Re: fixing non-NS darwin (was: Re: your emacs/src/keyboard.h change)

[YAMAMOTO Mitsuharu]

>>>>> On Mon, 4 Aug 2008 12:11:31 -0400, Adrian Robert <address@hidden> said:

> I've also gone through the removal diff and wanted to ask about the  
> bit below since it may be something that all terms should do:

Well, careful (as opposed to casual) readers would pay attention to
such a comment (not written by me, IIRC).

A related discussion is found at:

  http://lists.gnu.org/archive/html/emacs-pretest-bug/2004-04/msg00141.html

Still I can't clearly explain why it is ok to do a longjmp and not
execute the call to set_alarm from alarm_signal_handler.  But this
change does fix a particular problem and at least does not make things
worse, and it has been working well with Carbon Emacs 22 for a long
time.

                                     YAMAMOTO Mitsuharu
                                address@hidden

> Index: atimer.c
> ===================================================================
> RCS file: /sources/emacs/emacs/src/atimer.c,v
> retrieving revision 1.28
> retrieving revision 1.27
> diff -u -r1.28 -r1.27
> --- atimer.c  27 Jul 2008 18:24:40 -0000      1.28
> +++ atimer.c  14 May 2008 07:49:08 -0000      1.27
> @@ -368,7 +368,9 @@
>
>         t = atimers;
>         atimers = atimers->next;
> +#ifndef MAC_OSX
>         t-> fn (t);
> +#endif
>
>         if (t->type == ATIMER_CONTINUOUS)
>       {
> @@ -380,6 +382,10 @@
>         t->next = free_atimers;
>         free_atimers = t;
>       }
> +#ifdef MAC_OSX
> +      /* Fix for Ctrl-G.  Perhaps this should apply to all platforms.  
> */
> +      t->fn (t);
> +#endif
>
>         EMACS_GET_TIME (now);
>       }