[00:23:33] --- abo has become available [00:26:11] --- Simon Wilkinson has become available [02:26:30] --- abo has left [03:16:04] --- dev-zero@jabber.org has left: Replaced by new connection [03:16:08] --- dev-zero@jabber.org has become available [03:26:46] --- abo has become available [03:58:34] --- stevenjenkins has left [03:58:51] --- stevenjenkins has become available [04:02:40] --- reuteras has become available [04:11:13] --- abo has left [05:34:26] --- abo has become available [05:58:08] --- sxw has become available [06:13:40] --- SecureEndpoints has become available [06:29:05] rats. Darwin 8 doesn't have vprintf in the kernel - /afs/inf.ed.ac.uk/user/s/sxw/Public/openafs-darwin8-vprintf.patch RT #124167 [06:32:54] done [06:33:35] shall I close that bug? [06:34:04] --- matt has become available [06:35:03] sure [07:03:40] --- dmontuori has become available [07:45:15] Bug #124116 fixes our compilation on OpenSolaris 2008.11, but it does so by renaming struct conn *, to afs_conn_t. [07:45:30] It would seem more consistent, stylistically, to use afs_conn. [07:45:42] Anyone have any thoughts? Anyone care? [07:45:56] (sorry that should be 'to use struct afs_conn' ) [07:46:03] struct afs_conn == afs_conn_t in my mind [07:46:59] My issue was that we generally use the '_t' to be a more abstract type, whereas 'struct' gets used when somethings clearly one type. I guess I don't care that much. I just need that code to test the inode cleanup on Solaris. [07:47:06] We don't currently use the _t convention, and there is no good reason to introduce it. [07:47:08] So, I might end up porting it to 1.5 myself. [07:47:17] struct conn should be struct afs_conn, and we could opt to also provide afs_conn_t as a typedef of struct afs_conn. [07:47:22] do we? [07:48:07] Oh, hm, it seems we do. In which case ==shadow, since that's the convention. [07:49:09] But where we're currently using struct conn, we should use struct afs_conn - we shouldn't switch to afs_conn_t, is my argument. [07:49:17] I _like_ knowing when things a structures :) [07:49:28] agree [07:49:53] And further, yeah, that -- we don't seem to gratuitously define types for every struct, at least within the CM code. [07:50:53] well, the ones which have "common" names are an ongoing risk. [08:01:40] sxw: I would prefer "struct afs_conn" and to disregard afs_conn_t [08:02:35] (seems like that was already the consensus, sorry to spam) [08:11:57] --- reuteras has left [08:25:49] Okay. I'm going to put sed to work and see what falls out the other side ... [08:38:48] simon, fyi. Back at school, and can haven an AIX box up and ready soon, if you want me to test that patch [08:39:21] summatusmentis: Cool. Actually have two for you to take a look at. [08:42:13] struct conn rename : /afs/inf.ed.ac.uk/user/s/sxw/Public/openafs-rename-struct-conn.patch [08:42:27] (also does kdump.c, which I think was missing from Dale's original patch) [08:42:46] I've tested on Darwin, just waiting on my OpenSolaris box to install enough packages to actually be usable. [08:52:01] d'you mind shooting me an email about it? I'll have to get the box up sometime tonight [08:52:39] Remind me what your email address is? [08:52:48] this nick at gmail [08:52:59] or theba004 at morris.umn.edu [08:54:52] Cool. Thanks. Should be able to give you some patches soon. Just trying to verify that things build on Solaris at the moment. [08:56:20] alright, whenever works [09:26:11] --- sxw has left [09:49:33] --- manfred has left [09:57:40] --- dev-zero@jabber.org has left [10:14:00] --- sxw has become available [10:43:44] --- Russ has become available [10:45:38] --- manfred has become available [10:57:10] --- dev-zero@jabber.org has become available [11:48:26] --- shadow@gmail.com/owl2BB11B64 has left [11:53:44] --- shadow@gmail.com/owlB0D38068 has become available [12:25:20] --- mcohan has left [12:25:51] --- dmontuori has left [12:25:53] --- dmontuori has become available [12:26:37] --- stevenjenkins has left [12:26:38] --- stevenjenkins has become available [12:26:43] --- mcohan has become available [12:39:18] --- dmontuori has left [13:05:08] --- dmontuori has become available [13:36:35] --- manfred has left [13:46:42] --- manfred has become available [13:52:53] Russ: Sorry - you must be getting buried in bug mail at the moment ... [13:56:21] --- SecureEndpoints has left [14:05:35] Oh, no, this is great. [14:05:40] --- SecureEndpoints has become available [14:05:40] This will really help me see what's going on. [14:06:52] really, openafs-cvs does far more (worse) mailbombing [14:07:52] Well, once my mail client catches up, that's all of the outstanding disconnected bugs (of which I'm aware) in RT. [14:07:54] Yeah, and I'm already used to that. [14:19:35] --- manfred has left [14:21:57] Everything going to -cvs should be legitimate [14:22:06] it is [14:22:12] there is just a lot of it [14:23:10] mailbomb. not spam :) [14:23:56] if you want to send me spam, sent it to canned-ham@secure-endpoints.com :-) [14:24:29] thansk [14:39:22] --- dev-zero@jabber.org has left [15:13:09] Are we thinking that OpenAFS will apply as a GSoC mentoring organisation this year? [15:13:29] yes [15:13:56] I'm probably going to be away until after the sucessful mentoring organisations have been announced. Are we collating ideas anywhere? [15:14:05] although I'm not sure that I will be a mentor this year. working with all of the students of other mentors was enough for me [15:14:23] we still have ideas from last year that were not completed [15:14:25] I'd be interested in mentoring again. [15:14:28] Indeed. [15:14:33] feel free to submit additional ones [15:14:40] Where are we collecting them? [15:14:48] send them to RT or to me [15:14:54] or to openafs-gatekeepers [15:15:09] Okay. Cool. Just wanting to get some in before I leave... [15:15:12] they will get put up on a gsoc web page that does not currently exist [15:15:26] k [15:15:42] One of the ideas I'm thinking of is doing partial disconnection, similar to what Coda has. [15:16:00] things I would like to see are a new file manager for either the n8xx or android [15:16:16] Yes, although I don't think those would really be our projects. [15:16:32] At least for the n810, the new file manager should be properly platform independent. [15:16:34] they would be because we need them to support our work [15:17:12] I guess, but I would have though they would be better supported from within those communities, than from within ours. [15:17:23] --- dmontuori has left [15:17:24] I would also like to see a prototype user interface for disconnected mode [15:17:26] --- abo has left [15:17:30] --- dmontuori has become available [15:17:37] --- stevenjenkins has left [15:17:45] --- mcohan has left [15:17:55] --- mcohan has become available [15:18:09] perhaps, but then those communities won't understand the needs of a network file system [15:18:14] --- abo has become available [15:18:18] apparently they got it wrong before [15:19:08] I wouldn't expect what is produced to be distributed by Nokia or Google, but it might provide them a user experience for network file systems that they can try to match [15:19:22] Ah. Okay. [15:19:52] What I'd like to see is hildon-fm reworked so that it can deal with network-fs's, and that pulled in to the upstream distribution. [15:20:37] --- stevenjenkins has become available [15:20:53] I started on that at the Google Hackathon, but rapidly got lost in a world of gnome-vfs layer insanity (why have one filesystem tree in memory when you can have 3 subtly different versions of the same data?) [15:28:57] agreed [15:29:57] I also think we need to start working on the user interface to support disconnected operations. We really don't know how to make it work well from a user perspective and the design of the user interactions really will impact the functionality we require in the cache manager. [15:31:49] Agreed-ish. [15:33:04] I'm not really interested at the moment in producing a GUI style user interface. I think a GUI interface for AFS would be useful, but I'd like to have one that we could add disconnected operations on to, rather than building one just for disconnected AFS. GUI interfaces in Unix also have the problem that we'd have to build them OS-by-OS. [15:33:51] The command line user interface has the real issue that you need to be root to change the state of the machine. I don't really see any way around this - taking a machine from online to offline is something that the super user has to do. [15:41:43] 1.5.x doesn't build on OpenSolaris due to type conflicts in afs_nfsdisp.c. I can fix them pretty quickly - do we care? [15:45:04] What do you mean? [15:45:58] I recall that file as being the place where struct conn issue arose? [15:46:25] Let me take a closer look ... [15:47:30] I would appear to be losing my mind. Building the wrong tree. [15:51:07] matt: Would you mind taking a look at /afs/inf.ed.ac.uk/user/s/sxw/Public/openafs-cleaner-cfileopen.patch on your collection of BSDs? It's the cache manager change to abstract out a cache inode. [15:52:19] Sure. It may take a day or so... [16:11:17] --- dmontuori has left [16:21:59] > 1.5.x doesn't build on OpenSolaris due to type conflicts [16:22:02] we should caree [16:22:04] care [16:23:22] I was wrong. A failure to cvs up in the correct build tree. [16:23:46] When you get a moment, though, if you could try the cache inodes patch on IRIX, that would be cool. [16:25:54] (RT124184) [16:26:38] i'll try it [16:27:02] Thanks. I suspect if there's anything wrong, it just won't build. [16:37:56] But I can confirm that 1.5.x tip now builds on Linux, Darwin and Open Solaris - so no repeat of the 1.5.56 problem ... [16:38:53] see release-team mail ~seconds ago [16:40:49] I'll see if I can do RPM builds tomorrow. [16:42:04] Before I leave, I'm going to add the 2.6.29 kernel patch to the 1.4.8 SRPM, so hopefully they'll keep building when Fedora 10 gets its new kernel - just so you know ... [16:51:17] --- mdionne has become available [16:53:44] mdionne: Just the man. If you've got a moment, could you have a look at /afs/inf.ed.ac.uk/user/s/sxw/Public/openafs-cleaner-cfileopen.patch , and see what you think? It's the reworking of cache inodes inspired by the LINUX_USE_FH stuff. [16:54:05] yay [16:54:18] yay? [16:54:23] 2.6.29 yay [16:54:29] keep building yay [16:54:45] Hopefully yay. Won't be here to care if nay :) [17:04:11] we will hunt you down if necessary [17:18:13] --- matt has left [17:25:15] Simon, having a look at it now [17:33:45] --- mdionne has left [17:36:37] --- dlc has become available [17:41:18] --- sxw has left [17:42:31] --- mdionne has become available [17:44:00] --- mdionne has left [18:57:20] simonwilkinson: might not get a chance to look at that patch until friday, FYI [19:36:37] brain tired. over 8 hrs spent in core dumps today. not fun. [21:17:08] --- dev-zero@jabber.org has become available [21:30:36] --- Russ has left: Disconnected [22:34:39] --- dev-zero@jabber.org has left [23:14:52] --- reuteras has become available [23:39:02] --- dev-zero@jabber.org has become available