[Openais] Re: possible leak?

Mark Haverkamp markh at osdl.org
Thu Sep 23 08:58:46 PDT 2004


On Wed, 2004-09-22 at 18:17, Steven Dake wrote:
> Mark
> I was finishing up the expiration and duration timers for the checkpoint
> service.  During that effort I found a leak in the timer list (ugh!) but
> it was easily fixed.
> 
> During my stress tests with valgrind and all test applications, I came
> across this possible leak in the event service:
> 
> ==823== 276 bytes in 1 blocks are possibly lost in loss record 10 of 15
> ==823==    at 0x1B902D9B: malloc (vg_replace_malloc.c:131)
> ==823==    by 0x804FB7D: create_channel (evt.c:449)
> ==823==    by 0x804FC7C: evt_open_channel (evt.c:498)
> ==823==    by 0x8050B46: lib_evt_open_channel (evt.c:1449)
> ==823==    by 0x8049964: poll_handler_libais_deliver (main.c:563)
> ==823==    by 0x8053B91: poll_run (aispoll.c:385)
> ==823==    by 0x804A16C: main (main.c:957)
> 
> Is there some logic to free the channel?  I had exited all of the
> applications using the channel.
> 
> In the checkpoint service, I keep a reference count on the open
> checkpoint.  When it reaches zero, I start the retention timer (which
> erases the checkpoint eventually).  Is there something like this for the
> event service channels?

Not yet.  It is on my todo list to do something like this.

Mark.

> 
> Thanks
> -steve
-- 
Mark Haverkamp <markh at osdl.org>




More information about the Openais mailing list