nrwt: derive per-thread, per-worker stats from result summaries
Created attachment 177367 [details] Patch
Comment on attachment 177367 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=177367&action=review > Tools/ChangeLog:21 > + Also, this changes the stats so that we report only the time taken > + to actually run the test, rather than the time taken to run the test > + plus run matching reftests, compute diffs, etc. It's unclear which > + number is actually more interesting, but I suspect most people > + expect the stats to report the former; we can always add the latter > + back in if needed. I think for the individual test timings we care about the time actually spent in DRT. But for the aggregate numbers we care about the whole time. For example, all your numbers for the time spent in workers or per-directory will now not include time spent doing the other work, no? That seems less than ideal.
(In reply to comment #2) > (From update of attachment 177367 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=177367&action=review > > > Tools/ChangeLog:21 > > + Also, this changes the stats so that we report only the time taken > > + to actually run the test, rather than the time taken to run the test > > + plus run matching reftests, compute diffs, etc. It's unclear which > > + number is actually more interesting, but I suspect most people > > + expect the stats to report the former; we can always add the latter > > + back in if needed. > > I think for the individual test timings we care about the time actually spent in DRT. But for the aggregate numbers we care about the whole time. For example, all your numbers for the time spent in workers or per-directory will now not include time spent doing the other work, no? That seems less than ideal. Yeah, you're right. I was being a little lazy. I will collect both numbers.
Created attachment 177522 [details] collect both test_run_time and total_run_time
Comment on attachment 177522 [details] collect both test_run_time and total_run_time View in context: https://bugs.webkit.org/attachment.cgi?id=177522&action=review > Tools/Scripts/webkitpy/layout_tests/models/test_results.py:54 > + self.total_run_time = 0 # The time taken to run the test plus any references, compute diffs, etc. Where do we set this?
(In reply to comment #5) > (From update of attachment 177522 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=177522&action=review > > > Tools/Scripts/webkitpy/layout_tests/models/test_results.py:54 > > + self.total_run_time = 0 # The time taken to run the test plus any references, compute diffs, etc. > > Where do we set this? Here, next to where we set the worker name and shard name: > Tools/Scripts/webkitpy/layout_tests/controllers/layout_test_runner.py:304 > + result.total_test_time = time.time() - start
total_test_time != total_run_time
Created attachment 177535 [details] fix typos, lint
Created attachment 177537 [details] fix typos, lint, verify manually
(In reply to comment #7) > total_test_time != total_run_time Whoops. Fixed now. Clearly I need some tests for this and don't have them, but if it's okay with you I'd like to address that in a separate patch. I verified things look right manually.
Committed r136579: <http://trac.webkit.org/changeset/136579>