[Modules] mod_gnutls: does gnutls_allocate_* the counterpart gnutls_free_*

Nikos Mavrogiannopoulos nmav at gnutls.org
Thu Nov 18 18:14:18 EST 2010


On 11/18/2010 10:20 PM, Hardy Griech wrote:
> On 18.11.2010 10:45, Nikos Mavrogiannopoulos wrote:
> :
>> Can you use valgrind to trace the leak?
> valgrind shows as the main leak:
> 
> ==23601== 19,733,881 (2,399,856 direct, 17,334,025 indirect) bytes in 
> 692 blocks are definitely lost in loss record 1,743 of 1,743
> ==23601==    at 0x482328F: calloc (vg_replace_malloc.c:467)
> ==23601==    by 0x4F1D13D: gnutls_init (in /usr/lib/libgnutls.so.26.16.10)
> ==23601==    by 0x4ED44A0: mgs_hook_pre_connection (gnutls_hooks.c:711)
> ==23601==    by 0x148E3F: ap_run_pre_connection (in 
> /usr/lib/apache2/mpm-prefork/apache2)
> ==23601==    by 0x148E93: ap_process_connection (in 
> /usr/lib/apache2/mpm-prefork/apache2)
> ==23601==    by 0x159189: ??? (in /usr/lib/apache2/mpm-prefork/apache2)
> ==23601==    by 0x1594C0: ??? (in /usr/lib/apache2/mpm-prefork/apache2)
> ==23601==    by 0x159C4E: ap_mpm_run (in 
> /usr/lib/apache2/mpm-prefork/apache2)
> ==23601==    by 0x12AFA1: main (in /usr/lib/apache2/mpm-prefork/apache2)
> Does that help?

Pretty strage... It does mean that gnutls_deinit isn't called. Are you
sure it is the reason for the leak... I mean here do you really see it
being leaked as many times as the connections you have?

regards,
Nikos


More information about the Modules mailing list