.fileno()])
+
+            if stdout.fileno() in fds:
+                util.readpipe(stdout)
+            if stderr.fileno() in fds:
+                util.readpipe(stderr)
+        elif action == 'readline':
+            stdout.readline()
+        else:
+            raise error.Abort(_('unknown action: %s') % action)
+
+    if peer:
+        peer.close()
+
+    if proc:
+        proc.kill()