For some reason, running new-run-webkit-tests on the webkit mac port is unstable with more than 4 threads in parallel. I think this is a problem w/ DumpRenderTree somehow.
When you say "unstable" you mean that tests fail? DRT crashes?
Mostly DRT seems to hang as if it's blocking on locks. I haven't attempted to debug it yet, but it seems very reproducible at least on my 8-core Mac Pro.
sample DumpRenderTree should give us some information as to what's going wrong.
Created attachment 51802 [details] ignore this - wrong bug
Comment on attachment 51802 [details] ignore this - wrong bug ignore this patch - webkit-patch upload put it on the wrong bug.
You may be hitting the same issue that I hit when I did something like this. I fixed it in: https://bugs.webkit.org/show_bug.cgi?id=10906 To my memory, you just need to do something like this for each sepearate worker instance (it gave each DRT its own home directory): $ENV{CFFIXED_USER_HOME} = $testResultsDirectory . "/home" . WorkCrew::workerNumber();
init(In reply to comment #6) > You may be hitting the same issue that I hit when I did something like this. I > fixed it in: https://bugs.webkit.org/show_bug.cgi?id=10906 > > To my memory, you just need to do something like this for each sepearate worker > instance (it gave each DRT its own home directory): > $ENV{CFFIXED_USER_HOME} = $testResultsDirectory . "/home" . > WorkCrew::workerNumber(); Initial testing looks promising. Thanks!
Note that https://bugs.webkit.org/show_bug.cgi?id=36899 may also fix this, as might https://bugs.webkit.org/show_bug.cgi?id=33153 .
I am not currently working on this, nor do I expect to be in the near future, so I am disclaiming ownership in case someone else wants to take a look.
This should be fixed as of r79062.