MaskRay added inline comments.
================ Comment at: clang/test/Driver/darwin-version.c:217 // RUN: FileCheck --check-prefix=CHECK-VERSION-TNO-OSV1 %s -// CHECK-VERSION-TNO-OSV1: overriding '-mmacos-version-min=10.6' option with '-target x86_64-apple-macos10.11.2' +// CHECK-VERSION-TNO-OSV1: overriding '-mmacos-version-min=10.6' option with '-target x86_64-apple-macos10.11.2' [-Woverriding-t-option] ---------------- hans wrote: > dexonsmith wrote: > > MaskRay wrote: > > > dexonsmith wrote: > > > > Why would we want to use the old name here? An alias seems strictly > > > > better to me. > > > Making `overriding-t-option` an alias for `overriding-option` would make > > > `-Wno-overriding-t-option` applies to future overriding option > > > diagnostics, which is exactly what I want to avoid. > > > > > I understand that you don't want `-t-` to apply to work on future > > overriding option diagnostics, but I think the platform divergence you're > > adding here is worse. > > > > Having a few Darwin-specific options use `-Woverriding-t-option` (and > > everything else use `-Woverriding-option`) as the canonical spelling is > > hard to reason about for maintainers, and for users. > > > > And might not users on other platforms have `-Woverriding-t-option` > > hardcoded in build settings? (So @dblaikie's argument that we shouldn't > > arbitrarily make things hard for users would apply to all options?) > > > > IMO, if we're not comfortable removing `-Woverriding-t-option` entirely, > > then it should live on as an alias (easy to reason about), not as > > canonical-in-special-cases (hard to reason about). > > IMO, if we're not comfortable removing -Woverriding-t-option entirely, then > > it should live on as an alias (easy to reason about), not as > > canonical-in-special-cases (hard to reason about). > > +1 if we can't drop the old spelling, an alias seems like the best option. Making `overriding-t-option` an alias for `overriding-option`, as I mentioned, will make `-Wno-overriding-t-option` affect new overriding-options uses. This is exactly what I want to avoid. I know there are some `-Wno-overriding-t-option` uses. Honestly, they are far fewer than other diagnostics we are introducing or changing in Clang. I can understand the argument "make -Werror users easier for this specific diagnostic" (but `-Werror` will complain about other new diagnostics), but do we really want to in the Darwin case? I think no. They can remove the version from the target triple like https://github.com/facebook/ocamlrep/blame/abc14b8aafcc6746ec37bf7bf0de24bfc58d63a0/prelude/apple/apple_target_sdk_version.bzl#L50 or make the version part consistent with `-m.*os-version-min`. This change may force these users to re-think how they should fix their build. I apology to these users, but I don't feel that adding an alias is really necessary. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D158301/new/ https://reviews.llvm.org/D158301 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
