165173 2016-11-29 16:22:28 -0800 run-benchmark should have a switch to avoid killing browser process unconditionally 2016-12-05 15:40:01 -0800 1 1 1 Unclassified WebKit Tools / Tests WebKit Nightly Build Unspecified Unspecified NEW P2 Normal --- 164792 1 clopez webkit-unassigned dewei_zhu lforschler rniwa oldest_to_newest 1254558 0 clopez 2016-11-29 16:22:28 -0800 After r209090 <http://trac.webkit.org/changeset/209090> run-benchmark in Linux doesn't execute a "killall $browser" when it is asked to run the benchmarks with browser $browser. It instead keeps track of the browser PID it executed and only kill this PID. This is a desired behaviour when the user is testing on the machine. However for a performance bot running tests automatically it isn't. After discussing this in bug 164792 it seems we can add a command line switch to avoid issuing the killall, so users that want to avoid this behaviour of the script can just activate the switch. And at the same time we can keep as default behaviour killing all process with name $browser. 1256555 1 rniwa 2016-12-05 15:40:01 -0800 (In reply to comment #0) > After r209090 <http://trac.webkit.org/changeset/209090> run-benchmark in > Linux doesn't execute a "killall $browser" when it is asked to run the > benchmarks with browser $browser. It instead keeps track of the browser PID > it executed and only kill this PID. > > This is a desired behaviour when the user is testing on the machine. However > for a performance bot running tests automatically it isn't. > > After discussing this in bug 164792 it seems we can add a command line > switch to avoid issuing the killall, so users that want to avoid this > behaviour of the script can just activate the switch. And at the same time > we can keep as default behaviour killing all process with name $browser. Alternatively, we can add an option like --kill-all to mean kill all existing browser instances to be used on bots.