The testing EWS bots should upload layout-test-results.zip to bugs after failures
Created attachment 92337 [details]
Comment on attachment 92337 [details]
View in context: https://bugs.webkit.org/attachment.cgi?id=92337&action=review
> def handle_script_error(cls, tool, state, script_error):
> + # FIXME: Why does this not exit(1) like the superclass does?
Probably a bug.
> + # FIXME: This probably belongs at a layer above AbstractPatchQueue, but shared by CommitQueue and the EarlyWarningSystem.
above => below.
Committed r85802: <http://trac.webkit.org/changeset/85802>
There were a few random git.py fixes which got rolled into this as I successfully made it possible to commit while not being in your git checkout. The rest of webkit-patch is designed to work even when your PWD is not inside a git tree, so landing should too. :) Turns out there were an epic number of places where we were missing cwd=self.checkout_root!
Reopened, because it killed the Qt EWS. See https://bugs.webkit.org/show_bug.cgi?id=74929 for details.
I don't think if it is a good idea to upload 8-10 Mb sizes diffs to the bugzilla as attachment many times. For example there are 73 Mb sizes EWS
attachment in https://bugs.webkit.org/show_bug.cgi?id=25206 and its XML
size is 105Mb. And unfortunately EWS tries to fetch the whole bug 3 times.
It would be better if EWS uploads the results to an other
storage, and adds only its URL to the bugzilla as comment.
We really should filter files out of the zip so it's just the interesting ones. That should cut down on the size dramatically.
As far as I know, bugzilla did not use to inline attachments into the XML. Since it's started doing that, EWS speeds have gone way down. We either need to move away from ctype=xml, or find a way to tell bugzilla not to inline (base64 encoded) attachments into the ctype=xml response.
The bug is fixed in https://bugs.webkit.org/show_bug.cgi?id=74929 . (https://trac.webkit.org/changeset/116934)