[00:22:46] --- haba has become available [00:44:00] --- dev-zero@jabber.org has left [00:44:42] --- reuteras has left [00:45:39] --- reuteras has become available [00:54:33] --- manfred has become available [01:49:39] --- dev-zero@jabber.org has become available [01:51:13] --- haba has left [03:24:12] --- haba has become available [04:00:24] fsx fails on current master. I suspect it's my fault. I'm investigating now. [04:53:30] --- Simon Wilkinson has left [05:49:33] --- dev-zero@jabber.org has left [05:59:57] --- meffie has become available [06:51:26] I think that may have been a false alarm. A clean install of master no longer appears to have the problem. [06:54:14] --- reuteras has left [06:57:27] --- Jeffrey Altman has left: Replaced by new connection [06:57:57] --- jaltman has left: Replaced by new connection [06:57:57] --- jaltman has become available [07:05:20] --- deason has become available [07:39:12] --- abo has left [07:39:28] --- abo has become available [07:45:30] --- jaltman has left: Replaced by new connection [07:45:32] --- jaltman has become available [07:50:45] --- stevenjenkins has left [07:53:48] --- stevenjenkins has become available [08:01:48] --- sxw has left [08:27:19] --- stevenjenkins has left [08:27:59] --- stevenjenkins has become available [09:12:01] --- haba has left [09:15:58] --- manfred has left [09:18:26] --- meffie has left [09:51:29] could I get added to RT 125507? I have some information on it [09:52:05] hang on [09:52:38] NIL is kinda like NULL but spelled badly. [09:53:21] done [09:54:00] tx [09:54:30] deason: i presume if you add things to that RT it will e-mail my co-worker (who started the thing). we're particularly interested in figuring this one out. [09:56:09] kula, yes, it will [09:59:25] excellent. [10:02:30] i see dan is scared of git [10:04:05] well, it's scary [10:39:58] --- jaltman has left: Replaced by new connection [10:39:59] --- jaltman has become available [10:40:16] --- jaltman has left: Disconnected [10:47:05] --- haba has become available [10:54:29] --- stevenjenkins has left [10:54:44] --- stevenjenkins has become available [11:13:43] andrew, what's the patch for 1)? what *should* happen is callback_rxconn should be allowed to be swapped, but never be null, probably [11:13:54] er, not what's. where's. [11:14:25] jeff is correct, but, ... [11:17:01] well, if hosts aren't supposed to be used if HOSTDELETED is set, then we appear to be wrong in quite a few places, and in theory that should be fixed [11:17:37] it should be. [11:17:40] the "fix" for 1 is in two places; for removeAddr_r, you just don't delete it or set callback_rxcon to anything; h_TossStuff_r will destroy it when all of the references go away [11:17:55] there's another instance in h_GetHost_r, which is basically the same thing [11:18:40] the one in h_GetHost_r may not actually be a problem, since I think that host is just locally allocated, and nothing else references it [11:19:36] or, no, I misspoke about the removeAddress_r case [11:19:45] what I was describing was h_GetHost_r [11:20:02] in removeAddress_r, we would only destroy callback_rxcon when we are about to create a new one to take it's place [11:20:06] otherwise you don't touch it [11:20:12] which is what i expected. [11:20:24] --- stevenjenkins has left [11:21:15] "to take its place" [11:21:21] --- stevenjenkins has become available [11:26:03] --- meffie has become available [11:29:24] also, while I was looking at these... this block of code comes up a few times, which does not make any sense to me: [11:29:32] struct client *client; /* * If rx_DestroyConnection calls h_FreeConnection we will * deadlock on the host_glock_mutex. Work around the problem * by unhooking the client from the connection before * destroying the connection. */ client = rx_GetSpecific(rxconn, rxcon_client_key); rx_SetSpecific(rxconn, rxcon_client_key, (void *)0); rx_DestroyConnection(rxconn); [11:30:41] I don't understand what that does, since rxcon_client_key doesn't have a destructor, so setting it to 0 shouldn't affect rx_DestroyConnection [11:31:49] and h_FreeConnection is not something I've heard of; I assume that was a typo, but I don't know what it intended to say [11:34:34] it's legacy; there was code which we used to call to clean up, which was hooked into .... mumble (some rx key) which let our cleanup hook be called [11:34:47] it *did* have a destructor [11:35:00] (and that was what it was called) [11:43:15] okay [12:08:01] apologies, had to change venues. [12:23:23] gerrit 716 has the HOSTDELETED check for that one instance; 717 has the "don't destroy callback_rxconn" approach [12:25:53] "i confuse myself easily" ?? dude [12:26:38] hey, I don't trust myself after looking at the host package code all morning [12:26:58] actually, i think the "real" issue is not the [12:27:01] hang on [12:27:24] having nothing to do with the host package i need another beer [12:28:06] also a better network [12:30:35] 717 is closer to what i was going to say [12:38:05] --- haba has left [12:39:04] I agree with 717, since I don't see why we'd need to destroy those conns so early, but we still should fix all of the h_Lock_r callers, right? [12:39:14] --- manfred has become available [12:39:21] --- stevenjenkins has left [12:39:47] --- stevenjenkins has become available [12:58:23] sorry, ex-ex-employer (cmu) asking me questions about osx afs client. [12:58:41] the h_Lock_r callers should probably be fixed, yes [13:55:27] --- dev-zero@jabber.org has become available [15:14:29] --- deason has left [15:40:40] --- deason has become available [16:06:07] --- Simon Wilkinson has become available [16:39:56] --- meffie has left [18:13:42] --- jaltman has become available [18:22:55] --- jaltman has left: Disconnected [19:35:52] --- jaltman has become available [20:16:22] --- jaltman has left: Disconnected [21:56:21] --- deason has left [22:07:26] --- manfred has left [23:37:54] --- jaltman has become available