[chromium] fix image diffing in NRWT
Created attachment 69710 [details] Patch
This explains why the flakiness dashboard hasn't been generating new image diffs.
...at least on Linux. Haven't tested on other platforms.
Comment on attachment 69710 [details] Patch patch LGTM, although I'm not a reviewer.
Comment on attachment 69710 [details] Patch ok, I believe the dpanke's comment :-)
Comment on attachment 69710 [details] Patch Clearing flags on attachment: 69710 Committed r69066: <http://trac.webkit.org/changeset/69066>
All reviewed patches have been landed. Closing bug.
(In reply to comment #6) > (From update of attachment 69710 [details]) > Committed r69066: <http://trac.webkit.org/changeset/69066> I rolled it out. If image_diff exited with neither 0 nor 1, NRWT stopped by a ValueError.
(In reply to comment #8) > (In reply to comment #6) > > (From update of attachment 69710 [details] [details]) > > Committed r69066: <http://trac.webkit.org/changeset/69066> > > I rolled it out. > If image_diff exited with neither 0 nor 1, NRWT stopped by a ValueError. An example: http://build.chromium.org/buildbot/waterfall.fyi/builders/Webkit%20(webkit.org)/builds/35915/steps/webkit_tests/logs/stdio
(In reply to comment #9) > (In reply to comment #8) > > (In reply to comment #6) > > > (From update of attachment 69710 [details] [details] [details]) > > > Committed r69066: <http://trac.webkit.org/changeset/69066> > > > > I rolled it out. > > If image_diff exited with neither 0 nor 1, NRWT stopped by a ValueError. > > An example: > http://build.chromium.org/buildbot/waterfall.fyi/builders/Webkit%20(webkit.org)/builds/35915/steps/webkit_tests/logs/stdio That's too bad. Maybe flushing the temp files isn't enough on Windows. We might have to switch to real files that we can close(). I'll try testing on Windows. Thanks for trying to land this for me.
Created attachment 69825 [details] Patch
It turns out that on Windows, you can't open the the file created by NamedTemporaryFile. So I updated the patch to just make a temp dir and open the files using open().
Committed r69153: <http://trac.webkit.org/changeset/69153>
how annoying. I'll have to remember that for next time.
also, it sure seems lately that I (at least) would benefit from try servers that ran test-webkitpy. I'll see if I can add that step to the layout bots ...