All this talk of inbound bustage, reminded me of a problem I had a few weeks ago.
I landed a change that was OK on try and inbound, but had actually introduced a serious bug for NUWA processes (sorry!). Over 24 hours later the problem was found by a B2G test run. I am not sure if this is purely automated or involves some manual steps. Fortunately, I had already fixed the problem in a follow-up patch, as that particular bit of the code had caused a different issue, which was reported by email as part of the inbound runs (thanks to jld for spotting that this had fixed the NUWA bug and dvander for the quick patch review). Even though I had not broken inbound, my patch would have had to have been backed out, because the bug was so serious. As the patch hit quite a lot of code, given the delay, this might have been pretty painful. I am currently totally reliant on try and inbound for catching B2G issues and I imagine it's the same for a lot of people. It would be really useful if we had even a small set of tests on try that were more realistic and didn't rely on emulators. Cheers, Bob _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
