mstorsjo added inline comments.
================ Comment at: clang/test/Driver/cl-options.c:567 -// This test was super sneaky: "/Z7" means "line-tables", but "-gdwarf" occurs -// later on the command line, so it should win. Interestingly the cc1 arguments -// came out right, but had wrong semantics, because an invariant assumed by -// CompilerInvocation was violated: it expects that at most one of {gdwarfN, -// line-tables-only} appear. If you assume that, then you can safely use -// Args.hasArg to test whether a boolean flag is present without caring -// where it appeared. And for this test, it appeared to the left of -gdwarf -// which made it "win". This test could not detect that bug. +// If We specify both /Z7 and -gdwarf we should get dwarf, not codeview. // RUN: %clang_cl /Z7 -gdwarf /c -### -- %s 2>&1 | FileCheck -check-prefix=Z7_gdwarf %s ---------------- smeenai wrote: > bogner wrote: > > smeenai wrote: > > > bogner wrote: > > > > rnk wrote: > > > > > I think Meta folks depend on a mode that emits both codeview and > > > > > DWARF. +@smeenai > > > > Hopefully if that's the case there's a better test somewhere than this > > > > one that discusses in detail how `/Z7` is an alias for > > > > `-gline-tables-only`, which hasn't been true since 2017. > > > Thanks for the headers up :) We don't depend on that mode any more > > > though, so no problems on our end. > > Also note that this change does not affect what happens if we specify both > > `-gcodeview` and `-gdwarf` together. That continues to pass both > > `-gcodeview` and `-dwarf-version=...` to `clang -cc1`. However, I don't see > > any tests that actually check that behaviour, so if that is a mode that's > > useful to keep working we definitely have some gaps there. > > > > This change means that if you want to ask `-cc1` for dwarf *and* codeview > > with `/Z7` or `/Zi`, you'll need to specify `-gdwarf` and `-gcodeview` now. > > This matches how you would get both sets of options to render with `-g`, > > which I think makes sense. > Yeah, that sounds good. IIRC when we were using the joint codeview + DWARF > mode it was with the gcc-style Clang driver and not clang-cl, and we passed > `-gdwarf -gcodeview` or similar to enable it anyway. We don't need that > anymore though; @mstorsjo would know if there's anything on the MinGW side > that cares for it. I'm not aware of mingw usecases of using both codeview and DWARF at the same time - I've occasionally mentioned that I know some people do use that, but I haven't actually guided anybody into doing it. For mingw, as long as `-g -gcodeview` on the driver level generates codeview (and not DWARF) like it used to, mingw should be fine. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D157794/new/ https://reviews.llvm.org/D157794 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits