Greg Ames <[EMAIL PROTECTED]> writes:
> We got a second seg fault running 2.0.24 tonight with mod_include
> instrumented. Dump 4 looks like the first two, except we have the
> "bbstr" variable set at key places. It shows that Cliff is right - we
> are using an empty brigade, and what we think is a bucket is a sentinel.
#0 0x281c2a76 in find_start_sequence (dptr=0x81f0c50, ctx=0x823700c, bb=0x81f0c2c,
do_cleanup=0xbfbf7cd8)
at mod_include.c:212
The brigade is definitely empty
(gdb) p *bb
$6 = {p = 0x81f000c, list = {next = 0x81f0c30, prev = 0x81f0c30}}
but the bucket we're working with could only be the sentinel of a
brigade at 0x81f0c4c (since sentinel is at brigade+0x04).
Here is that brigade:
(gdb) p *(apr_bucket_brigade *)0x81f0c4c
$7 = {p = 0x81f000c, list = {next = 0x810a320, prev = 0x810a320}}
I don't see any references to it in the call tree, so I don't know how
dptr got set to 0x81f0c50. There are only a bazillion places in the
caller that set dptr.
---/---
[probably nothing but] In all three similar dumps, the browser is IE
5.01 or 5.5, and Accept-Language is not en (zh-cn, zh-tw, ru).
Gross!
--
Jeff Trawick | [EMAIL PROTECTED] | PGP public key at web site:
http://www.geocities.com/SiliconValley/Park/9289/
Born in Roswell... married an alien...
