[00:50:57] --- Russ has left: Disconnected [02:22:24] --- haba has become available [03:33:55] --- lars.malinowsky has become available [03:45:10] --- reuteras has become available [04:53:28] --- haba has left [04:53:43] --- haba has become available [05:21:17] --- haba has left [06:32:46] --- ksumner has become available [06:39:32] --- pod has become available [07:23:28] --- deason has become available [08:13:50] --- Simon has become available [08:20:30] --- reuteras has left [08:23:49] --- Simon has left [09:56:46] --- Russ has become available [10:08:21] If we want to support Fedora 15 when it is released, we need to do a 1.4.14.1 very very soon. [10:48:03] --- Russ has left: Disconnected [11:50:09] hmm, so on an aix box, aklog with -noprdb aborts for me, because the new token mechanisms try to xdr-encode stuff, without anything initializing rx, so osi_malloc_mutex has not been initialized [11:50:30] at least, I think that's what it is; I get a failed assert on trying to lock that mutex, and it makes sense since I don't see where it would have been initialized [11:50:40] That sounds logical [11:50:59] I'm not sure where we'd want the initialization to take place... is calling pthread_once stuff too much to call for each alloc/free? [11:51:12] or should I just force something to rx_Init() before that happens? [11:51:29] Can we kill the malloc mutex and use atomics instead? [11:52:07] Although then we get into the "where do we init the atomic mutex" problem... [11:53:48] yes, currently the atomic mutex is initialized in the same place the current malloc mutex is, so.... [11:54:56] I would be very tempted to just remove the osi_alloccnt, and the osi_allocsize counters. In this day and age, if you want malloc debugging/accounting, we should just LD_PRELOAD a malloc library that gives us that. [11:57:37] yeah, I was going to suggest that, too... nothing uses the counters unless you're looking at them in a core or something [11:58:07] Yeah. There's no reference to those symbols in the code. And there are better was of finding out things like the processes current heap size. [11:58:14] s/was/ways/ [12:35:01] --- mfelliott has left [14:08:33] --- mfelliott has become available [15:25:38] --- deason has left [15:34:07] --- pod has left [15:46:50] --- steven.jenkins has left [16:39:14] --- jaltman/FrogsLeap has left: Disconnected [18:05:37] --- Russ has become available [18:36:09] --- jaltman/FrogsLeap has become available [23:26:59] --- reuteras has become available [23:34:59] --- Russ has left: Disconnected