[00:22:36] --- reuteras has left [00:55:51] --- Russ has left: Disconnected [01:41:11] --- haba has become available [02:05:53] --- pod has become available [02:36:54] --- Jeffrey Altman has left: Replaced by new connection [02:36:54] --- Jeffrey Altman has become available [02:38:17] --- jaltman has left: Replaced by new connection [02:38:20] --- jaltman has become available [03:04:24] --- jaltman has left: Disconnected [03:14:30] --- jaltman has become available [03:32:37] --- jaltman has left: Replaced by new connection [03:32:39] --- jaltman has become available [03:45:55] --- jaltman has left: Disconnected [03:46:27] Just a quick warning for Mentors. I seem to have a large number of new notifications on socghop, none of which have been emailed to me. [03:49:43] --- Jeffrey Altman has left: Replaced by new connection [03:51:08] --- jaltman has become available [04:12:33] --- jaltman has left: Disconnected [04:55:40] --- haba has left [05:04:53] --- haba has become available [05:26:45] --- jaltman has become available [06:05:08] --- haba has left [06:05:08] --- haba has become available [06:06:36] --- reuteras has become available [06:10:35] --- reuteras has left [06:31:39] --- meffie has become available [07:00:51] --- meffie has left [07:14:08] --- shadow@gmail.com/owl33D1E86D has left [07:19:52] --- deason has become available [07:36:32] --- shadow@gmail.com/owlB8C24CDF has become available [07:42:31] --- jaltman has left: Replaced by new connection [07:42:35] --- jaltman has become available [08:23:04] --- jaltman has left: Disconnected [08:41:38] the ka_ChangePassword thing was probably a pointer size issue, e.g. the 0 was a pad to fill out something to 64 bits. as to why, i dunno. we got that port from neale ferguson, who i assume still works for SNA [08:44:23] In which case, having proper prototypes is almost certainly the right fix. [08:44:27] --- phalenor has left: Lost connection [08:46:46] --- phalenor has become available [08:47:46] the specific change for the 390 special case appears to have come from carsten jacobi, 5e8ee41b [08:49:42] http://www.mail-archive.com/openafs-devel@openafs.org/msg01017.html [08:50:16] and it was a pointer issue, see the immediate previous message [08:51:27] So, would the protoyping fix it, now that the compiler knows that what's being passed is a pointer, not an int? [08:52:17] well, if nothing else i assume it would make an error be thrown if there was still a real error [08:52:41] so it should be safe to just take the "undoes that" patch [08:53:00] in the change to remove the s390 special-case, there is still a call to ubik_Call [08:53:24] so I would imagine it'd still be problematic [08:53:56] we can fix that now [08:54:27] ubik_Call->ubik_KAM_SetPassword [08:54:50] yeah [08:58:13] 1793 [09:01:17] --- kaj has left [09:11:50] The world just got too much for gerrit's tiny little brain. [09:16:59] --- haba has left [09:16:59] --- haba has become available [10:02:49] I have a change which will make OpenAFS builds determine their release by running 'git describe' [10:03:21] However, this is going to require release tarballs having a file which contains the release information. Would the CML stuff be an appropriate way to do this? [10:38:06] --- haba has left [11:26:25] --- meffie has become available [11:28:21] > Would the CML stuff be an appropriate way to do this? yes. make_release should generate one. [11:29:41] What should a CML file look like? [11:30:28] CML file: hang on [11:40:27] --- Russ has become available [13:07:37] --- steven.jenkins has left [13:10:58] --- steven.jenkins has become available [13:50:49] --- jaltman has become available [13:53:08] --- mdionne has become available [14:00:21] --- jaltman has left: Disconnected [14:02:06] --- dwbotsch has left [14:05:13] --- dwbotsch has become available [14:44:53] --- meffie has left [15:20:26] --- pod has left [15:34:21] --- deason has left [16:36:08] --- deason has become available [16:58:00] --- deason has left [17:01:18] --- shadow@gmail.com/owlB8C24CDF has left [17:01:29] --- shadow@gmail.com/owlEE34A744 has become available [17:02:24] --- deason has become available [17:04:29] RFC 5864 on DNS SRV Resource Records for AFS published. [17:07:38] yay [17:08:07] tom's rx clear and my pts extended names drafts have just been reissued and are looking for reviews. [17:12:00] --- jaltman has become available [17:30:38] I think that the FUSE work for the Red Hat packaging should be developed and applied separately, since it needs additional changes like the new configure switch, the new build dependency on the FUSE development package (and I don't know what it's called on Red Hat), etc. My patch shouldn't break anything. [18:15:50] shouldn't it be separate for debian as well? [18:15:59] (I mean, in terms of packages) [18:19:27] Yes. [18:19:38] I'm not bothering right now for the experimental packages, but I'll break it out before it goes into unstable. [18:19:49] It probably needs to use alternatives. [18:22:03] But doing that properly also requires init script work to not try to load the kernel module and not abort if the kernel module is unavailable if the FUSE variant is installed. [18:22:12] And I'm not quite sure how I'm going to do that yet. [18:22:30] Hm, I can probably just check to see where /etc/alternatives/afsd is pointing. [18:25:20] --- jaltman has left: Disconnected [18:26:16] fwiw, you can mount fuse filesystems just using fstab iirc; though it's probably incredibly obnoxious with the options you tend to pass to afsd [18:27:39] I also wasn't really thinking if having it needed some wider discussion or something... it pretty much just exists now as something I was using to test libuafs [18:28:44] I think having a FUSE method is really, really useful in some situations where you don't feel like dealing with the kernel module. [18:29:06] For instance, hosted Xen VMs sometimes make it annoyingly hard to build the kernel module for the kernel that you're running, since you don't necessarily know what source that kernel was built from. [18:31:04] I know the advantages of having one (er, well, some of them); I just mean, I don't really know how well it works, and I hadn't thought too much about integration with everything [18:31:11] i'm really looking forward to the libuafs talk at the workshop [18:31:31] Oh, yeah, I was going to ask -- does it register the ioctl so that you can get PAGs and whatnot? [18:31:42] no, it's completely unauthenticated right now [18:31:50] Ah, okay. [18:31:57] (I’ll go farther and say that I hope one day the kernel module meets the painful fiery death that it has deserved for many years. But maybe that’s just me.) [18:32:27] the kernel module is awesome. less so on platforms with arbitrarily constricted interfaces. [18:32:30] andersk: The solution there is kafs. [18:32:36] doing authenticated access and support for the normal userland binaries will be... interesting [18:32:50] deason: I wonder if the userspace support for the NFS translator helps you. [18:32:58] It seems like it at least potentially could. [18:33:09] it would be easier if we did something like xattrs for pioctls or something, but it's but a distant thought [18:33:10] If we have to extend FUSE to do it, then great: we’ll make FUSE better for everyone. [18:33:12] Make all the userland utilities think the FUSE afsd is a remote NFS translator service. [18:33:50] kula: cool! I better have something to say, then :) [18:34:07] Okay, 1.5.74-2 uploaded to Debian experimental with afsd.fuse installed in openafs-client. Also hopefully fixing the build problems on s390 and powerpc. [18:36:19] okay, but my focus/target isn't really fuse right now; libuafs perl bindings will be coming in when I get time to forward port it [18:44:49] --- Russ has left: Disconnected [19:21:32] --- Russ has become available [19:24:53] That sounds pretty cool. [19:37:16] 1.5.74-2 is built in the Ubuntu PPA. [19:38:21] --- mdionne has left [20:56:30] Okay, build problem on s390 confirmed fixed. [20:56:33] powerpc hasn't tried yet. [21:40:43] --- kaj has become available [22:08:18] --- Born Fool has become available [22:12:07] --- Born Fool has left [22:12:11] --- deason has left [22:13:17] --- reuteras has become available [22:33:50] --- Born Fool has become available [22:55:29] --- Born Fool has left [22:59:37] --- kaj has left [23:53:36] I'm not convinced that kafs is going to be the solution to our problems. I think it's the solution to our "Argghhh, they broke our interfaces" problem, but I don't think it's ever going to match the feature set of the OpenAFS client. Just look at the issues that David has with getting even simple things in. [23:56:44] Yeah, there is that. [23:57:37] Admittedly, to some extent I agree with the objections to some of the ways AFS does things, and the approach the kernel community takes to forcing people to clean up all their dark corners (and some of their light ones that just happen to offend their sense of taste) does have some things going for it. [23:57:50] But for our purposes that may always go a bit too far. [23:57:56] --- kaj has become available [23:58:29] The Linux kernel module is a much better citizen now than than it was a couple of years ago. By and large, we now use the "proper" interfaces, and do things in slightly more standard ways. [23:58:49] But that's not going to stop them from breaking us in their crusade against all things non-GPL. [23:58:52] Yup. [23:59:00] Which is just annoying.