Re: [bug-gawk] Behavior of fflush with SIGPIPE on stdout [PATCH]

[FERRIEUX Alexandre - IMT/OLN]

On 25/03/2017 20:21, Andrew J. Schorr wrote:
> On Sat, Mar 25, 2017 at 08:16:08PM +0100, FERRIEUX Alexandre - IMT/OLN wrote:
> >  OK, so, on behalf of the worst behavior (of a few non-posix oddball
> >  OSes), you'd rather sacrifice the unix instance of ... awk ? All of
> >  this for a "hard to maintain" 3-liner
> ..
> >  # define SIGPIPE_EXIT_CODE   1 /* my OS is too dumb to tolerate more than one 
> > bit here *:
> ...
> >  Seriously...
>
> No need to get snippy. We are discussing the best way to resolve this issue.
> You obviously feel very strongly that the fate of humanity depends on the 141
> exit code from a broken pipe. I personally never check the exit code other than
> for success (zero) vs failure (non-zero).

Apologies for the tone. It's just that taking Windows as a limiting factor gets 
me nervous.

Of course the 141 vs 1 is no big deal. But if it comes for free, what about it ?

By "for free" I mean an #ifdef appropriate to cover all of true unix. Something 
like

        #ifdef unix
        # define SIGPIPE_EXIT_CODE (128|SIGPIPE)
        #else
        # define SIGPIPE_EXIT_CODE 1
        #endif