new-run-webkit-tests: add an --additional-drt-flag option
Created attachment 89827 [details] Patch
Comment on attachment 89827 [details] Patch What sort of flags? Should we just pass everything after -- to DRT? Should we have the flags listed one at a time in the NRWT arguments?
(In reply to comment #2) > (From update of attachment 89827 [details]) > What sort of flags? Should we just pass everything after -- to DRT? Should we have the flags listed one at a time in the NRWT arguments? The syntax would be new-run-webkit-tests --additional-drt-flag=--foo=bar test1.html test2.html We can't use "--" to separate things easily because we need to be able to specify tests on the command line. Also, optparse would get confused, and so we'd have to do some preprocessing of sys.argv which could be a bit messy. This is intended to replace at least most if not all of the chromium-specific options (incl. the GPU flags), e.g., the dozen or so in run_webkit_tests.py:224, like --stress-opt, --enable-hardware-gpu, --accelerated-2d-canvas, --gp-fault-error-box, etc.
FYI, the same sort of discussion just went on in bug 58370 about adding such a flag to webkit-patch. I think Adam convinced me to stay away from such for now. But I don't have hugely strong opinions here. I think it's somewhat awkward to tell devs the should use --additiona-drt-flag=--gpu instead of --enable-hardware-gpu for example. But this is really up to you all since this doesn't really affect non-chromium ports.
(In reply to comment #4) > FYI, the same sort of discussion just went on in bug 58370 about adding such a flag to webkit-patch. I think Adam convinced me to stay away from such for now. But I don't have hugely strong opinions here. I think it's somewhat awkward to tell devs the should use --additiona-drt-flag=--gpu instead of --enable-hardware-gpu for example. But this is really up to you all since this doesn't really affect non-chromium ports. Hm. I looked through that bug but didn't see the sort of discussion I was expecting; was that the right number? I actually think this may make life slightly easier for devs since they can use the DRT flag names directly without having to remember how they are mapped through NRWT. Plus it allows us to get the python out of the loop when people are adding new options to test all these wacky GPU configurations. Note that this is largely for individual developers testing on their machines. This is *not* for the bots (which is perhaps the opposite of the case in bug 58370). The bots need to have predictable sets of flags which are largely determined by the port name; NRWT shouldn't grow other command line args for those cases (and I agree that we should minimize the # of flags we have to pass to NRWT).
Comment on attachment 89827 [details] Patch OK.
Committed r84207: <http://trac.webkit.org/changeset/84207>