On Tuesday, January 14, 2014 01:09:42 PM Michel Dänzer wrote: > On Mon, 2014-01-13 at 14:12 -0800, Dylan Baker wrote: > > I don't see a problem with always using --quick, however, I believe that > > -- > > quick is in effect when loading the quick profile. > > > > Here is how I verified it: > > cd into the piglit root and run the attached script with python. It should > > print ['--quick'] > > Running quick.tests: > > daenzer 28641 53.7 2.9 438544 242484 pts/0 Tl 12:38 1:05 > /home/daenzer/src/piglit-git/piglit/bin/glean -o -v -v -v -t +texCombine > > No --quick. > > > I killed it with Ctrl-C after several minutes (completes in 3 seconds > with --quick), which pointed out another problem: Ctrl-C kills > individual piglit tests, but records them as 'pass' and doesn't abort > the piglit run. I think this may be another recent regression, but I > haven't isolated which change introduced it. I noticed Ctrl-C no longer > aborted a piglit run last week, but I only noticed today that the killed > tests are recorded as 'pass'.
The conversion first threadpool replacement caused, and using python's included multiprocessing.dummy also causes ctrl-c to kill an individual process, use ctrl-\ to kill the whole run. I should have mentioned it in the commit message. However, that aborted tests are recorded as pass is concerning, I'll look into that. I also know what the problem with glean tests is and have a solution in mind.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
