httpd-test-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stas Bekman <>
Subject Re: Failures in with A::T 1.09
Date Thu, 11 Mar 2004 18:01:17 GMT
William McKee wrote:
> On Wed, Mar 10, 2004 at 11:46:02PM -0800, Stas Bekman wrote:
>>>   optimize='-O3',
>>that could be your problem. Try rebuilding with -02 instead.
> Hi Stas,
> Well, I tried recompiling 5.8.2 with optimize=02. No go.
> I tried to compile 5.8.3 (with O3) - no good.
> I tried to compile 5.8.3 (with O2) - no good.

Are you sure it was actually compiled with -02? Did you check 'perl -V'?

> I'm just going to comment it out for now. At least I'm up to 5.8.3 now
> ;).

That would be a pain, won't it. Let's try to nail it down, continueing where 
we started:

 >>Since, I can't reproduce it, I'm not sure what is abnormal about your perl,
 >>I'd suggest to start truncating the code that fails, till you get to the
 >>smallest possible case which you will be able to reproduce in a standalone
 >>script. Then we can ask p5p if there is some sort of bug, triggered in
 >>certain cases.
 > I copied the entire list of modules and the set_ulimit_via_sh function
 > into a test script but was unable to reproduce the problem. Also tried
 > messing with eval{} to no avail.

No, that's not what I was suggesting. I suggested that you take the failing 
A-T as is, and start cutting chunks of the code leading to failure and 
retrying, untill you get a minimal case. This is much better than starting 
from the other side. Using that approach you have a constant failure, and you 
know that removing some code doesn't affect it.

 > I was able to find another poor soul who has experienced my pain[1].
 > Unfortunately there was no solution. Found several threads from back in
 > the mid-90's which addressed several problems before the die, warn and
 > carp statements were allowed to come after an exec w/o triggering the
 > warning message. Still not enough to help me reproduce the error when
 > running with TEST.

Once we have a standalone short test case, we can ask at p5p.

Stas Bekman            JAm_pH ------> Just Another mod_perl Hacker     mod_perl Guide --->

View raw message