Hi On Tue, Mar 01, 2016 at 07:47:49PM +0100, Jakub Jelinek wrote: > On Tue, Mar 01, 2016 at 07:39:18PM +0100, Martin Jambor wrote: > > as Jakub requested, this patch deals with HSA "excess errors" in the > > libgomp library testsuite by passing -Wno-hsa to all of them. IIUC, > > that passing it in the second parameter of dg-runtest (as opposed to > > the third) means that it will apply even tests that have their own > > dg-options, which is presumably easier for everyone, provided that hsa > > will get is own libgomp testsuite directories. > > What is the difference betwee the $flags and $default-extra-cflags > arguments to dg-runtest?
well, exactly what I wrote in the original email and what you have quoted (and me as well) above. But let me quote the dejagnu source comment of dg-runtest, which is perhaps more clear: # FLAGS is a set of options to always pass. # DEFAULT_EXTRA_FLAGS is a set of options to pass if the testcase # doesn't # specify any (with dg-option). So if I changed DEFAULT_EXTRA_FLAGS rather than FLAGS, I'd have to go through all testcases specifying dg-options and add -Wno-hsa there too. Moreover, we'd have to add -Wno-hsa to all appropriate future testcases if they specify their own dg-options. Perhaps we should be using dg-additional-options in libgomp testsuite wherever possible but there certainly are testcases using dg-options. > You seem to stick -Wno-hsa into the former, > which to me looks like it will be mentioned as part of the test > names (e.g. when cycling through -O* options, -Wno-hsa would be printed > along with -O2 etc.)? Yes, that is an unfortunate side-effect. Furthermore, automated comparison scripts might be confused by the change (mine was, reporting all testcases as newly passed/xfailed and old as disappeared). But again, I do not have a strong preference, I can change the patches to use DEFAULT_EXTRA_FLAGS and am willing to be watching for fallout and fixing dg-options if you prefer that. So let me know what you consider nicer and I'll do it. Thanks, Martin