# HG changeset patch
# User Luke Schierer <lschiere@pidgin.im>
# Date 1053953105 0
# Node ID 10920646988f0e53f0b5194cc91d4857d676d5d5
# Parent  2c4c975620f000f0a4d3f9186e822aa76431008b
[gaim-migrate @ 5931]
Ambrose LI (acli) writes:
" This is w.r.t. the 0.64cvs as of around 20:25 EDT.

When gaim starts up, it displays the signon window, and
then immeidately segfaults. The problem is that
tobase64() in util.c uses a size_t (which could very
likely be unsigned) as a second parameter, but proxy.c
calls it with a -1. "

at my questioning the specific change he made:
"Actually the initialization is not the matter (just
something I did while debugging and did not remove, since it
makes sense in any case), it's the other change that matters.

In util.h, the javadoc-like comments states that passing len
= -1 means that tobase64() should calculated len =
strlen(buf). However, since size_t is (or can be) unsigned,
the "if (len < 0)" check never succeeds, so passing
-1
actually results in a very large integer being passed.

So the "if (len < 0)" check actually need to be
changed so
that the passed "-1" will actually get recognized."

so since most of us aren't segfaulting on start, i'm guessing this is a
platform/os specific problem, but that the change shouldn't affect the
rest of us anyway.

committer: Tailor Script <tailor@pidgin.im>

diff -r 2c4c975620f0 -r 10920646988f src/util.c
--- a/src/util.c	Mon May 26 08:30:48 2003 +0000
+++ b/src/util.c	Mon May 26 12:45:05 2003 +0000
@@ -353,10 +353,10 @@
 /* This was borrowed from the Kame source, and then tweaked to our needs */
 char *tobase64(const unsigned char *buf, size_t len)
 {
-	char *s, *rv;
+	char *s = NULL, *rv = NULL;
 	unsigned tmp;
 
-	if(len < 0)
+	if(len == (size_t)(-1))
 		len = strlen(buf);
 
 	s = g_malloc((4 * (len + 1)) / 3 + 1);