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

[Peter Stephenson <p.w.stephenson@xxxxxxxxxxxx>]

> fn () {
>   emulate -L zsh
>   setopt errreturn pipefail
>   false | {
>     true
>   }
>   print "Shouldn't get here, status $?"
> }

A bit more prodding reveals it's not actually timing that
has this effect --- the problem is this_noerrexit being
set to indicate something like "we're doing that complicated
internal shell stuff and we've discovered the status is OK".
This time it isn['t.

Not sure if I need the second case or not, certainly
the first case is the one we're hitting here (the
STAT_CURSH is the giveaway).  My guess is that it's
simply irrelevant because if there's no STAT_CURSH
we aren't going to hit those cases where this_noerrexit
becomes non-zero.

Obviously, there could be even more complicated cases that
still cause problems, but I think we'll just have to discover
those.  I'll write some tests that might help.

Note I fixed some indentation.

pws

diff --git a/Src/jobs.c b/Src/jobs.c
index dd7bba405..a3b9f667a 100644
--- a/Src/jobs.c
+++ b/Src/jobs.c
@@ -427,11 +427,17 @@ storepipestats(Job jn, int inforeground, int fixlastval)
     }
 
     if (fixlastval) {
-      if (jn->stat & STAT_CURSH) {
-	if (!lastval && isset(PIPEFAIL))
-	  lastval = pipefail;
-      } else if (isset(PIPEFAIL))
-	lastval = pipefail;
+	if (jn->stat & STAT_CURSH) {
+	    if (!lastval && isset(PIPEFAIL)) {
+		if (inforeground)
+		    this_noerrexit = 0;
+		lastval = pipefail;
+	    }
+	} else if (isset(PIPEFAIL)) {
+	    if (inforeground)
+		this_noerrexit = 0;
+	    lastval = pipefail;
+	}
     }
 }