[Serdev] Question about memory stats

[Andreas Granig]

Jan Janak wrote:
>   No, they are sorted according to the starting address. Maybe I could
>   help you if you post the whole log.

I found the problem. It's because of nested mysql queries.

I created two different connections because I already suspected that 
there could be a problem using the same connection.

But although I have two different connection pointers obtained by 
dbf.init(db_url.s), they both share the same underlying connection (the 
tail-pointer in db_con_t) which is rather wired IMHO. Now when the inner 
query is freed, the res-pointer of the outer result is set to NULL by 
the inner one, so when freeing the outer result, mysql_free_result() is 
called with a NULL pointer, and that's why the leak doesn't show up in 
the memory dump.

The attached little diff which disables the mysql pool solves the issue, 
  but I think the pool handling should really be reviewed...

Andy
-------------- next part --------------
Index: modules/mysql/my_pool.c
===================================================================
RCS file: /cvsroot/ser/sip_router/modules/mysql/Attic/my_pool.c,v
retrieving revision 1.1
diff -u -r1.1 my_pool.c
--- modules/mysql/my_pool.c	26 Jan 2004 12:46:15 -0000	1.1
+++ modules/mysql/my_pool.c	1 Sep 2005 14:51:22 -0000
@@ -69,6 +69,7 @@
 	id = new_my_id(url);
 	if (!id) return 0;
 
+#if 0
 	ptr = pool;
 	while (ptr) {
 		if (cmp_my_id(id, ptr->id)) {
@@ -79,6 +80,7 @@
 		}
 		ptr = ptr->next;
 	}
+#endif
 
 	DBG("get_connection(): Connection not found in the pool\n");
 	ptr = new_connection(id);