run-jsc-stress-tests supports a make-based runner and a shell-based runner. For local run both of them are supported. The goal is to make the make-based runner work (optionally) for remote testing too. It is much more faster than the shell-based runner and make is available on Linux, we can use it.
Created attachment 245713 [details] WIP patch
Comment on attachment 245713 [details] WIP patch View in context: https://bugs.webkit.org/attachment.cgi?id=245713&action=review > Tools/Scripts/run-jsc-stress-tests:-191 > when '--remote' > $copyVM = true > - $testRunnerType = :shell Now the default runner type is shell, which is good for iOS, but bad for Linux. But with this change, the default is make, which is good for Linux, but bad for iOS. I'm going to find a good way to have different default runner types based on OS or add some autodetection mechanism: use make if it is available, otherwise use shell. > Tools/Scripts/run-jsc-stress-tests:1408 > + remoteScript += "DYLD_FRAMEWORK_PATH=\\\"\\$(cd #{$testingFrameworkPath.dirname}; pwd)\\\" make -j4 -s -f Makefile" I'm going to find how to detect / set number of parallel tasks properly instead of hard coding.
new bug report for processor number detection: bug141196
Created attachment 245943 [details] Patch Complete patch, it depends on fix in bug141196 and bug141164. I'll set r? once these fixes landed.
Comment on attachment 245943 [details] Patch wrong patch
Created attachment 245944 [details] Patch
Comment on attachment 245944 [details] Patch webkit-patch upload is crazy :-/
Created attachment 245945 [details] Patch
Created attachment 246394 [details] Patch
Comment on attachment 246394 [details] Patch Clearing flags on attachment: 246394 Committed r179946: <http://trac.webkit.org/changeset/179946>
All reviewed patches have been landed. Closing bug.