case $? fails

case $? fails

[Roderick Schertler]

Running this script

    #!/bin/zsh -f

    print "version $ZSH_VERSION"

    false
    case $? in
	0)  print "case says \$? is 0";;
    esac

    false
    print "\$? was actually $?"

gives me this buggy result

    version 3.0.3-test4
    case says $? is 0
    $? was actually 1

on both freebsd and dgux.

-- 
Roderick Schertler
roderick@argon.org

Re: case $? fails

[Zoltan T. Hidvegi]

Roderick Schertler wrote:
> Running this script
>
>     #!/bin/zsh -f
>
>     print "version $ZSH_VERSION"
>
>     false
>     case $? in
>       0)  print "case says \$? is 0";;
>     esac
>
>     false
>     print "\$? was actually $?"
>
> gives me this buggy result
>
>     version 3.0.3-test4
>     case says $? is 0
>     $? was actually 1

while has the same problem unfortunately.  Similarily
for ((i=$?;;)) do echo $i; break; done
prints zero.  The patch below fixes these.  The for syntax is new in
zsh-3.1 so the first two hunks should be removed if the patch is applied
against zsh-3.0.3-test4.  Note that the case bug was not present in
zsh-3.0.2 but the while bug was there.

Zoltan


*** Src/loop.c  1997/02/11 14:23:47     3.1.2.0
--- Src/loop.c  1997/02/27 21:26:13
***************
*** 41,47 ****
      int val;
      LinkList args;

-     lastval = 0;
      node = cmd->u.forcmd;
      args = cmd->args;
      if (node->condition) {
--- 41,46 ----
***************
*** 58,63 ****
--- 57,63 ----
        for (x = pparams; *x; x++)
            addlinknode(args, ztrdup(*x));
      }
+     lastval = 0;
      loops++;
      pushheap();
      for (;;) {
***************
*** 267,277 ****

      olderrexit = noerrexit;
      node = cmd->u.whilecmd;
!     lastval = 0;
      pushheap();
      loops++;
      for (;;) {
-       oldval = lastval;
        list = (List) dupstruct(node->cont);
        noerrexit = 1;
        execlist(list, 1, 0);
--- 267,276 ----

      olderrexit = noerrexit;
      node = cmd->u.whilecmd;
!     oldval = 0;
      pushheap();
      loops++;
      for (;;) {
        list = (List) dupstruct(node->cont);
        noerrexit = 1;
        execlist(list, 1, 0);
***************
*** 295,300 ****
--- 294,300 ----
            lastval = 1;
            break;
        }
+       oldval = lastval;
      }
      popheap();
      loops--;
***************
*** 381,390 ****
      l = node->lists;
      p = node->pats;

-     lastval = 0;
      word = *p++;
      singsub(&word);
      untokenize(word);

      if (node) {
        while (*p) {
--- 381,390 ----
      l = node->lists;
      p = node->pats;

      word = *p++;
      singsub(&word);
      untokenize(word);
+     lastval = 0;

      if (node) {
        while (*p) {

Re: case $? fails

[Bart Schaefer]

On Feb 27,  4:50pm,  (Zoltan T. Hidvegi) wrote:
} Subject: Re: case $? fails
}
} prints zero.  The patch below fixes these.  The for syntax is new in
} zsh-3.1 so the first two hunks should be removed if the patch is applied
} against zsh-3.0.3-test4.

The first two hunks?  Or the second two hunks?

I've now tried this a couple of different ways and can't get it to apply
to 3.0.3-test4.  Could you send a patch specifically for that?

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

Re: case $? fails

[Zoltan T. Hidvegi]

Bart Schaefer wrote:
> } prints zero.  The patch below fixes these.  The for syntax is new in
> } zsh-3.1 so the first two hunks should be removed if the patch is applied
> } against zsh-3.0.3-test4.
>
> The first two hunks?  Or the second two hunks?
>
> I've now tried this a couple of different ways and can't get it to apply
> to 3.0.3-test4.  Could you send a patch specifically for that?

It applied cleanly for me after removing the first two hunks.  But here is
the patch for 3.0.3-test4 anyway.

Zoltan


*** loop.c      1996/12/29 01:47:11     3.0.2.1
--- loop.c      1997/02/28 16:16:48
***************
*** 168,178 ****

      olderrexit = noerrexit;
      node = cmd->u.whilecmd;
!     lastval = 0;
      pushheap();
      loops++;
      for (;;) {
-       oldval = lastval;
        list = (List) dupstruct(node->cont);
        noerrexit = 1;
        execlist(list, 1, 0);
--- 168,177 ----

      olderrexit = noerrexit;
      node = cmd->u.whilecmd;
!     oldval = 0;
      pushheap();
      loops++;
      for (;;) {
        list = (List) dupstruct(node->cont);
        noerrexit = 1;
        execlist(list, 1, 0);
***************
*** 196,201 ****
--- 195,201 ----
            lastval = 1;
            break;
        }
+       oldval = lastval;
      }
      popheap();
      loops--;
***************
*** 282,291 ****
      l = node->lists;
      p = node->pats;

-     lastval = 0;
      word = *p++;
      singsub(&word);
      untokenize(word);

      if (node) {
        while (*p) {
--- 282,291 ----
      l = node->lists;
      p = node->pats;

      word = *p++;
      singsub(&word);
      untokenize(word);
+     lastval = 0;

      if (node) {
        while (*p) {

Re: case $? fails

[Bart Schaefer]

On Feb 27,  4:50pm,  (Zoltan T. Hidvegi) wrote:
} Subject: Re: case $? fails
}
} while has the same problem unfortunately.  Similarily
} for ((i=$?;;)) do echo $i; break; done
} prints zero.  The patch below fixes these.  The for syntax is new in
} zsh-3.1 so the first two hunks should be removed if the patch is applied
} against zsh-3.0.3-test4.

Here's what I came up with as a patch for 3.0.3-test4 by manually applying
the patch from Zoltan.  I *did* include the first two hunks (with one
additional change that moved the increment of the "loops" counter, to
match the context in the patch) -- so this patch may be flawed.

Can somebody tell me if this patch is correct?  Or have I broken something
that I haven't encountered yet?

--- Src/loop.c.orig	Fri Jan 31 21:24:10 1997
+++ Src/loop.c	Fri Feb 28 08:25:09 1997
@@ -40,8 +40,6 @@
     char *str;
     LinkList args;
 
-    loops++;
-    lastval = 0;
     node = cmd->u.forcmd;
     args = cmd->args;
     if (!node->inflag) {
@@ -51,6 +49,8 @@
 	for (x = pparams; *x; x++)
 	    addlinknode(args, ztrdup(*x));
     }
+    lastval = 0;
+    loops++;
     pushheap();
     while ((str = (char *)ugetnode(args))) {
 	setsparam(node->name, ztrdup(str));
@@ -168,11 +168,10 @@
 
     olderrexit = noerrexit;
     node = cmd->u.whilecmd;
-    lastval = 0;
+    oldval = 0;
     pushheap();
     loops++;
     for (;;) {
-	oldval = lastval;
 	list = (List) dupstruct(node->cont);
 	noerrexit = 1;
 	execlist(list, 1, 0);
@@ -196,6 +195,7 @@
 	    lastval = 1;
 	    break;
 	}
+	oldval = lastval;
     }
     popheap();
     loops--;
@@ -282,10 +282,10 @@
     l = node->lists;
     p = node->pats;
 
-    lastval = 0;
     word = *p++;
     singsub(&word);
     untokenize(word);
+    lastval = 0;
 
     if (node) {
 	while (*p) {

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern

Re: case $? fails

[Bart Schaefer]

On Feb 28, 11:23am,  (Zoltan T. Hidvegi) wrote:
} Subject: Re: case $? fails
}
} Bart Schaefer wrote:
} > I've now tried this a couple of different ways and can't get it to apply
} > to 3.0.3-test4.  Could you send a patch specifically for that?
} 
} It applied cleanly for me after removing the first two hunks.  But here is
} the patch for 3.0.3-test4 anyway.

Zoltan, all the tabs in your email message have been converted to spaces.
That's why the patch failed to apply for me; this latest one fails, too.
If I re-tabify with `unexpand -a` then it applies.

I hope this doesn't have something to do with mailing from IBM's network.

If you can't figure out why this is happening, you may have to start
base64-ing or uuencoding you patches. :-(

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern