Elvis Stansvik

I just found a GitLab issue and it indeed seems to be a race condition: https://gitlab.com/gitlab-org/gitlab/-/issues/13134 Note that this issue mentions multiple pipelines being created when status updates come in...

Hm, if `GitLabStatusPush` was given a new `self.last_status : Dict[str, Deferred]` to hold the deferred corresponding to the last status report sent for a given commit, and its `sendMessage(...)` modified...

Something like: ```diff --- gitlab.py.orig 2022-07-10 10:26:36.474458392 +0200 +++ gitlab.py 2022-07-10 10:35:38.590691281 +0200 @@ -75,6 +75,7 @@ self.master, baseURL, headers={'PRIVATE-TOKEN': token}, debug=self.debug, verify=self.verify) self.project_ids = {} + self.prev_status = {}...

Meh nevermind, that probably won't work reliably since `sendMessage` yields to other things, before yielding to `createStatus`.

Thinking about it, isn't there a more fundamental problem in how `ReporterBase` works: Since (AFAICS) there is nothing in place to ensure statuses are sent out in the order in...

So maybe instead something like ```diff diff --git a/master/buildbot/reporters/base.py b/master/buildbot/reporters/base.py index 21efeb12e..8124b9fec 100644 --- a/master/buildbot/reporters/base.py +++ b/master/buildbot/reporters/base.py @@ -15,6 +15,7 @@ import abc +from collections import deque from twisted.internet import...

If the approach above is OK I can make a PR. Let me know if I should create one straight away and move discussion there. Could probably use some scrutiny...

Alright, I'll wait a bit and see if @tardyp has any input. I believe the patch has fixed the issue for us. We are running with it now and I've...

@g-arjones Thanks for the ping and for testing => https://github.com/buildbot/buildbot/pull/6603

I'm running into this problem. An additional complication for me is that I need everything to be done within a single "repeatable_read" transaction. So what I desire is something like...