Joseph Myers <josmy...@redhat.com> writes: > On Thu, 28 Aug 2025, Jeff Law via Gcc wrote: > >> > I find the classification this would provide useful, like I like the >> > (internal compiler error) classification we already have. I would >> > of course not duplicate all of th eabove message but only >> > '(unrecognizable insn)' in the above case. >> It's not the only consideration, but keep in mind that such output is not >> stable and will cause some headaches with scripting that compares two summary >> files. > > Classifications really don't belong in the test name (the thing after > "PASS: " or "FAIL: ", until end of line) at all, they belong as separate > metadata so the set of test names can be stable when the testsuite itself > doesn't change. Unfortunately DejaGnu output doesn't have any clean way > to provide such metadata for a test result that can be reliably associated > with it but is clearly separate from the test name.
This is not to disagree (as I do completely agree, in fact), but does this have any bearing on the current "internal compiler error" detection-and-rewriting? I was planning on extending that to -fcompare-debug failures as suggested by richi. Your message (along with some of the others) make me think that may not be desirable. I feel a bit nervous about us rejecting any changes to the current setup because DejaGnu is DejaGnu (what are going to do instead?), but I of course understand changes can be disruptive to comparison scripts too. Maybe that's a better topic for Iain's BOF at Cauldron though?
