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

[Eli Zaretskii]

> Date: Sat, 8 Apr 2017 20:03:00 +0200
> From: <address@hidden>
> CC: <address@hidden>, <address@hidden>
> 
> As I said, you're the Windows expert, but I still fail to see what reasoning 
> can bring the developer, upon discovering 
> that "invalid handle" exit status, to the proper thinking that it is a valid 
> handle to a pipe, with the other side closed.

They will not know that this was caused by a broken pipe, but they
will know (a) that the program exited abnormally, and (b) that the
abnormal exit was caused by some handle becoming invalid.  That's
close enough, and I just didn't find any better way.

> That said, I've seen so many Windows programs report "Unknown Error" that I 
> guess this lack of rigor in error reporting 
> is part of the culture. I won't sweat for Windows.

It's not due to lack of rigor, it's due to the difficulty of mapping
Posix-centric concepts (such as signals) into the equivalent Windows
concepts.