httpd-apreq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joe Schaefer <>
Subject Re: 2.03-dev-rc3 (was Re: 5.6.1 segfaults during perl tests)
Date Fri, 11 Jun 2004 15:22:31 GMT
Geoffrey Young <> writes:

> (gdb) thread 3
> [Switching to thread 3 (process 9176)]#0  0x08078101 in ap_get_module_config
> (cv=0x0, m=0xbb6480) at util_debug.c:105
> 105         return ((void **)cv)[m->module_index];
> (gdb) bt
> #0  0x08078101 in ap_get_module_config (cv=0x0, m=0xbb6480) at util_debug.c:105
> #1  0x00328fcf in mpxs_Apache__RequestRec_content_type (my_perl=0x9837fd0,
> r=0x99fd468, type=0x9a52fe4)
>     at Apache__RequestRec.h:23

Clearly cv shouldn't be NULL here.  Since this stack trace looks
similar to the 5.6.1 coredumps, I'm wondering if the sv_magic fix 
I committed to rc3 broke your build.  See if any of the other 
release candidates (rc1, rc2) still segfault (with your patch applied),
or just read the sv_magic comments in glue/perl/xsbuilder/apreq_xs_postperl.h 
and replace the 5.6.1-compatible call with the original code (that is now
commented out).


> I'm not sure if it's the same issue, but we went a few rounds over in
> mod_perl land with joe orton on something similar.  the (very long)
> threads starts here:
> the most interesting part (which you probably already know) is here
> I really didn't understand most of it, but joe and stas worked
> something out for the mod_perl build system IIRC.  maybe that needs to
> happen here as well? 

Thanks- that's a very good possibility, but let's eliminate the
sv_magic() issue first.

Joe Schaefer

View raw message