Re: err_exit/err_return regression

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Oct 1, 10:32pm, Joshua Krusell wrote:
} Subject: Re: err_exit/err_return regression
}
} On 29/09/15 at 06:09P, Bart Schaefer wrote:
} > I think what this boils down to is, "retflag" needs two values to
} > distinguish an actual return from an ERR_RETURN

I'm still not entirely happy with that patch, in case anyone has any
better ideas.  I'm tempted to check cmdstack[cmdsp-1] to see whether
it is one of CS_IF or CS_ELIF, but that's abusing the prompt mechanism
for program control, which seems wrong.

} With this patch it's still failing for me, but only when running the
} snippet I posted as a script. Seems to magically work with the sourced
} 'cut-and-paste' version.

The following seems like an obvious former thinko, in retrospect.  All
the regular tests still pass.  Anyone see a problem?

diff --git a/Src/exec.c b/Src/exec.c
index 154bbb8..235faf3 100644
--- a/Src/exec.c
+++ b/Src/exec.c
@@ -1364,7 +1364,8 @@ sublist_done:
 	 * we hit execcmd on the way down.  We're now
 	 * on the way back up, so don't restore it.
 	 */
-	noerrexit = (oldnoerrexit == 2) ? 0 : oldnoerrexit;
+	if (oldnoerrexit != 2)
+	    noerrexit = oldnoerrexit;
 
 	if (sigtrapped[SIGDEBUG] && !isset(DEBUGBEFORECMD) && !donedebug) {
 	    /*