On 16 March 2018 at 18:58, Jeroen Demeyer <[email protected]> wrote:
> On 2018-03-16 12:40, David Loeffler wrote: > >> IMHO, the patchbot should be reconfigured so that it re-tests ticket 0 >> after every upgrade, and if that test fails, the patchbot should stop >> > > Absolutely not. I agree that the current patchbot situation is not good, > but what you propose is much worse. Why is this "much worse"? Surely it's better for the patchbot to stop than for it to keep going and produce completely useless and misleading output? I emphasise that I'm *not* proposing that it *silently* stop, but rather that it report the fact that ticket 0 fails in some sensible way, rather than spamming a whole load of unrelated tickets with spurious failure reports. This issue has been discussed on this list before, and several people agreed that the patchbot should run tests on each new beta and halt if they don't pass, with Jeroen being the only person to defend the current system. For an analogy: imagine I am in my university office, marking a big pile of examination scripts, and suddenly I find that I can't go on, because my desk chair has collapsed on the floor and I can no longer see the papers on my desk. What shall I do? Should I (a) stop and report to maintenance that I need a new chair; (b) scrawl "FAIL" across all the remaining exam scripts (because I can't see them well enough to verify that they're correct), and hope somebody notices that something is amiss and brings me a new chair? I'm pretty sure that option (b) would get me fired rather quickly; and, of course, (b) is exactly isomorphic to the patchbot's current reaction when a new beta won't pass tests. The signal-to-noise ratio of the patchbot is dreadful right now; for instance, on #24086 -- a rather straightforward bug-fix in the modular forms code -- the last twelve patchbot runs, on five different machines, have all reported failures, and every single one of these is totally unrelated to the ticket. At this stage the patchbot is worse than useless (strictly worse, because it creates uncertainty among ticket reviewers and thus prevents perfectly good code from getting into Sage). David -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
