From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from groucho.cs.psu.edu ([130.203.2.10]) by hawkwind.utcs.toronto.edu with SMTP id <2223>; Sat, 15 Aug 1992 20:54:48 -0400
Received: from localhost by groucho.cs.psu.edu with SMTP id <2613>; Sat, 15 Aug 1992 20:54:25 -0400
To:	netapp!byron@netcom.com (Byron Rakitzis)
cc:	rc@hawkwind.utcs.toronto.edu
Subject: Update on memory leaks 
In-reply-to: Your message of "Sat, 15 Aug 92 01:45:03 EDT."
             <9208150545.AA17514@netapp.netapp> 
Date:	Sat, 15 Aug 1992 20:54:18 -0400
From:	Scott Schwartz <schwartz@groucho.cs.psu.edu>
Message-Id: <92Aug15.205425edt.2613@groucho.cs.psu.edu>


It looks like it is Sun's yacc's fault.  It mallocs space for the parse
stack at the top of yyparse, and frees it when YYACCEPT or YYABORT is
called.  Unfortunately, longjmp defeats that strategy.  Sun's ansi yacc
mallocs once and saves the pointer in global.  Berkeley yacc statically
allocates things.  Bison uses all sorts of techniques, but it seems to
work ok.