On Mon, May 13, 2013 at 6:22 PM, Richard Trieu <[email protected]> wrote:

>
>
> ================
> Comment at: lib/AST/ASTContext.cpp:1358
> @@ +1357,3 @@
> +  unsigned Align = EltInfo.second / CharWidth;
> +  Width = llvm::RoundUpToAlignment(Width, Align);
> +  return std::make_pair(CharUnits::fromQuantity(Width),
> ----------------
> Richard Smith wrote:
> > This shouldn't be necessary; EltInfo.first / CharWidth should have been
> a multiple of EltInfo.second / CharWidth. Maybe convert this to an
> assertion?
> This was added in revision 130242
> http://llvm.org/viewvc/llvm-project?view=revision&revision=130242
>
> in response to PR5637
> http://llvm.org/bugs/show_bug.cgi?id=5637
>
> where Clang was changed to follow the behavior of GCC.  Otherwise,
> omitting this line will cause test/Sema/align-x86.c and
> test/Sema/align-x86-64.c to fail.


Wow. That's impressively broken. OK then!
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to