nonlocal goto patch

[Janis Johnson]

Andrew MacLeod responded to my mail:

>> Nothing ever happened with my patch from May for builtin_longjmp on
>> ia64.  Unless you're already doing something with it, I'll update that
>> patch to work with your new nonlocal goto fix and submit it again.
>
> Yeah, I know. The problem was that we were seeing similar problems
> on a couple of other ports, so instead of patching each port,
> this should remedy the problem at it source.
>
> This incorporates the parts of your patch that were needed. With
> the correction to the parameters of gen_nonlocal_goto () in
> expand_builtin_longjmp, we shouldn't need the builtin_longjmp
> pattern anymore, or any of the other changes.
>
> You can verify that for me tho. :-)

I haven't been able to build on ia64-linux to verify it, but I looked
through your patch more carefully and it does appear to fix the nonlocal
goto problems on ia64.  Thanks!

Janis