00:03 <LePhilousophe> it's not an error

00:03 <LePhilousophe> you can't provide the information because it isn't there

00:04 <LePhilousophe> what do you provide to the try scheduler? if it's not a github PR you can't push anything to it

00:05 <kkleine> LePhilousophe, let me ask differently. Who normally provides the information of repo_owner and repo_name? :)

00:05 <LePhilousophe> the sourcestamp

00:05 <kkleine> LePhilousophe, I didn't fully grok this concept.

00:06 <kkleine> LePhilousophe, project = sourcestamp['project']

00:06 <kkleine> repository = sourcestamp['repository'] seem to be relevant and so the question can be asked who is supposed to provide this information?

00:07 <LePhilousophe> when you have an build triggered, a sourcestamp is generated for the code checked out

00:07 <LePhilousophe> it contains various informations like repository, commit id, etc.

00:08 <kkleine> LePhilousophe, so that sounds like it is the scheduler's job?

00:08 <LePhilousophe> scheduler and change hook I think

00:09 <LePhilousophe> but what do you provide through the TryScheduler?

00:09 <LePhilousophe> I think that if you want to push to GitHub you must use a GitHubPullrequestPoller or sth like that

00:11 <kkleine> LePhilousophe, "but what do you provide through the TryScheduler?" you mean what properties I provide when calling "buildbot try --property=key=value?"

00:12 <LePhilousophe> yes

00:12 <kkleine> LePhilousophe, it looks as if this function of the Try Scheduler handles the repo and project, right? https://github.com/buildbot/buildbot/blob/e630745771/master/buildbot/schedulers/trysched.py#L128

00:12 <kkleine> One second for the params

00:12 <kkleine> I meant properties

00:13 <LePhilousophe> in fact the question is more: what do you try to achieve

00:15 <LePhilousophe> because if I understand correctly try scheduler, it's for patch provided through buildbot try command. It's not something which is in a Github PR

00:15 <kkleine> LePhilousophe, I try to kick off a build from "anywhere" using "buildbot try" and after the build was run or when it was kicked off have a message appear as a regular comment in a PR that is associated with it. I noticed that amongst my properties passed to buildbot try, there's no "project".

00:16 <LePhilousophe> you should use GitHubPullrequestPoller instead

00:16 <LePhilousophe> or the Github change hook

00:17 <LePhilousophe> maybe you can hack the thing by defining manually the missing properties

00:18 <LePhilousophe> and note that I am just a user of buildbot :)

00:26 <kkleine> LePhilousophe, my setup is quite exotic. We want to be able to build any PR on any try builder upon user request. I think the GitHubRequestPoller or Github Change Hook won't allow to react upon this. At the moment a try build is kicked off when I have a comment like "/must-build trybuiler0" in my PR. Strange setup I know.

00:30 <kkleine> LePhilousophe, I have added the missing project and repository properties to the buildbot try call but the problem is still the same in buildbot

00:36 <kkleine> LePhilousophe, at the moment github cannot publically reach the buildbot server, which is why I have a self-hosted actions runner that federates the request to issue a try build when it sees a "/must-build-on trybuiler" comment.

00:37 <kkleine> I think this disqualifies the buildbot change hook setup.

00:40 <kkleine> LePhilousophe, but I still wonder if there's anything I can do to teach "teach" the GitHubCommentPush function which repo to push the comment to. Be it through buildbot try properties or some setting in the scheduler or so. No real user is calling the buildbot try command in our scenario. Instead we hide it behind a comment on a PR. This is to hide complexity of buildbot try but also to not expose it directly at all to anybody.

00:46 <kkleine> Will go to bed and try more on Monday. Happy Weekend and thanks for your help LePhilousophe .

11:55 <shurdeek> can I remove duplicate jobs? I have github hooks both for upstream and a fork. When I open a PR and then push to the branch, it runs every job twice: once because a branch in the fork was modified, and once because the PR was modified.

11:55 <shurdeek> this is an unnecessary waste of resources