Date: Fri, 23 Apr 2004 12:02:22 +0200
From: Alex Kiesel <alex.kiesel@document-root.de>
To: htdig-general@lists.sourceforge.net
Subject: Re: [htdig] 'htfuzzy soundex' segfaulting

On Thu, 2004-04-22 at 19:02, Alex Kiesel wrote:
> htfuzzy soundex then segfaults. I ran it with strace:
> $ strace htfuzzy -v soundex -c /etc/htdig/htdig.conf

I've produced a more informational backtrace by compiling it from source
and trying again. The problem occurs when one has activated the option

allow_numbers:		true

and tried to run 'htfuzzy soundex' on the database files. As numbers
from a string are being skipped in Soundex::generateKey(), a segfault
occurs as the char *word is being increased by

while (!isalpha(*word))
  word++;

if the given word contains numbers only (or: contains only nonalpha
chars, to be precise).

The attached diff[1] fixes this problem for me, it's a diff against
htdig 3.1.6, file htdig-3.1.6/htfuzzy/Soundex.cc

Regards,
-Alex

[1] Also at:
http://document-root.de/patches/htdig-htfuzzy-segfault-20040423.diff


--- Soundex.cc.orig	Fri Apr 23 11:42:11 2004
+++ Soundex.cc	Fri Apr 23 11:45:53 2004
@@ -46,10 +46,11 @@
 	return;
       }
 
-    while (!isalpha(*word))
+    while (!isalpha(*word) && *word != '\0') {
       word++;
+    }
 
-    if (word)
+    if (*word)
     {
 	key << *word++;
     }