ORWT's httpd locking and scheduling mechanism was implemented to be able to run multiple run-webkit-tests sessions besides each other without interfering, so the the tests needing httpd should mutually exclude each other. Since http://trac.webkit.org/changeset/90419 this mechanism is broken when using --child-processes=1 in two ways: 1) The httpd is started without using the checking mechanism implemented in http://trac.webkit.org/changeset/89414 and the script is crashing when there is no httpd present. 2) The httpd locks are acquired even if not running any http or websocket tests, so one testing session entirely blocks another.
Oh? Do we need to turn off NRWT for Qt for the meanwhile?
Hm. Neither of these things should be true. There was a bug introduced yesterday on chromium windows where we weren't checking for the http server properly, but the rest of this stuff should be working. I will investigate further.
Created attachment 100065 [details] Patch
Yup, I definitely broke it. This patch should fix things.
Comment on attachment 100065 [details] Patch Clearing flags on attachment: 100065 Committed r90651: <http://trac.webkit.org/changeset/90651>
All reviewed patches have been landed. Closing bug.