# perl-dev mailing list archives

##### Site index · List index
Message view
Top
From Markus Wichitill <ma...@gmx.de>
Subject Re: time for 1.99_15?
Date Thu, 26 Aug 2004 22:38:13 GMT
Stas Bekman wrote:
>> Markus Wichitill wrote:
>>> prematurely: SV 0x6cc4a14, Perl interpreter: 0x6c90cfc at
>>> request has failed (the response code was: 500)
>>>
>>> Seems to have crashed Apache, as all later tests failed (but I forgot
>>> to look at the process list to check if it was really gone). No crash
>>> popup this time, so no JIT debugging. Didn't see anything in the
>>> error_log. Can't reproduce it anymore.
>>
>> I now always get this. Apache exits, but doesn't crash in a way that
>> would allow me to debug it. If I attach the debugger during the "nmake
>> test" execution, all I see is:
>>
>> [...]
>> The thread 0x3BC has exited with code 9 (0x9).
>> The thread 0xEC has exited with code 9 (0x9).
>> The thread 0x5FC has exited with code 9 (0x9).
>> The program 'C:\Dev\debug\apache\bin\apache.exe' has exited with code
>> 9 (0x9).
>
> If you can reliably get this problem, we need to figure what tests
> sequence causes it, before we can even start solving it. You could try
> to do that manually (with a binary search). Or if you can get hold of
> some error messages in the error_log you could adjust TestSmoke.pm to
> check those after runnin each test. It already checks for a core file
> after each test is run and considers the test to be failure if it finds
> one. So you could add another hook, and check error_log (for which it
> already holds the pointer in the filehandle, which is moved after each
> test) for custom error messages.

Manual search would be cruel, there's nothing in the log, and there's no
coredumps on Win32 - but since this, unlike the leak issue, is a real
failure, I've run

perl t\SMOKE -order=rotate -times=1 -verbose

which yields the reduced sequence

perl t\TEST -verbose t\filter\in_bbs_msg.t t\filter\in_bbs_underrun.t