develooper Front page | perl.perl5.porters | Postings from July 2009

Re: GitLive-maint-5.10-1655-g4197df1 breaks 100s of modules

Thread Previous | Thread Next
Craig A. Berry
July 29, 2009 09:10
Re: GitLive-maint-5.10-1655-g4197df1 breaks 100s of modules
Message ID:
On Wed, Jul 29, 2009 at 4:24 AM, demerphq<> wrote:
> 2009/7/28 David Favor <>:
>> As an example here's a compile of PPI which works
>> and one which fails.
>>    perl-5.10.1-1641
>>    perl-5.10.1-1655
> That was a doc patch. Can you try your test again from a few commits earlier?

Thanks for the report, David, and thanks for testing.  After climbing
back into the chair from which I'd fallen and looking up what PPI is,
I realized that nothing "breaks 100s of modules."  Something breaks
one module, PPI, which now apparently has pretty basic difficulties
doing its job of parsing other modules.

As Yves says, narrowing this down further would help (maybe use git
bisect?).  You might also look into why on your run with failures, the
manifest checker finds cache files from a previous run:

Not in MANIFEST: t/data/18_cache/6/64/64568092e7faba16d99fa04706c46517.ppi
Not in MANIFEST: t/data/18_cache/a/ab/abcdef1234567890abcdef1234567890.ppi

Does clearing those out manually make any difference in the test
failures?  Even if it doesn't, the clean targets in PPI may need a
little attention.

The tests that fail spew a lot of warnings that weren't there before.
Can you follow the clues they give about where the trouble might be
and create a smaller reproducer?

I'm cc'ing Adam Kennedy, who appears to be the PPI maintainer, in case
he has any advice or insight.

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About