Overall I think this idea makes sense. Is it your intention that the try run should be 100% complete, or just that one exists and is associated with the review? In some cases I find myself invoking Autoland when I feel reasonably confident about the results.
On 22 January 2016 at 12:12, Daniel Minor <dmi...@mozilla.com> wrote: > The only requirement would be that a try run was made. Which jobs to run > would be at the discretion of the person requesting the try run. We would > also not enforce having a green try run, I think the results are still too > noisy at this point for that to be useful. Before I invoke Autoland, I make an effort to pin and tag intermittents and infra issues on my own try run to give an indication to sheriffs/reviewers that I’ve paid attention to the results. But I agree the results are far too flaky at the moment for this to be a requirement. As an aside, at some point it would be good to have an auto-starring bot for infra issues. > * Try jobs are also triggered outside of MozReview, having to run them > again inside MozReview would be annoying and/or wasteful of resources. Precisely for this reason, I often find myself submitting a contributor’s patch to MozReview to get the benefits of a good review tool, try landing, and integration. However, when Autoland is invoked to land the patches it wants to rewrite the commit messages to r=me. Because I obviously didn’t author the patches this is confusing and prevents me from using Autoland to land these patches. Either I need to tell contributors explicitly to submit their patches to MozReview, which in itself isn’t unreasonable but could drag out in time if the contributor isn’t responsive. Or I need to manually push to try and land it on inbound. It would be nice if it was possible to submit other people’s patches to MozReview. _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
