So for the past couple of days we (qa) had things that come up that made us re-examine this backout immediately policy: 1) the security bug and the regression that the patch caused: bug 1173284 and bug 1189455 <https://bugzilla.mozilla.org/show_bug.cgi?id=1189455> respectively. a) In this instance, the regression that the patch causes gives a better situation than the original situation. we prefer not to back out in this instance. 2) https://bugzil.la/1189261 ; gfx defect; I was able to chat with the dev on irc and there's a bug 1188995 that has a dependency on it (as I was told ) for the Mac Os X platform. I asked him to create a patch for disabling this on the b2g side in which he quickly did so and it is in : bug 1189399 3) We (QA) came to realize recently that we also need to reevaluate what we mean by dogfood-blocker; as there are some things that some people don't regularly use (like bluetooth ( bug 1182644 <https://bugzilla.mozilla.org/show_bug.cgi?id=1182644> ) and NFC ( bug 1188639 <https://bugzilla.mozilla.org/show_bug.cgi?id=1188639> ) and a few other things).
The change in guideline that I am thinking about putting myself through is this: 1) If a regression is not worse than the original issue, I would prefer to keep the patch and then ask the dev who regressed to fix the regression as well. 2) I think I would prefer to backout if I can't get in touch with the developer in time for the next build that's coming out. 3) If I can get in touch with them, I would like to give them a chance to either turn it off for the next build. 4) If they cannot patch in time, then I will back out. This probably takes a bit more legwork; at the same time, I think it will help us move forward better and cause less tension of me backing out everything all the time. The end goal is to deliver at least 1 good dogfood build a week w/ the majority of the device working so one could use it as a regular device. If I could get some feeback on this approach, I would apprecate it. Regards, Naoki <https://bugzilla.mozilla.org/show_bug.cgi?id=1189455> On Mon, Jul 27, 2015 at 2:30 PM, Naoki Hirata <[email protected]> wrote: > Fair enough. > > I'll take it as a learning experience that I need to backout the related > tests as well as the issue and I take full responsibility for not backing > out the tests as well, which caused reds. > > Having said that, having cooperation from developers would be > appreciated. We're trying to get a working build out to our end users. > Not argue where the issues are. > > If there's no cooperation, I'll back out of the whole thing next time and > possibly asking sheriffs to help. > > Regards, > Naoki > > > > On Mon, Jul 27, 2015 at 2:21 PM, L. David Baron <[email protected]> wrote: > >> On Monday 2015-07-27 12:03 -0700, Naoki Hirata wrote: >> > The dev knew about the crash locally: >> > https://bugzilla.mozilla.org/show_bug.cgi?id=1186965#c24 >> > So shouldn't that prevent him from landing in the first place if he's >> > crashing locally? >> >> That comment was from after the regression was found, not before. >> >> Developers are not responsible for manually testing things across >> multiple devices and platforms unless there's particular reason to >> think that such testing is needed (e.g., because the change has >> platform-specific or device-specific code). >> >> The underlying problem here is that it's relatively easy to have a >> build that we consider unacceptable for our nightly users that still >> passes all of our automated tests. (In this case that appears to be >> because we're only running Firefox OS mochitests on top of Android >> ICS, which isn't what we're shipping to our nightly users.) If the >> automated test coverage covered most of the things that we consider >> essential for our nightly users, as it does for Firefox, we'd be >> much less likely to have these problems. >> >> And people doing backouts are still responsible for backing out >> changes correctly, and in a way that doesn't break other tests. >> >> -David >> >> -- >> 𝄞 L. David Baron http://dbaron.org/ 𝄂 >> 𝄢 Mozilla https://www.mozilla.org/ 𝄂 >> Before I built a wall I'd ask to know >> What I was walling in or walling out, >> And to whom I was like to give offense. >> - Robert Frost, Mending Wall (1914) >> > >
_______________________________________________ dev-b2g mailing list [email protected] https://lists.mozilla.org/listinfo/dev-b2g
