[patch][c++/PR 31743]

[Mark Mitchell]

Dave Brolley wrote:

> Yes, this works. It also removes the need to call complete_type_or_else
> at all in build_new_1. So, here's my latest patch, ChangeLog and test
> case. The patch fixes the problem and causes no regressions in the test
> suite.

Great!

This comment isn't quite accurate:

> +  /* PR 31743: Make sure the array type has a known size.  */
> +  if (!complete_type_or_else (type, NULL_TREE))
> +    return error_mark_node;

because this check isn't just about array types.  Better would be:

/* The type allocated must be complete.  If the new-type-id was "T[N]"
   then we are just checking that "T" is complete here, but that is
   equivalent, since the value of "N" doesn't matter.  */

(I'm not a fan of putting PR numbers in comments; the comment ought to
stand on itself.)

OK with that change; no need to retest.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713