mstorsjo added a comment. In D132444#3742305 <https://reviews.llvm.org/D132444#3742305>, @mstorsjo wrote:
> In D132444#3742295 <https://reviews.llvm.org/D132444#3742295>, @thakis wrote: > >> Do we have precedent for "platform" for this? For fuse-ld=, one is supposed >> to use `-fuse-ld=` (without anything after the `=`) to get the default ld. >> That's not great (...but it can't collide with actual linker names, i >> suppose). >> >> Using "platform" (or any other self-descriptive name) for this seems easier >> to understand than passing an empty value. But it'd be nice if we could use >> this consistently in our various flags. > > Yes, `"platform"` is an existing option handled in a bunch of places already > - see e.g. > https://github.com/llvm/llvm-project/blob/main/clang/lib/Driver/ToolChain.cpp#L832-L838. > It's just that these cases hadn't been updated to take it into account. Then again, the quoted code comment says that `"platform"` only is meant to be used for overriding `CLANG_DEFAULT_RTLIB` in tests, but I don't see why one can't use it for overriding `CLANG_DEFAULT_RTLIB` (or an earlier `-rtlib` option on the command line) in real world uses too. (I had a user wanting to use my builds of clang for MSVC use cases, where it failed due to my Clang defaulting to `-rtlib=compiler-rt`, with no way of overriding it back to default, see https://github.com/mstorsjo/llvm-mingw/issues/267) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D132444/new/ https://reviews.llvm.org/D132444 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
