The regrtest returns "success", even if the test suite is not run completely.
The last running test is "test_unittest".


http://www.python.org/dev/buildbot/all/builders/x86%20FreeBSD%20trunk/builds/171
...
51 tests OK.
2 tests skipped:
    test_ascii_formatd test_py3kwarn
1 skip unexpected on freebsd6:
    test_ascii_formatd
[294584 refs]
program finished with exit code 0
elapsedTime=1082.949696

http://www.python.org/dev/buildbot/all/builders/x86%20FreeBSD%20trunk/builds/168
...
37 tests OK.
5 tests skipped:
    test_bsddb185 test_macostools test_py3kwarn test_startfile
    test_winreg
1 skip unexpected on freebsd6:
    test_bsddb185
[184731 refs]
program finished with exit code 0
elapsedTime=507.756584

I wonder if unittest.test.test_break runs and the signals.SIGINT it sends terminates a process early? Do you have access to this platform to try it Florent?

Sorry, but I don't have such environment.

Proposed patch will add a notice if the regrtest is interrupted (Ctrl+C).
And the exitcode is set to 1, meaning test failed for the buildbot.

That sounds very useful from a diagnostic point of view. If it does turn out to be the problem then we will have to disable the test for that platform. At the moment test_break runs on all platforms that have os.kill(...).

Slightly different, with a count of omitted tests
And removed "All .. tests OK" is there's some tests omitted.

Regrtest.py fixed with r79497 to report keyboard interrupt as a "failure".

I can reliably reproduce the SIGINT problem on a FreeBSD 6.3 box, if (and so far only if) some test that uses threading is run first (see also issue 3864).  So I suspect this is a freebsd6 specific problem and just needs a skip added.

The test_break unit test is now skipped for freebsd6 on trunk in r80155.  I suspect this affects py3k as well on freebsd6.

See also issue #8432.

Confirmed on FreeBSD-6.4/py3k. Threading causes the problem, the skip
works here, too.

Skip committed to py3k in r80331.