Re: [python-committers] Revert changes which break too many buildbots
On 15 June 2017 at 23:38, Victor Stinner wrote: > 2017-06-15 5:31 GMT+02:00 Nick Coghlan : >> For example, something that would be genuinely helpful would be a bot >> monitoring PR comments that could automate the "custom-build" dance >> for core developers (i.e. we'd be able to write something like >> "BuildBot: test custom build", and it would go away, kick off a custom >> BuildBot run by pushing the PR to the "custom-build" branch, and then >> report back the links for failed builds like >> http://buildbot.python.org/all/builders/x86%20Tiger%20custom/builds/5 >> ) > > I don't think that buildbots have enough resources to run tests on > each PR. We get too many PR and each PR is updated regulary. On some > buildbots, a single build can take 1 hour (or longer)... I wasn't proposing running every PR through BuildBot, I was suggesting having a more straightforward way of triggering custom builds directly from the PR we want to test after post-merge CI has indicated it's necessary. However, I think there's a way of handling that which doesn't require any new automation code to be written. Instead, we can have the policy be to post a link to https://docs.python.org/devguide/buildbots.html#custom-builders on the PR that is being reverted. That way, even if folks didn't notice the problem with the buildbots themselves: 1. They get a ping on a PR that they're necessarily following 2. They get a reminder of how to test the revised PR across the buildbot fleet prior to merging it again Hopefully reversions will continue to be rare (since relatively few changes are likely to be as platform dependent as PEP 538, and Windows/*nix differences are already covered in pre-merge CI), but when they do come up, the reminder of how to manually trigger pre-merge cross-platform CI is likely to be useful, and doesn't require much additional effort on the part of the folks doing the reversion. Cheers, Nick. -- Nick Coghlan | [email protected] | Brisbane, Australia ___ python-committers mailing list [email protected] https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
Re: [python-committers] Revert changes which break too many buildbots
2017-06-16 10:37 GMT+02:00 Nick Coghlan : > Hopefully reversions will continue to be rare (since relatively few > changes are likely to be as platform dependent as PEP 538, and > Windows/*nix differences are already covered in pre-merge CI), but > when they do come up, the reminder of how to manually trigger > pre-merge cross-platform CI is likely to be useful, and doesn't > require much additional effort on the part of the folks doing the > reversion. I'm also tolerant. A regression on macOS Tiger matters me less than a regression on most Linux buildbots. I tried to focus on recent versions of Windows, Linux, macOS and FreeBSD. AIX, OpenIndiana/Solaris, OpenBSD, etc. are broken for month. It's ok, I can live with that :-) But regulary, I propose to drop support for these platforms ;-) For AIX, I tried to skip a few tests known to fail on AIX. For OpenIndiana, we should simply remove this buildbot and replace it with a new IllumOS buildbot. Victor ___ python-committers mailing list [email protected] https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
Re: [python-committers] Revert changes which break too many buildbots
On Fri, 16 Jun 2017 at 03:40 Victor Stinner wrote: > 2017-06-16 10:37 GMT+02:00 Nick Coghlan : > > Hopefully reversions will continue to be rare (since relatively few > > changes are likely to be as platform dependent as PEP 538, and > > Windows/*nix differences are already covered in pre-merge CI), but > > when they do come up, the reminder of how to manually trigger > > pre-merge cross-platform CI is likely to be useful, and doesn't > > require much additional effort on the part of the folks doing the > > reversion. > > I'm also tolerant. A regression on macOS Tiger matters me less than a > regression on most Linux buildbots. I tried to focus on recent > versions of Windows, Linux, macOS and FreeBSD. > > AIX, OpenIndiana/Solaris, OpenBSD, etc. are broken for month. It's ok, > I can live with that :-) But regulary, I propose to drop support for > these platforms ;-) For AIX, I tried to skip a few tests known to fail > on AIX. For OpenIndiana, we should simply remove this buildbot and > replace it with a new IllumOS buildbot. > https://www.python.org/dev/peps/pep-0011/ backs you up in dropping support if we can't get a buildbot *and* someone to maintain the support (and it sounds like we don't have the latter ATM for those platforms). Maybe we should amend PEP 11 to say that whomever volunteers to maintain a platform must make sure that platform's buildbot is not red for longer than a month (to give volunteers some time to notice the fail and fix it in case it happens while they are e.g. on vacation)? Otherwise we will consider the platform unsupported. ___ python-committers mailing list [email protected] https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
Re: [python-committers] Revert changes which break too many buildbots
On 06/16/2017 09:48 AM, Brett Cannon wrote: Maybe we should amend PEP 11 to say that whomever volunteers to maintain a platform > must make sure that platform's buildbot is not red for longer than a month [...] How about three? Some life changes need more than a month to recover from... (death in the family, life in the family, job loss, etc.) -- ~Ethan~ ___ python-committers mailing list [email protected] https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
[python-committers] Bedevere now automatically removes "needs backport to *" labels
When you create a backport PR, if the title is formatted as, e.g. "[3.6] stuff that changed (GH-1234)", then Bedevere will remove the "needs backport to 3.6" label on the GH-1234 PR and leave a comment linking to the backport PR that triggered the label removal. ___ python-committers mailing list [email protected] https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
Re: [python-committers] Revert changes which break too many buildbots
On Fri, 16 Jun 2017 at 13:24 Ethan Furman wrote: > On 06/16/2017 09:48 AM, Brett Cannon wrote: > > > Maybe we should amend PEP 11 to say that whomever volunteers to maintain > a platform > > must make sure that platform's buildbot is not red for longer than a > month [...] > > How about three? Some life changes need more than a month to recover > from... (death in the family, life in the family, > job loss, etc.) > True. I guess as long as we are upfront that the platform's buildbot is known to be failing and the clock has started (so we all know to ignore the failure), then I'm fine with 3 months. ___ python-committers mailing list [email protected] https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
