Re: Failed pipeline doesn't err_exit/return if complex command last

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Thu, Jul 6, 2023 at 2:44 AM Peter Stephenson
<p.w.stephenson@xxxxxxxxxxxx> wrote:
>
\> fn() {
>   emulate -L zsh
>   setopt errreturn pipefail
>   false | { true }
>   print "Shouldn't get here, status $?"
> }
>
> The final part of the pipeline is run in the current shell, so the
> job control is a bit fiddly.  It looks like the relevant code,
> including the function storepipestats(), is run too late in this
> case.

It's not that it's run too late, it's that it's run too SOON:
pipestatus has already been updated and lastval set, but then we get
here:

4186            if (!errflag) {
4187                int ret = execbuiltin(args, assigns, (Builtin) hn);
4188                /*
4189                 * In case of interruption assume builtin status
4190                 * is less useful than what interrupt set.
4191                 */
4192                if (!(errflag & ERRFLAG_INT))
4193                lastval = ret;
4194            }

and clobber lastval with the result of execbuiltin().  Or that's how
it currently appears from pass through gdb.  Per "fiddly" it's a bit
hard to be sure.