Patch for PR c/11234

[Zack Weinberg]

"Joseph S. Myers" <jsm@polyomino.org.uk> writes:

> On Thu, 8 Jan 2004, Zack Weinberg wrote:
>
>> The changes to except.c are OK.  I would prefer that you change the
>> output field of struct insn_data to a union, and restructure
>
> Without C99 designated initializers you can only initialize the first
> element of a union (so at least the stage1 compiler would need to
> initialize these fields at runtime).

You could make it a struct if the host compiler doesn't support
designated initializers, and a union if it does - just an #ifdef in
recog.h and some tweaks to the generated code; users wouldn't even
notice.  The table in insn-output.o would be bigger in stage1 but
that's harmless.

>> weak-[67].c so that foo doesn't need to return a pointer to itself.
>
> The testcases don't have comments saying exactly what they are meant to be
> testing for and why foo returns a pointer to itself, but I'd be wary of
> whether such a change would change the effect of the testcases.

Fair enough, then check in this part of the patch as is.

zw