develooper Front page | perl.perl5.porters | Postings from October 2016

Re: Perl 5.24.1-RC4 is now available!

Thread Previous | Thread Next
James E Keenan
October 18, 2016 11:58
Re: Perl 5.24.1-RC4 is now available!
Message ID:
On 10/17/2016 10:51 AM, James E Keenan wrote:
> On 10/12/2016 05:50 PM, Steve Hay via perl5-porters wrote:
>> ***Please, please, please*** test your code against perl-5.24.1-RC4.
>> This is your last chance to point out any critical regressions before
>> v5.24.1 ships "for real," and we'd rather fix things for v5.24.1 than
>> for v5.24.2!
> perl-5.24.1-RC4
> testing command: 'sh ./Configure -des -Dusedevel -duseithreads && make
> -j8 test_harness'
> FreeBSD-10.3: PASS
> FreeBSD-11.0: FAIL: failed test 2 in t/op/alarm.t
> When t/op/alarm.t was run individually with:
> cd t; ./perl harness -v op/alarm.t; cd -
> ... it PASSed.
> These tests were performed in a virtual machine run by VMware sitting on
> Ubuntu Linux 16.04 LTS.  That may have adversely affected timings during
> make_harness.  So this may or may not be a showstopper.

I re-ran Configure and make test_harness on this FreeBSD-11 VM and got a 
PASS.  I think we can chalk up the one failure reported earlier to 
timing issues.  So I don't think this is a blocker for 5.24.1.

However, on list (or at upcoming P5P hackathon) we should probably have 
a broader discussion of test timing issues.  I know that Jarkko has been 
thinking about this recently (commit 
877b982928bd23c44e063a6506419b9c5e214b15, et. al.).  My hunch is that in 
a case like this where, in the hope of getting *some* coverage of blead 
on a given platform, we rely on a VM installed on someone's laptop, we 
are likely to get intermittent failures in timing-dependent tests.

Thank you very much.
Jim Keenan

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