Re: gawk: other double free(_wstr)

[Karel Zak]

On Sat, Jan 13, 2007 at 03:17:20PM -0500, Andrew J. Schorr wrote:
> On Sat, Jan 13, 2007 at 08:52:30PM +0200, Aharon Robbins wrote:
> > I think I'm going to undo the part of free_wstr that only zeros the
> > fields if the flag is set.
> 
> Hmmm, this code that you plan to restore (zeroing wstptr even if the
> WSTRCUR flag is not set) seems to conflict with a statement you made
> back in July:

 I agree with Aharon. It's more robust. There are places in code which
 expect this behavior.

> 
> On Tue, Jul 18, 2006 at 10:22:41PM +0300, Aharon Robbins wrote:
> > I have not yet read all this thread. The intent is that n->wstptr is NOT
> > valid if the WSTRCUR bit is clear. Or to flip it around, only if the bit
> > is set should you mess with n->wstptr.
> 
> That statement makes a lot of sense to me.  The NODE is a big messy union; can

 Yes, the NODE is nothing nice and it should be splited to more
 structs, but it's not trivial task.

> you be certain that zeroing wstptr won't stomp on some other meaningful part 
> of
> the union?

 Yes, but the NODE.type is the right way how gawk should be check type
 of the node. I think the others ways are workarounds.

 Suggestion:

--- node.c.kzak 2007-01-15 12:53:07.000000000 +0100
+++ node.c      2007-01-15 12:53:39.000000000 +0100
@@ -778,6 +778,8 @@
 void
 free_wstr(NODE *n)
 {
+       assert(n->type == Node_val);
+
        if ((n->flags & WSTRCUR) != 0) {
                assert(n->wstptr != NULL);
                free(n->wstptr);



-- 
 Karel Zak  <address@hidden>