On Thu, 2016-07-14 at 16:55 +0200, Jakub Jelinek wrote:
> Hi!
>
> As mentioned in the PR, this testcase behaves differently on
> powerpc*/spu/s390* targets, because in non-iso modes they predefine
> conditional macros like bool and that affects the behavior of the
> suggestions.
>
> We should certainly discuss what to do with the conditional macros,
> but in the mean time IMHO it doesn't hurt to just avoid this
> problem.
>
> Tested on x86_64-linux and i686-linux, ok for trunk?
Thanks. I think I can approve this with my "libcpp/diagnostic messages
" maintainer hat on.
Did you test the case with a build for one of the affected targets to
verify the fix?
> 2016-07-14 Jakub Jelinek <ja...@redhat.com>
>
> PR testsuite/71865
> * gcc.dg/diagnostic-token-ranges.c: Add -std=c11 to dg-options.
> (wide_string_literal_in_asm): Use __asm instead of asm, adjust
> expected diagnostics.
>
> --- gcc/testsuite/gcc.dg/diagnostic-token-ranges.c.jj 2016-07
> -14 11:06:23.000000000 +0200
> +++ gcc/testsuite/gcc.dg/diagnostic-token-ranges.c 2016-07-14
> 14:47:50.399907380 +0200
> @@ -1,4 +1,4 @@
> -/* { dg-options "-fdiagnostics-show-caret -Wc++-compat" } */
> +/* { dg-options "-fdiagnostics-show-caret -Wc++-compat -std=c11" }
> */
>
> /* Verify that various diagnostics show source code ranges. */
>
> @@ -68,11 +68,11 @@ foo (unknown_type param); /* { dg-error
>
> void wide_string_literal_in_asm (void)
> {
> - asm (L"nop"); /* { dg-error "wide string literal in 'asm'" } */
> + __asm (L"nop"); /* { dg-error "wide string literal in 'asm'" } */
> /*
> { dg-begin-multiline-output "" }
> - asm (L"nop");
> - ^~~~~~
> + __asm (L"nop");
> + ^~~~~~
> { dg-end-multiline-output "" }
> */
> }
>
> Jakub
