http://www.python.org/dev/buildbot/builders/x86 FreeBSD 3.x/builds/211/steps/test/logs/stdio

Example:
======================================================================
FAIL: test_send_signal (test.test_subprocess.POSIXProcessTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/home/db3l/buildarea/3.x.bolen-freebsd/build/Lib/test/test_subprocess.py", line 770, in test_send_signal
    self.assertIn(b'KeyboardInterrupt', stderr)
AssertionError: b'KeyboardInterrupt' not found in b''

----------------------------------------------------------------------

See also issue #8263.

I get this same failure on Mac OS X 10.6.3 as well.

Since it only happens on 3.x, a possible culprit is “[svn r78946] Replaces the internals of the subprocess module from fork through exec on
POSIX systems with a C extension module”.

NB: recent URL of deterministic failure under FreeBSD 6.4:
http://www.python.org/dev/buildbot/builders/x86%20FreeBSD%203.x/builds/492

http://www.python.org/dev/buildbot/builders/x86%20FreeBSD%203.x/builds/492

On this report, both implementations (Python and C) failed:

FAIL: test_send_signal (test.test_subprocess.POSIXProcessTestCase)
FAIL: test_send_signal (test.test_subprocess.ProcessTestCasePOSIXPurePython)

--

The problem is maybe that the SIGINT is ignored on these particular platforms (FreeBSD and Mac OS X). I worked on this topic recently (see issue #3137) and I cannot guarantee that exceptions are still ignored somewhere. I can investigate if you give me access to the platform.

My procedure was to run python in gdb, set a break point, and send the signal at different points until the KeyboardInterrupt is no more ignored anywhere.

> The problem is maybe that the SIGINT is ignored on these particular
> platforms (FreeBSD and Mac OS X). I worked on this topic recently (see
> issue #3137) and I cannot guarantee that exceptions are still ignored
> somewhere. I can investigate if you give me access to the platform.

These tests exist and pass on 2.7 and 3.1.

> These tests exist and pass on 2.7 and 3.1.

Ok. In r78946, for the pure Python implementation, I only see the restoration of the signals which can be related to this issue.

Issue 8714 could be related as well: On OpenBSD, after a KeyboardInterrupt
one has to hit <Return> before the traceback appears.

This was introduced in (or exposed by) r68460. When Python is compiled
without threads, the bug does not occur.

Also occurs on Python 2.7, with "x86 FreeBSD 7.2" buildbot.
(I found it on many builds since revision 83725 at least. Older builds are not available)
Seen on "x86 FreeBSD" buildbot, too.


test test_subprocess failed -- Traceback (most recent call last):
  File "/usr/home/db3l/buildarea/2.7.bolen-freebsd7/build/Lib/test/test_subprocess.py", line 723, in test_send_signal
    self.assertIn('KeyboardInterrupt', stderr)
AssertionError: 'KeyboardInterrupt' not found in ''

http://www.python.org/dev/buildbot/builders/x86%20FreeBSD%207.2%202.7/builds/89
http://www.python.org/dev/buildbot/builders/x86%20FreeBSD%207.2%202.7/builds/51

The test fails mostly on very slow machines (not only BSD ones, but also the Debian sparc buildbot).
Therefore, I've attempted a fix in r84910 by relaxing timings, and simply skipping the test if it seems we couldn't send the signal in time.

The fix succeeded on the "sparc Debian 3.x" buildbot while test_subprocess had been failing consistently before.

test_subprocess now succeeds on all 3.x buildbots. Will backport and then close.