Front page | perl.perl5.porters |
Postings from March 2022
On 3/10/22 08:31, Paul "LeoNerd" Evans wrote: > [snip]> > Can we just turn this on now? > > Furthermore, can we get this in in time to test it on 5.35.10 so it's > part of 5.36? > Without commenting on the merits of the specific change you're proposing, I'd just like to say ... We seem to be in a rush to cram as many changes as possible into the upcoming release, notwithstanding the fact that we're already more than two weeks past the "Contentious changes freeze" point. This is risky. It assumes a greater degree of ability to discover flaws in our code in a short period of time than we currently have. That ability to discover flaws has two aspects: (1) smoke-testing of the core distribution; (2) and testing of downstream impact of changes in blead via CPANtesters. Our smoke test aggregators have never fully recovered from the test-smoke.org disk failure of a couple of years ago, nor are we getting the volume of smoke test reports that we used to. That means that problems that several years ago would have been detected within hours now may take days to be detected. As for CPANtesters ... while we still get a large number of reports against older versions of perl, the number we get against dev releases is much smaller. It can take several months for failures of CPAN distros against dev releases to percolate up to the creation of Blead Breaks CPAN (BBC) tickets. Once a BBC ticket is filed, we have to bisect the problem, then determine what portion of the problem is in blead and what portion is in the CPAN code. That takes time. Rather than asking, "How much more can we cram in in the next 10 days?", the PSC and core committers should be asking, "Is what we have in blead already truly rock-solid and non-contentious?" and "Do we have the capacity to detect and remedy remaining flaws in the last two months before the production release?" Thank you very much. Jim KeenanThread Next