Operation Buildborg

k900 commented

2024-08-31 11:15:21 +00:00

Member

aka building ofborg on top of buildbot

Core ofborg features we need in rough order of complexity:

eval checks
- should be made part of the normal buildbot eval (release-checks?)
rebuild count reports
- special buildbot job that does two evals and runs a diff
- refactor service to post results to gerrit
testing the affected packages
- can we somehow use buildbot's change tracking for this? need investigating
magic ofborg build/eval/test commands
- add a ForceScheduler for starters, then we can maybe do a webhook to start builds from gerrit commands
pinging maintainers
- oof

aka building ofborg on top of buildbot Core ofborg features we need in rough order of complexity: - eval checks - should be made part of the normal buildbot eval (release-checks?) - rebuild count reports - special buildbot job that does two evals and runs a diff - refactor service to post results to gerrit - testing the affected packages - can we somehow use buildbot's change tracking for this? need investigating - magic ofborg build/eval/test commands - add a ForceScheduler for starters, then we can maybe do a webhook to start builds from gerrit commands - pinging maintainers - oof

k900 referenced this issue from the-distro/infra

2024-08-31 11:16:34 +00:00

Run ofborg on bagel-infra #4

jade commented

2024-09-02 00:31:08 +00:00

Owner

Pinging maintainers should possibly be part of a separate tagging bot tbh, but I can also see wanting to just shove it in buildbot to use buildbot to deal with the gerrit API bits.

It looks like Google has a thingy here that has a web plugin for scheduling "tryjobs" that they use on chromium/gerrit/go/others https://chromium.googlesource.com/infra/gerrit-plugins/buildbucket/+/refs/heads/main/web/. We could potentially take (inspiration from) some of that code for a clean Gerrit integration.

Pinging maintainers should *possibly* be part of a separate tagging bot tbh, but I can also see wanting to just shove it in buildbot to use buildbot to deal with the gerrit API bits. It looks like Google has a thingy here that has a web plugin for scheduling "tryjobs" that they use on chromium/gerrit/go/others https://chromium.googlesource.com/infra/gerrit-plugins/buildbucket/+/refs/heads/main/web/. We could potentially take (inspiration from) some of that code for a clean Gerrit integration.

k900 commented

2024-09-02 11:01:00 +00:00

Author

Member

The reason I want pinging maintainers in ofborg and not in some other place is for it to be consistent with what packages we consider "affected" by a certain PR (i.e. what gets rebuilt etc)

lilyinstarlight commented

2024-09-16 15:13:25 +00:00

Owner

I'll probably comment with more ofborg load-bearing quirks, but not listed yet above is eval cleanliness checks, i.e. no trace warnings or anything during normal package-set eval with aliases disabled. In ofborg this happens during each outpath calculation step and during check-meta-queue-builds step, but not during other eval checks (and e.g. the aliases-enabled check definitely needs to not check for trace output)

As long as it happens at all though, I'm indifferent where/how it happens in Buildborg

I'll probably comment with more ofborg load-bearing quirks, but not listed yet above is eval cleanliness checks, i.e. no trace warnings or anything during normal package-set eval with aliases disabled. In ofborg this happens during each outpath calculation step and during check-meta-queue-builds step, but not during other eval checks (and e.g. the aliases-enabled check definitely needs to not check for trace output) As long as it happens at all though, I'm indifferent where/how it happens in Buildborg

Operation Buildborg #20