|Submitter||Diff||Changes||Open Issues||Last Updated|
|Error loading review requests:|
58 bytes, text/x-review-board-request
|Details | Review|
Looking at my infraherder prototype (http://wlach.github.io/treeherder/ui/infra.html#/last-finished), it seems like the mochitest dt9 chunk is frequently the last to finish on Linux debug. Looking at this example: https://treeherder.mozilla.org/#/jobs?repo=try&revision=89fccfd367377df5296d16f41b9933217e8c63f2 dt1: 44 minute(s) dt2: 56 minute(s) dt3: 53 minute(s) dt4: 26 minute(s) dt5: 63 minute(s) dt6: 33 minute(s) dt7: 49 minute(s) dt8: 56 minute(s) dt9: 83 minute(s) It seems like reshuffling the chunks is in order. It looks like dt9 is tests in the `devtools/client/debugger/test/mochitest/` directory. :chmanchester suggested that we may need to split that up, since it's run-by-dir.
According to ahal, chunk-by-runtime won't split manifests, so we might be able to fix this by splitting the tests in that dir into two (or more) manifests. We should probably re-normalize the chunks as well; it looks like this suite hasn't been touched for 11 months: https://hg.mozilla.org/mozilla-central/filelog/tip/testing/runtimes/mochitest-devtools-chrome.runtimes.json
Here's a try run with devtools/client/debugger/test/mochitest split into two manifests: https://treeherder.mozilla.org/#/jobs?repo=try&revision=359e706d089c Runtimes on linux64 debug for all chunks are: 18, 31, 27, 20, 18, 15, 12, 18, 26, 24 So this seems to have the desired effect.
Comment on attachment 8784397 [details] Bug 1294489 - Split directory into two manifests for better chunking, https://reviewboard.mozilla.org/r/73860/#review73562 Seems fine to me (sorry, was at a conference and PTO recently). It's worth noting that we are about to land a new debugger and working on an entirely new test suite, so in the next month or two we can retire these old debugger tests.
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/autoland/rev/d819b0431209 Split directory into two manifests for better chunking, r=jlongster