Remove old-run-webkit-tests support from webkitpy It's just confusing things. Only Apple Win still uses old-run-webkit-tests, and the Apple-Win EWS does not run tests. I believe partially-broken NRWT support is at the root of bug 103839, in part because NRWT is currently a second-class "experimental" citizen in the EWS code. Since NRWT has now long been the default, we should make it a real boy and kill ORWT (since I'm very likely to break it by re-working this code anyway). At least the following tools will break: 1. sheriff-bot, webkit-patch will no longer understand ORWT results files (from Apple-Win buildbots, etc.) 2. webkit-patch build-and-test (something no one uses) will not work on AppleWin 3. The AppleWin EWS cannot graduate to a testing-EWS w/o moving to NRWT (that's probably for the best, given how impossibly slow it is already). There is no urgency to this, but again, I think it's likely I may break ORWT support in webkitpy by accident as part of fixing bug 103839.
I should note that this will *not* affect the buildbots, or those calling run-webkit-tests directly. Only webkit-patch (and associated commands) which currently know how to run old-run-webkit-tests or interpret its results (as I've listed above).
We could also investigate modifying ORWT to output NRWT-compatible stuff. I don't know if that's more or less work than finishing getting NRWT running on win.
Created attachment 181044 [details] Patch
Comment on attachment 181044 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=181044&action=review I suspect there is more than this. :) But this is a start. > Tools/Scripts/webkitpy/common/net/buildbot/buildbot.py:88 > return LayoutTestResults.results_from_string(results_file) This could call results_from_json directly now, I think. > Tools/Scripts/webkitpy/common/net/layouttestresults.py:63 > # FIXME: run-webkit-tests should store the --exit-after-N-failures value This seems like this would be trivial for one of us to fix. I'm kinda surprised this is still here. :( > Tools/Scripts/webkitpy/layout_tests/controllers/layout_test_runner.py:-182 > - # This differs from ORWT because it does not include WebProcess crashes. I would leave this in? I'm not sure if WebKit2 folks care about this or not.
(In reply to comment #4) > (From update of attachment 181044 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=181044&action=review > > I suspect there is more than this. :) But this is a start. > > > Tools/Scripts/webkitpy/common/net/buildbot/buildbot.py:88 > > return LayoutTestResults.results_from_string(results_file) > > This could call results_from_json directly now, I think. Searching 1977 files for "results_from_json" 0 matches across 0 files If you mean ResultsJSONParser.parse_results_json, we could but we'd still need to construct a LayoutTestResults interface to wrap the results. > > Tools/Scripts/webkitpy/common/net/layouttestresults.py:63 > > # FIXME: run-webkit-tests should store the --exit-after-N-failures value > > This seems like this would be trivial for one of us to fix. I'm kinda surprised this is still here. :( Yeah. I think it might actually be in the JSON already. > > Tools/Scripts/webkitpy/layout_tests/controllers/layout_test_runner.py:-182 > > - # This differs from ORWT because it does not include WebProcess crashes. > > I would leave this in? I'm not sure if WebKit2 folks care about this or not. Fixed.
Created attachment 181045 [details] Patch for landing
Thank you. Related bug 103839 is on my short-list for next week. (Although you're welcome to tackle it before then if you are in the mood.)
Comment on attachment 181045 [details] Patch for landing Clearing flags on attachment: 181045 Committed r138634: <http://trac.webkit.org/changeset/138634>
All reviewed patches have been landed. Closing bug.