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

[Andrew J. Schorr]

Hi,

On Mon, Mar 27, 2017 at 12:50:07PM -0600, address@hidden wrote:
> I think I have it all straightened out now.  If not, it's good enough
> until some future date or someone else wants to tackle it.

Wow, that's almost as sleazy as something I would do. :-)
#define die_via_sigpipe() (signal(SIGPIPE, SIG_DFL), kill(getpid(), SIGPIPE))

I think you have a trivial bug in main.c:usage where it says:

        if (ferror(fp)) {
                /* don't warn about stdout/stderr if EPIPE, but do error exit */
                if (errno != EPIPE) {
                        if (fp == stdout)
                                warning(_("error writing standard output 
(%s)"), strerror(errno));
                        else if (fp == stderr)
                                warning(_("error writing standard error (%s)"), 
strerror(errno));
                } else if (errno == SIGPIPE)
                        die_via_sigpipe();

                // some other problem than SIGPIPE
                exit(EXIT_FAILURE);
        }

I think that "if (errno == SIGPIPE)" test should be removed.

Regards,
Andy