The buildbot must :
- Run the full test suite on master on a regular basis.
- Bissect test failure to find the root cause.
- Report the test to the person who made the diff.
Alternatively, if the test is flaky, we may want to report to the test owner.
Description
Description
| Status | Assigned | Task | ||
|---|---|---|---|---|
| Open | None | T261 Upper level roadmap | ||
| Open | None | T258 Quality Assurance Improvements | ||
| Open | deadalnix | T37 Setup 3 stage testing | ||
| Resolved | schancel | T38 Implement a buildbot for master |
Event Timeline
Comment Actions
Do we specifically want to use buildbot? Or is that being used loosely? How many different build configurations do we need?
Comment Actions
I can think of something ultra-simple that meets these requirements:
Write some scripts that integrate with Travis and just auto-pull new diffs out and submit them to travis (a little fuzzy on the particulars of what travis's API is like.. but I am pretty sure it's doable).
Does that sound evil/stupid/good/doable/reasonable/crazy?
Comment Actions
I don't care what we use as long as it works. I think the best option for us is to have some homebrew thing on a machine somewhere. Starting a container and building in it isn't too hard and we can have it do whatever we want with it.
Using travis is another option if API allows.