DBIx::Class and minimum perl requirements (especially EPEL)

Paul Howarth paul at city-fan.org
Tue Oct 28 17:14:51 UTC 2008


Chris Weyl wrote:
> On Mon, Oct 27, 2008 at 3:05 PM, Nigel Metheringham
> <nigel.metheringham at dev.intechnology.co.uk> wrote:
>> Chris,
>>
>> Nice to meet you - and commiseration's on being responsible for packaging
>> DBIx::Class (having packaged it myself for our internal systems I know it is
>> a bitch of module to package with a tendancy to pull in all of CPAN as a
>> dep).
> 
> Nice to meet you as well -- DBIx::Class was definitely one of those
> packages that had a, well, very large dependency tree when I submitted
> it for review :)
> 
>> Minutes before seeing your message I committed some proposed changes to both
>> documentation and test suite to cover these issues.
>>
>> I would appreciate your comments on svn commits 5009 & 5010 to
>> http://dev.catalyst.perl.org/repos/bast/DBIx-Class/0.08/branches/doc_mods
>>
>> http://dev.catalyst.perl.org/svnweb/bast/revision?rev=5009
>> http://dev.catalyst.perl.org/svnweb/bast/revision?rev=5010
>>
>> I don't have access to my test RHEL5 VMs from here (home), so have not
>> tested the tests on an affected system, but it *should* work - will check it
>> tomorrow morning.
> 
> The diffs (documentation) looked good to me.  I pulled down
> t/99rh_perl_perf_bug.t (r5014) and tried it out on a couple systems.
> 
> On a F-9 system (perl-5.10.0-34.fc9), both tests passed cleanly (as expected).
> 
> On a F-8 system with the latest perl (perl-5.8.8-41fc8), the first
> test failed, but the second test passes.  This makes me worry, as
> we're failing a test here on an F-8 system that doesn't actually have
> this problem.
> 
> Given that these tests look for something that is specifically a
> RH/Fedora issue, would it be unreasonable to look for specific
> Fedora/RH information in determining the first test pass/fail?
> Unfortunately, it looks like prior to F-9, patchlevel.h wasn't upated
> with any applied patch info; so we can't use that...  But we could
> test for the presence of /etc/{fedora,redhat}-release, and use rpm to
> query for the rpm package version.  If any of these bits aren't there
> for us to query, no problem, we're not on an impacted system, and we
> can pass the first test.

Don't use rpm in these tests. Running rpm within the build environment 
(which would happen if the test suite did it) is bad because the rpmdb 
could be set up by a different version of rpm (on the buildhost) than 
the one in the target build environment (the target OS/arch) and hence 
could have an incompatible database format.

Paul.




More information about the Fedora-perl-devel-list mailing list