Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!sdd.hp.com!apollo!vinoski
From: vinoski@apollo.hp.com (Stephen Vinoski)
Newsgroups: comp.lang.perl
Subject: possible perl bug (was Re: compiling perl patch 10 on HPUX)
Message-ID: <1991Jun19.211543.17886@apollo.hp.com>
Date: 19 Jun 91 21:15:43 GMT
References: <sherman.677300831@foster>
Sender: netnews@apollo.hp.com (USENET posting account)
Organization: Hewlett-Packard Company, Apollo Division - Chelmsford, MA
Lines: 22
Nntp-Posting-Host: zep.ch.apollo.hp.com

In article <sherman.677300831@foster> sherman@unx.sas.com (Chris Sherman) writes:
>For those who have HPUX on their HP 9000/3xx or 9000/7xx, here is what I did
>to get perl to pass its tests:
>
>HP 9000/700:
>
>  o  Use perl's malloc.  Otherwise, op/pat will fail test 44-47.

Interesting.  While I was trying to track down another bug, my perl built with a
paranoid malloc on Domain/OS sr10.3.5 on a DN3500 kept failing this very test. 
If I build with perl malloc, it passes.  It fails with two different paranoid
malloc implementations.

I believe this test tickles a bug involving a bad pointer dereference.  When the
regexec() function is called from do_match() at line 197 the variable "s"
sometimes points to garbage.  Exactly why I don't know yet.


-steve

| Steve Vinoski  (508)256-0176 x5904       | Internet: vinoski@apollo.hp.com  |
| HP Apollo Division, Chelmsford, MA 01824 | UUCP: ...!apollo!vinoski         |