Thanks for the encouragement!

That sort of clean is is definitely something I want to do. It’s not a small task though, and isn’t going to happen in the next few months. Despite the state things are in, people still throw patches our way, and I hope what I am doing this quarter will encourage that.

I was just catching up on newsgroups and shaver mentioned that in the Cairo project the reviewers tend to do things like style fixes, adds tests, etc. instead of r-’ing and bouncing back to contributors, and that it makes for a better experience. Your “test Monday” idea seems very much in the same spirit.

I am very interested in hearing about your experiences with this, and I’ll bet it would encourage others to do this sort of thing as well.

BuildSlaves.py is one example. You can’t even run buildbot checkconfig locally without trying to make up some file that doesn’t even have a sample in the upstream repo.

The staging dirs in the buildbot-configs repo is another example. We should get to better factorization there. I still think that a try repo for that is better than some dirs with config changes that are mostly stuff that could be factored into BuildSlaves.py. Like, the slaves are all in BuildSlaves, but which slave belong to which build pool is still in master.cfg, and varies from staging to staging2 to production. The pools could be nicely configured inside BuildSlaves.py. I spearheaded that in my l10n master v2, which has the slaves for my builders encoded in http://hg.mozilla.org/users/axel_mozilla.com/l10n-master/file/e6c4bf4e3699/BuildSlaves.py. Trivial sample, yeah, and I really don’t want parallelism in that one, but you get the idea.

If we get the setup data factored out, we might be able to drop the staging dirs alltogether. Like, we don’t have staging for buildbotcustom, too. Which kinda confuses me.

If people wanted to do real testing locally, being able to just focus testing on particular features would be cool. And more tests and testing environments would help. I try (and more often than not fail) to write tests for my code. I just realized today that once again my builds pass, but my tests don’t. But there are ways to set up scenarios which can significantly help in testing off-site and off-stage.

I hope this announcement comes with support from management; these are the types of initiatives I had always wanted to undertake, but it was, lamentably, never a viable option.