OSError: [Errno 3] No such process

[ormandi.matyas@…]

How to Reproduce

While doing a GET operation on /browser, Trac issued an internal error.

(please provide additional details here)

Request parameters:

{'path': '/'}

User agent: Chrome/23.0.1271.64

System Information

Trac	1.0
Genshi	0.6 (without speedups)
GIT	1.7.9
pysqlite	2.4.1
Python	2.6.8 (unknown, Jun 9 2012, 11:30:32) [GCC 4.5.3]
setuptools	0.6c11
SQLite	3.7.13
jQuery	1.7.2

Enabled Plugins

Python Traceback

Traceback (most recent call last):
  File "build/bdist.cygwin-1.7.17-i686/egg/trac/web/main.py", line 497, in _dispatch_request
    dispatcher.dispatch(req)
  File "build/bdist.cygwin-1.7.17-i686/egg/trac/web/main.py", line 214, in dispatch
    resp = chosen_handler.process_request(req)
  File "build/bdist.cygwin-1.7.17-i686/egg/trac/versioncontrol/web_ui/browser.py", line 391, in process_request
    dir_data = self._render_dir(req, repos, node, rev, order, desc)
  File "build/bdist.cygwin-1.7.17-i686/egg/trac/versioncontrol/web_ui/browser.py", line 559, in _render_dir
    entries = [entry(n) for n in node.get_entries()
  File "build/bdist.cygwin-1.7.17-i686/egg/tracopt/versioncontrol/git/git_fs.py", line 585, in get_entries
    historian)
  File "/usr/lib/python2.6/contextlib.py", line 23, in __exit__
    self.gen.next()
  File "build/bdist.cygwin-1.7.17-i686/egg/tracopt/versioncontrol/git/PyGIT.py", line 930, in get_historian
    terminate(p[0])
  File "build/bdist.cygwin-1.7.17-i686/egg/tracopt/versioncontrol/git/PyGIT.py", line 58, in terminate
    return terminate_nix(process)
  File "build/bdist.cygwin-1.7.17-i686/egg/tracopt/versioncontrol/git/PyGIT.py", line 54, in terminate_nix
    return os.kill(process.pid, signal.SIGTERM)
OSError: [Errno 3] No such process

[Christian Boos]

Well, we don't really support Trac on Windows with Cygwin's Python. You're on your own with that setup (PatchWelcome, but make sure the changes don't break the regular Unix platforms).

[Jun Omae]

repos:jomae.git:ticket10958, 5d27bd6d/jomae.git ignores a OSError of ESRCH.

On POSIX platforms, it is able to send signals to the terminated process which has not been waited (zombie process). However, cannot on cygwin.

Linux

$ python
Python 2.4.3 (#1, Jan  9 2013, 06:49:54)
[GCC 4.1.2 20080704 (Red Hat 4.1.2-54)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from subprocess import Popen, PIPE
>>> import signal, os
>>> def handler(sig, frame):
...   print 'Caught SIGCHLD\n'
...
>>> signal.signal(signal.SIGCHLD, handler)
0
>>> proc = Popen(('/bin/cat',), stdin=PIPE, stdout=PIPE)
>>> proc.stdin.close()
>>> Caught SIGCHLD


>>> os.kill(proc.pid, signal.SIGTERM)  # Be able to send signals until wait
>>> os.kill(proc.pid, signal.SIGTERM)
>>> os.kill(proc.pid, signal.SIGTERM)
>>> proc.wait()
0
>>> os.kill(proc.pid, signal.SIGTERM)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
OSError: [Errno 3] No such process

Cygwin

$ /usr/bin/python
Python 2.7.3 (default, Dec 18 2012, 13:50:09)
[GCC 4.5.3] on cygwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from subprocess import Popen, PIPE
>>> import signal, os
>>> def handler(sig, frame):
...   print 'Caught SIGCHLD\n'
...
>>> signal.signal(signal.SIGCHLD, handler)
0
>>> proc = Popen(('/bin/cat',), stdin=PIPE, stdout=PIPE)
>>> proc.stdin.close()
>>>
Caught SIGCHLD

>>> os.kill(proc.pid, signal.SIGTERM)  # Have yet to wait, however, cannot send any signals
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OSError: [Errno 3] No such process
>>> proc.wait()
0

[Christian Boos]

So OK, this ignores the exception, but what happens with the process, will it be left dangling? Wouldn't we better off using terminate_win for that, assuming we can get the appropriate Windows pid from the cygwin pid (​cygwin_winpid_to_pid with ctypes? Well, you're the ctypes hacker ;-) ).

[Christian Boos]

Replying to cboos:

(​cygwin_winpid_to_pid …

Sorry, got it backwards. Googled a bit more: this one sounds promising:

winpid = cygwin_internal(CW_CYGWIN_PID_TO_WINPID, cygpid);

(from ​so:1679337)

[Jun Omae]

Replying to cboos:

So OK, this ignores the exception, but what happens with the process, will it be left dangling?

No, it will be not left dangling, I think. The process is fork/exec'ed from cygwin python. I think that os.kill() correctly works.

• ​sig_send in cygwin/sigproc.cc
• ​_pinfo::kill in cygwin/signal.cc

>>> proc = Popen(('/bin/cat',), stdin=PIPE, stdout=PIPE)
>>> os.kill(proc.pid, signal.SIGTERM)  # the child is still alive
>>> Caught SIGCHLD                     # ... it terminates

>>> proc.returncode
>>> proc.wait()
-15
>>> proc.returncode
-15
>>> os.kill(proc.pid, signal.SIGTERM)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OSError: [Errno 3] No such process

Also, I just tried to call ctypes.windll.kernel32.TerminateProcess when cygwin. cygwin_internal works well, however, Python on cygwin doesn't provide ctypes.windll.

[Christian Boos]

Replying to jomae:

Replying to cboos:

So OK, this ignores the exception, but what happens with the process, will it be left dangling?

No, it will be not left dangling, I think. The process is fork/exec'ed from cygwin python. I think that os.kill() correctly works.

I see. Then I suggest to rephrase the comment like this:

            os.kill(process.pid, signal.SIGTERM) 
        except OSError, e: 
            # If the process has already finished and has not been
            # waited for, killing it raises an ESRCH error on Cygwin

[Jun Omae]

Thanks. My patch with your suggestions is committed in [11710-11711].