+1 Tested against a wide range of applications and no obvious problems found so far.
For our regular test against master: WIP, ETA around end of September (hopefully faster). Thanks, Marco On August 2, 2018 9:50:49 PM GMT+02:00, Wes McKinney <wesmck...@gmail.com> wrote: >I think it's OK to proceed with the next RC (RC1) if the code >described in https://issues.apache.org/jira/browse/ARROW-2963 works >properly after the fix. Please let us know tomorrow if you run into >any more issues. Thank you! > >On Thu, Aug 2, 2018 at 3:38 PM, Marco Neumann ><ma...@crepererum.net.invalid> wrote: >> I'll test the PR tomorrow (Friday, until 15:00 UTC). Thanks for the >quick fix! >> >> @Wes Might be doable, I'll check how we can improve there. Sorry for >catching this problem that late. >> >> I'm totally fine with the "no veto" policy. It's a bug for which no >test existed beforehand, and a behavior / feature that was just >implicitly assumed to exist (fork stability). So the regression kinda >"normal". >> >> Marco >> >> On August 2, 2018 8:40:05 PM GMT+02:00, Phillip Cloud ><cpcl...@gmail.com> wrote: >>>Marco, what would be even better is if you would test your >application >>>against Antoine's PR before it gets merged. >>> >>>On Thu, Aug 2, 2018 at 2:37 PM Phillip Cloud <cpcl...@gmail.com> >wrote: >>> >>>> I'll cut another one. Can someone review Antoine's PR, like ASAP, >so >>>that >>>> I can cut another RC? >>>> >>>> On Thu, Aug 2, 2018 at 2:32 PM Wes McKinney <wesmck...@gmail.com> >>>wrote: >>>> >>>>> -1 >>>>> >>>>> It's up to Phillip whether he wants to cancel the RC, but note >that >>>>> releases cannot be vetoed. >>>>> >>>>> @Marco, thanks for reporting -- is there a process which could >have >>>>> surfaced this issue sooner (e.g. testing your application >regularly >>>>> against master)? >>>>> >>>>> Thanks, >>>>> Wes >>>>> >>>>> On Thu, Aug 2, 2018 at 2:29 PM, Antoine Pitrou ><anto...@python.org> >>>>> wrote: >>>>> > >>>>> > Le 02/08/2018 à 20:15, Li Jin a écrit : >>>>> >> Antoine, >>>>> >> >>>>> >> Do you think we should fail this RC because of Arrow-2963? >>>>> > >>>>> > It's a regression, so ideally it should be fixed. >>>>> > Furthermore, the issue can be reproduced quite easily in Python >>>with the >>>>> > (popular) multiprocessing package. >>>>> > >>>>> > Regards >>>>> > >>>>> > Antoine. >>>>> >>>>
