Hi Jakub, On 20 Jan 2025, at 10:15, Jakub Jelinek wrote:
> On Mon, Jan 20, 2025 at 08:52:17AM +0000, Simon Martin wrote: >> On 12 Jan 2025, at 12:10, Simon Martin wrote: >> >>> While testing an unrelated C++ patch with "make check-c++-all", I >>> noticed that r15-6760-g38a13ea4117b96 added a test case that fails >>> with >>> -fimplicit-constexpr. >>> >>> The problem is that this test unconditionally expects an error >>> stating >>> that a non-constexpr function is called, but that function is >>> auto-magically constexpr'd under -fimplicit-constexpr. > > Is that the only testcase that regresses with -fimplicit-constexpr? > I'd expect tons of others... Actually yes, I test all my patches with check-c++-all and this is the only case that fails in that mode and does not with standard make check. >>> This patch adapts the test to also pass with -fimplicit-constexpr. >>> >>> Successfully tested on x86_64-apple-darwin19.6.0 with >>> make check-c++-all RUNTESTFLAGS="dg.exp=constexpr-asm-5.C" >> Ping? This is borderline obvious but since I’ve never played with >> inline assembly, I’d appreciate a second pair of eyes. > > Wouldn't it be easier to just add -fno-implicit-constexpr to > dg-additional-options? That would work indeed, and it’s true that it’s probably more in the spirit of that test. I’m happy to do that instead - I’ll do another testsuite run with that and I can merge it later today if that works for you. Thanks, Simon
