Bug#476011: libpoe-api-peek-perl: FTBFS: Failed 1/9 test scripts, 88.89% okay. 2/102 subtests failed, 98.04% okay.

[Niko Tyni]

clone 476011 -1
reassign -1 libpoe-perl 1.0000-1
retitle -1 libpoe-perl: hiding KR_SESSIONS breaks libpoe-api-peek-perl
tag -1 upstream patch
notforwarded -1
block 476011 with -1
thanks

On Wed, Apr 16, 2008 at 10:52:34PM +0300, Niko Tyni wrote:

> > > > > t/07-kernel........
> > > > > #   Failed test 'event_list() returns populated hashref'
> > > > > #   in t/07-kernel.t at line 54.
> > > > > 
> > > > > #   Failed test 'event_list() returns correct list of sessions and events'
> > > > > #   in t/07-kernel.t at line 58.
> > > > > #     Structures begin differing at:
> > > > > #          $got->{2} = Does not exist
> > > > > #     $expected->{2} = ARRAY(0x849a2d8)
> > > > > # Looks like you failed 2 tests of 11.
> 
> > upstream test reports indicate it's caused by POE 1.0000.
> 
> Verified this and reported upstream as CPAN #35129. No fix yet,
> and I'm not even sure which package is doing the wrong thing.

OK, the regression is caused by this change in libpoe-perl 1.0000:

 http://poe.svn.sourceforge.net/viewvc/poe?view=rev&revision=2245

 svn diff -r2244:2245 https://poe.svn.sourceforge.net/svnroot/poe/trunk

which removes the initialization of
$POE::Kernel::poe_kernel->[KR_SESSIONS] (and KR_FILENOS, but that's not
the issue here) that Poe::API::Peek::event_list() needs.

I don't know if this is intentional hiding or a bug. We should ask
upstream. POE::Resource::Sessions (where KR_SESSIONS lives) is documented
to be an internal module without any public interface.

A quick patch attached; I assume upstream intends to remove those
_preload() functions instead of calling them but this pinpoints the problem.

I have verified that the patch fixes libpoe-api-peek-perl while not
breaking the libpoe-perl test suite itself.

Will notify both upstreams probably tomorrow.

Cheers,
-- 
Niko Tyni   ntyni at debian.org