[PATCH] D53457: clang-cl: Add "/Xdriver:" pass-through arg support.

Reid Kleckner via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon Oct 22 10:37:07 PDT 2018


rnk added a comment.

In https://reviews.llvm.org/D53457#1269769, @hans wrote:

> I haven't started looking at the code yet.
>
> I'm not completely convinced that we want this. So far we've used the strategy of promoting clang options that are also useful in clang-cl to core options, and if someone wants to use more clang than that, maybe clang-cl isn't the right driver for that use-case.
>
> But I suppose an argument could be made for having an escape hatch from clang-cl if it doesn't add too much complexity to the code.


Personally, I really want to get out of the role of being a gatekeeper for GCC-style driver flags. I want users to be able to solve their own problems without building clang-cl from scratch or waiting for the next 6 month release.

I'm personally in favor of moving clang-cl to a blacklist model instead of a whitelist model so that we go even further in this direction, but it doesn't address what to do about conflicting or ambiguous options like `-MD`. An escape hatch like `[-/]Xdriver:` seems like a useful thing to have when that arises.

> I'm not sure I'm a fan of calling it /Xdriver: though, because what does it mean - clang-cl is the driver, but the option refers to the clang driver. The natural name would of course be -Xclang but that already means something else.  Maybe we could just call it /clang:

Huh, I like that. You'd also be able to spell it `-clang:`, so the leading `/` is less of a disambiguator, but the colon is clearly a sign that this is a CL-style option.


Repository:
  rC Clang

https://reviews.llvm.org/D53457





More information about the cfe-commits mailing list