Bug #85421 for Module-CPANTS-Analyse: (Hard to repeat) failure when running multiple analysers simultaneously under HARNESS

Sat May 18 16:15:17 2013 KENTNL [...] cpan.org - Ticket created

Subject:

(Hard to repeat) failure when running under HARNESS_OPTIONS=j5

I got this error on one pass, but its hard to reproduce, because it appears to be some sort of race condition between 2 tests trying to both read/write the same file, or something like that.

The fail is kinda lenghty, so I'll put it in its own reply so RT doesn't get all "helpful" and hide the entire bug report due to being "too long".

Sat May 18 16:15:46 2013 KENTNL [...] cpan.org - Correspondence added

Message body is not shown because it is too large.

Sat May 18 16:18:04 2013 KENTNL [...] cpan.org - Correspondence added

RT, what the fuck. What the actual fuck. Why am I getting HTML when I ask to view the message.

Helpful.

Sat May 18 16:26:47 2013 KENTNL [...] cpan.org - Correspondence added

On 2013-05-19 08:18:04, KENTNL wrote:
Show quoted text

> RT, what the fuck. What the actual fuck. Why am I getting HTML when I ask to
> view the message.
>
> Helpful.

Ok, Have a way to repeat it reliably now:

make clean ; rm Debian_CPANTS.txt ; rm t/corpus/Debian_CPANTS.txt ; perl Makefile.PL ; make ; HARNESS_OPTIONS=j5 make test

The reason its hard to repeat is t/corpus/Debian_CPANTS.txt is provisioned in the first pass, and the problem is r/w with that file, so if you simply repeat the tests it will pass on successive runs.

Sat May 18 16:48:38 2013 ether [...] cpan.org - Subject changed from '(Hard to repeat) failure when running under HARNESS_OPTIONS=j5' to '(Hard to repeat) failure when running multiple analysers simultaneously under HARNESS_OPTIONS=j5'

Sat May 18 16:48:38 2013 ether [...] cpan.org - Queue changed from Test-Kwalitee to Module-CPANTS-Analyse

Sat May 18 16:48:38 2013 ether [...] cpan.org - Severity Normal added

Sat May 18 16:48:38 2013 ether [...] cpan.org - Broken in 1.06 deleted

Sat May 18 16:49:43 2013 ether [...] cpan.org - Correspondence added

On 2013-05-18 13:26:47, KENTNL wrote: Show quoted text

> Ok, Have a way to repeat it reliably now: > > make clean ; rm Debian_CPANTS.txt ; rm t/corpus/Debian_CPANTS.txt ; > perl > Makefile.PL ; make ; HARNESS_OPTIONS=j5 make test > > The reason its hard to repeat is t/corpus/Debian_CPANTS.txt is > provisioned in > the first pass, and the problem is r/w with that file, so if you > simply repeat > the tests it will pass on successive runs.

This sounds like a problem deep in Module::CPANTS::Analyse (Test::Kwalitee doesn't do much other than run those indicators), so I'm bumping this ticket over to them... The repro steps pertain to Test::Kwalitee 1.06, which has several tests that run kwalitee metrics in succession.

Sat May 18 16:49:43 2013 The RT System itself - Status changed from 'new' to 'open'

Sat May 18 16:53:52 2013 ether [...] cpan.org - Correspondence added

13:52 < kentnl> even a test that just does bail_out if $ENV{HARNESS_OPTIONS} =~ /j(\d+)/ and $1 ne 1 ; that would be great 13:53 <@ether> the first thing that comes to mind is that each call to the analysers should use its own randomly-named temp file, rather than using one -- so simultaneous runs don't stomp all over each other

Fri Jul 26 11:18:00 2013 ether [...] cpan.org - Correspondence added

Also reported in https://github.com/harleypig/Dist-Zilla-Plugin-Test-Kwalitee/issues/3

Sun Jun 29 11:25:34 2014 ishigaki [...] cpan.org - Correspondence added

Closing this as it looks like this has been resolved since 0.88/0.91 when the problematic debian metrics were removed. Thanks.

Sun Jun 29 11:25:37 2014 ishigaki [...] cpan.org - Status changed from 'open' to 'resolved'

Bug #85421 for Module-CPANTS-Analyse: (Hard to repeat) failure when running multiple analysers simultaneously under HARNESS_OPTIONS=j5