run_webkit_tests.py shouldn't have platform-specific logic
Created attachment 53040 [details] Patch
Created attachment 53041 [details] Patch
Comment on attachment 53041 [details] Patch Looks basically good to me, although I'm really not that familiar with what the Executive package does; I'll have to look into that more now that this code is rather dependent on it.
Thanks. Executive.cpu_count shouldn't exist. It only exits because the multiprocessing package is in Python 2.6 and later. We shouldn't have a bunch of different pieces of code that work around the lack of the multiprocessing. In principle, we could just implement our own version of multiprocessing, like we do for os.relpath, but we decide to put it on executive so it could be non-static (which is better for testing). The main point of this patch is that it makes sense for a WebKit port to recommend a default number of DRT instances. It doesn't make a lot of sense for a WebKit port to tell you how many CPUs your machine has. Some lower-level piece of code should do that. The webkitpy.common.system package houses all the non-WebKit specific code for interacting with the operating system (that isn't supplied by the standard library), so the CPU counting code should live there. Now, what to do with the CPU count, that's a question for the layout_test package.
Comment on attachment 53041 [details] Patch I wish executive had more testing. OK. We need to kill the 4-limit hack. I have never reproduced the errors Dirk saw, but I would like to... so we can fix them.
Comment on attachment 53041 [details] Patch Clearing flags on attachment: 53041 Committed r57422: <http://trac.webkit.org/changeset/57422>
All reviewed patches have been landed. Closing bug.