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

[Eli Zaretskii]

> Date: Sun, 9 Apr 2017 11:32:00 +0200
> From: <address@hidden>
> CC: <address@hidden>, <address@hidden>
> 
> > I'm drawing a line where the EXCEPTION_* values defined in winbase.h
> > header end.  All the other values are AFAIK never triggered by fatal
> > exceptions that cause abnormal program termination.
> 
> Yet the values live in the same namespace (in ntstatus.h) so are guaranteed 
> no to collide.

Yes, the values of EXCEPTION_* are identical to the corresponding
STATUS_* values in ntstatus.h.  But that doesn't mean they are the
same and will be seen in the same situations.  The EXCEPTION_* values
are a small subset of the NTSTATUS values, and AFAIK only that subset
is generated by fatal exceptions that terminate programs.

> Anybody looking 0xC000014B up with a header grep or a google search will land 
> on "pipe broken".

As I explained earlier, the issue is not with grepping Windows headers
(which end-users likely won't have installed), the issue is with other
programs that interpret these exit codes when invoking Gawk as a
sub-process.  None of them know about the value you propose, while the
value I used _is_ known.

> So, "invalid handle" is simply not applicable at all.

I can only say that I disagree, and I explained why.