[00:01:28] --- kaj has become available [01:12:41] --- Russ has left: Disconnected [02:02:45] --- Simon Wilkinson has left [02:56:16] --- Simon Wilkinson has become available [03:08:30] --- dev-zero@jabber.org has become available [03:08:48] --- dev-zero@jabber.org has left [03:09:06] --- dev-zero@jabber.org has become available [04:17:18] --- Simon Wilkinson has left [04:17:18] --- Simon Wilkinson has become available [04:41:09] --- dev-zero@jabber.org has left [04:57:02] --- kaj has left [04:59:20] --- Jeffrey Altman has become available [05:23:09] --- Jeffrey Altman has left [05:51:30] --- meffie has become available [06:20:44] --- deason has become available [06:56:57] --- kaj has become available [07:05:20] --- reuteras has left [10:07:22] --- Simon Wilkinson has left [10:07:25] --- Simon Wilkinson has become available [10:32:47] --- meffie has left [10:38:51] --- kaj has left [10:41:28] --- kaj has become available [11:15:02] --- Russ has become available [12:59:56] --- jaltman has become available [13:22:43] --- meffie has become available [14:19:19] --- meffie has left [14:26:37] --- jaltman has left: Disconnected [14:36:57] --- dev-zero@jabber.org has become available [14:37:10] --- dev-zero@jabber.org has left [14:37:13] --- dev-zero@jabber.org has become available [14:58:59] --- deason has left [15:11:01] --- deason has become available [15:21:06] Oh, demand-attach still doesn't build with --enable-checking, huh? [15:21:14] * Russ changes his script to not use that for the time being. [15:21:17] Oh, I thought I'd fixed that. [15:21:20] Which platform? [15:21:33] Debian i386. [15:21:48] Hmmm. Or maybe it was fixed, and someone broke it. Lemme see. [15:21:52] Missing prototype for a pt_* function that has now scrolled off my screen. [15:21:59] 'cause I'm an idiot and didn't save the results. [15:22:24] One of my holiday projects was going to be to get buildbot going. Not sure if I'm going to make that happen in the remaining time. [15:25:46] all you have to do is slow the rotation of the earth. plenty of more time! [15:40:32] * Russ wonders if bos restricted mode is actually as weird as I think it is from looking at the code, or if something writes out the BosConfig file more frequently than I think, in which case it's just a poorly chosen interface. [15:40:49] No. it's wierd. [15:41:19] dafs should build with --enable-checking; I've been building with both options for verifying pretty much everything [15:41:29] * Russ suspects that having bos setrestricted as an RPC and representing the state in BosConfig was the core mistake, and it should be only a command-line option. [15:43:03] Yeh. bos setrestricted is wierd, in that you can only use it to set the state on. [15:44:29] The command line flag should toggle a different state flag to the RPC, if we're going to have both. And the command line's state flag shouldn't cause BosConfig to be modified, IMHO. [15:45:15] Current git master compiles fine with --enable-checking and --enable-demand-attach-fs on Mac OS, btw. [15:45:45] Okay, let me retry and see what's going on. [15:45:55] Yeah, I agree on both of those. [15:46:58] Of course, the problem with the code having been submitted aeons ago, is that it's hard to divine the original thinking behind it. [15:47:19] * Russ is inclined to go ahead and submit your patch and then deal with this in follow-up patches. [15:47:36] I think that's probably cleanest from a version control point of view. [15:47:46] Either that, or clean it up before pushing the patch that enables it. [15:48:34] Yeah, but that requires rebasing and there's a whole stack of patches. [15:49:13] I wouldn't worry about the stack unduly. They're not really dependent, so we can split them apart pretty easily if one at the beginning needs rebased. [15:49:28] (given we're cherrypicking, gerrit will let you submit out of order, anyway, iirc) [15:49:29] either way will require rebasing if you touch the same file twice, right? unless you wait for the enabled patch to go all the way to be cherry-picked [15:50:05] ChangeIds pretty much remove the pain of dealing with debased commits. [15:50:11] s/debased/rebased/ [15:52:12] deason: Well, I was going to submit the enabling patch now and then go poke at fixing the code. [15:53:10] ptutils.c:87: error: function declaration isn’t a prototype [-Wstrict-prototypes] ptutils.c:167: error: function declaration isn’t a prototype [-Wstrict-prototypes] ptutils.c: In function ‘pt_hook_write’: ptutils.c:167: error: old-style function definition [-Wold-style-definition] ptutils.c:169: error: comparison of distinct pointer types lacks a cast ptutils.c:171: error: assignment from incompatible pointer type ptutils.c: In function ‘ChangeEntry’: ptutils.c:1860: error: unused variable ‘centry’ [-Wunused-variable] ptutils.c:1856: error: unused variable ‘nptr’ [-Wunused-variable] [15:53:40] I suspect this actually has nothing to do with demand-attach, gomen. [15:53:59] It's supergroups. [15:54:05] Aha. Yes. [15:54:09] Sorry for the misaccusation. [15:54:15] Which are almost certainly broken on 64bit, btw. [15:55:25] If you want to blow your mind, look at the pointer manipulation voodoo that takes place in map.c [16:09:16] --- dev-zero@jabber.org has left [16:16:47] Wow, some of this code is... special. [16:17:03] Which code? restricted mode? [16:17:28] I particularly liked the bit where the structure changed and rather than change the code that used a variable that became a structure, it just wraps the old code in a #define that defines the old variable name to the new structure accessor, shadowing the variable declaration and making it unused. [16:17:30] Supergroups. [16:17:47] Oh, supergroups is just really really scary. [16:18:09] So far, the required fixes to make it build without warnings are actually really straightforward, though. [16:18:21] Which probably implies that the compiler needs a better sense of taste. :) [16:18:23] ... until you move to a 64bit platform ... [16:18:47] Well, yeah. [16:18:49] But one step at a time. [16:21:03] This all happened because I wrote a script to verify a Gerrit patch and decided to pass in all my standard configure options. [16:21:26] Sorry. [16:23:24] * Russ hehs. Not a problem. I wanted to do some AFS work during vacation. [16:45:17] I'm surprised map.c compiles anywhere other than on the BSDs, given that it depends on STDLIB_HAS_MALLOC_PROTOS which seems to only be defined for those platforms ... [16:48:34] Linux provides malloc.h. [16:48:48] --- jaltman has become available [16:49:21] I agree that assuming such a thing exists is not a good move. :) [18:01:57] --- jaltman has left: Disconnected [20:09:01] oh hey, the recent comments on simon's RT work reminded me; I can provide a list of a few tickets I think can be closed... should I give that to someone? [20:09:34] or I could poke the individual tickets... but at least a couple I think already have me poking for close [20:27:35] Yeah, that would be great! And I don't see any reason not to use openafs-devel for that purpose, unless someone else thinks that's a bad idea. [22:30:47] --- jaltman has become available [22:31:32] I think posting the list of tickets to this forum would be sufficient [22:48:19] Yeah, that would work too. [23:23:10] --- dev-zero@jabber.org has become available [23:23:26] --- dev-zero@jabber.org has left [23:23:39] --- dev-zero@jabber.org has become available [23:28:28] --- deason has left