From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <zsh-workers-return-7349-mason-zsh=primenet.com.au@sunsite.auc.dk>
Received: (qmail 26454 invoked from network); 2 Aug 1999 14:31:23 -0000
Received: from sunsite.auc.dk (130.225.51.30)
  by ns1.primenet.com.au with SMTP; 2 Aug 1999 14:31:23 -0000
Received: (qmail 22040 invoked by alias); 2 Aug 1999 14:31:12 -0000
Mailing-List: contact zsh-workers-help@sunsite.auc.dk; run by ezmlm
Precedence: bulk
X-No-Archive: yes
X-Seq: 7349
Received: (qmail 22033 invoked from network); 2 Aug 1999 14:31:12 -0000
Date: Mon, 2 Aug 1999 16:31:10 +0200 (MET DST)
Message-Id: <199908021431.QAA13865@beta.informatik.hu-berlin.de>
From: Sven Wischnowsky <wischnow@informatik.hu-berlin.de>
To: zsh-workers@sunsite.auc.dk
In-reply-to: "Bart Schaefer"'s message of Sun, 1 Aug 1999 07:39:19 +0000
Subject: Re: Bug: using send-break at "select" prompt breaks a bit too much


Bart Schaefer wrote:

> Check this out:
> 
> function oops() {
>    setopt localoptions localtraps
>    trap "echo got signal" 0 1 2 3 15 ZERR DEBUG
>    select x in a b c
>    do echo $REPLY
>    done
>    echo "finished select"
> }
> 
> Put that in your $fpath and run "oops".  Try ^C and ^\ (or ^_ or whatever
> you have sending QUIT).  Then use ^G (send-break).  Note that the whole
> function is killed, but "finished select" is never printed nor are any
> traps triggered.

The patch below makes errflag be restored after the call to zleread()
so that only the select is left.

I don't get a trap for QUIT, either (and I haven't found the place
where this is inhibited), however, I get the traps for SIGINT and
DEBUG.

But while playing with this I found a more serious bug:

  % TRAPQUIT() { echo quit }
  % foo() { setopt localtraps; trap 'echo foo quit' 3 }
  % foo

... and bang! dosavetrap() called shfunctab->removenode() which called 
removeshfuncnode() which called dosavetrap again -- removing the list
for the trap we want to get at the first call to dosavetrap(). We
ended up with a savetrap struct with the ZSIG_FUNC flag and a NULL
list pointer which gave us a SEGV in endtrapscope().

Bye
 Sven

diff -u os/loop.c Src/loop.c
--- os/loop.c	Mon Aug  2 11:44:46 1999
+++ Src/loop.c	Mon Aug  2 15:51:11 1999
@@ -178,8 +178,13 @@
 	for (;;) {
 	    if (empty(bufstack)) {
 	    	if (interact && SHTTY != -1 && isset(USEZLE)) {
+		    int oef = errflag;
+
 		    isfirstln = 1;
 		    str = (char *)zleread(prompt3, NULL, 0);
+		    if (errflag)
+			str = NULL;
+		    errflag = oef;
 	    	} else {
 		    str = promptexpand(prompt3, 0, NULL, NULL);
 		    zputs(str, stderr);
diff -u os/signals.c Src/signals.c
--- os/signals.c	Mon Aug  2 11:44:47 1999
+++ Src/signals.c	Mon Aug  2 16:22:38 1999
@@ -640,7 +640,11 @@
 	 */
 	char func[20];
 	sprintf(func, "TRAP%s", sigs[sig]);
-	st->list = shfunctab->removenode(shfunctab, func);
+	/* We call removehashnode() directly because otherwise
+	 * removeshfuncnode() would be called which in turn would
+	 * call us again so that we would end up with a NULL pointer
+	 * instead of the list for the trap. */
+	st->list = removehashnode(shfunctab, func);
     } else {
 	st->list = sigfuncs[sig];
 	unsettrap(sig);

--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de