The plan is to have a bisection tool to use Taskcluster as this is what we're moving to. bisection tool will be used with gecko revisions.
Usually what we do is pick dates then see where the first break is and then swap the gecko gaia to determine which is broken. I think we have to do the swap first and then do a git bisect with gaia or use the tool against gecko. The problem doing the swap first is if the regression window is too large, we might not have the apis from gecko and it might break in a different fashion. It's something that we need to figure out. On Mon, Jun 15, 2015 at 1:37 PM, Michael Henretty <mhenre...@mozilla.com> wrote: > > On Mon, Jun 15, 2015 at 11:29 AM, No-Jun Park <np...@mozilla.com> wrote: > >> One other might-be-related thing that comes into my mind is that if we’re >> moving away from pvtbuilds to taskcluster, we might need to think about how >> that is going to affect the bisection task. If it’s basically the same task >> with different build source, that’s great, but if there is some subtle >> differences (like how long each builds are stored) it might be nice to plan >> ahead about it. > > > Naoki, I seem to remember you were working on a gecko bisection tool > (something similar but better to what QAnalytics uses). Is this still in > progress, and will it be able to leverage task cluster artifacts? > >
_______________________________________________ dev-b2g mailing list dev-b2g@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-b2g
