Closed jarzec closed 1 month ago
1291 addresses one such failure.
Well, I was rather thinking about fixing the code than modifying the tests to make the faulty code pass. There is actually an LLVM bug I've reported that affects all (almost all) the Clang versions we use. Nevertheless an elegant workaround modifying the code itself seems achievable.
EDIT: I just saw I didn't click "Update" on my edit, so here it goes: on a second though a mod of the test could serve as a stepping stone making all tests pass until a better solution is available. I guess that would be @hsutter's call.
Well, I was rather thinking about fixing the code than modifying the tests to make the faulty code pass. There is actually an LLVM bug I've reported that affects all (almost all) the Clang versions we use.
I do wonder.
I noticed this while working on #526,
before requiring parameters to be named _
,
and lowering only the parameter's type.
I'm fairly sure that, at the time, I checked with the C++ Standard whether Clang was right. I think I found out Clang was indeed right, and as a result I opened no compiler bugs.
Outdated, replaced with #1324
Updates outdated regression test output files. Test failures to be addressed separately.