dblaikie added a comment. In D101805#2740150 <https://reviews.llvm.org/D101805#2740150>, @tlively wrote:
> @dblaikie, what's the best practice for making tests robust to this > difference? If you need a label or register (eg: if you want to check a branch branches to some particular spot) - the first mention of the label or register should use a pattern match (which I think this test already do generally for instructions/values - but not for labels) I'm not sure there's a canonical way to skip over the unused "entry" label in an asserts build while not tripping up the non-asserts build (perhaps you could look around at similar tests) - I guess probably not & people accept a bit of looseness in the tests by not using CHECK-NEXT for that first instruction/line in the test. You could add a CHECK-NOT (or --implicit-check-not) maybe for " = " which would catch some instructions (not all of them - those that don't return a value, for instance) if they snuck in between the start of the function and the first instruction the test was checking for. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D101805/new/ https://reviews.llvm.org/D101805 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
