[00:13:40] --- Russ has left: Disconnected [00:13:48] --- Russ has become available [00:21:45] --- Simon Wilkinson has become available [00:25:38] --- Simon Wilkinson has left [02:48:59] --- Russ has left: Disconnected [04:34:24] --- mmeffie has become available [04:56:01] next time red* is awake: why do you believe 0x8000 came from icl? [05:04:15] --- mmeffie has left [06:16:45] --- dev-zero@jabber.org has left: Replaced by new connection [06:16:46] --- dev-zero@jabber.org has become available [06:45:52] shadow - looking at the code of how %edx got set [06:46:19] something just occurred to me [06:46:33] did you have crashes with the distributed modules or only with self-compiled? [06:46:42] (e.g. "is there an optimizer bug") [06:47:08] I believe they were all self-compiled... either via dkms or via rpm --rebuild [06:47:24] you never got a crash from our modules? [06:47:38] never used your modules [06:47:59] would you be able to try that? [06:48:16] what's different about the way your modules are built? [06:48:49] which compiler version are you using? [06:49:50] possibly "different gcc version"; i don't know what's on simon's vm and at this point he's offline, i think, and presumably in ireland [06:50:09] I suspect a plane over the Atlantic [06:50:20] oh, this is the trip to nyc? [06:50:31] it's certainly worth finding out, first [06:50:40] trying out 1.4.8 is prolly also worth it [06:50:45] worth finding out which? [06:50:52] what os/gcc he uses to build [06:50:59] 1.4.8pre1 won't fix it. i'd try our 1.4.7 modules first [06:51:45] like, i'd argue knowing which compiler is more dull than knowing it no longer happens with our module and then tracking down environment differences [06:51:59] but, if it doesn't happen with ours i can stop looking for fixes to the kmod [06:52:12] bus... back in a bit [07:04:35] --- dmontuori has become available [07:12:39] back [07:15:32] do those kmods have debuginfo? [07:17:43] shadow - anyway, when I was tracing back, just looking backwards through the code, in afs_Analyze, there's only a few places %edx is set, so, looking at where it was previously set to something not 0x8000, it looks like it was set in a function called [07:18:19] you should be able to look and tell me if I'm reading the assembly completely wrong pretty quickly [07:28:12] debuginfo: dunno. don't care. if it reproduces the crash you have, we have debuginfo from one of those. if we get a new crash, well, seems these are reproducible anyway. if we get no crash, we start trying to match your environment to simon's [07:28:17] or vice-versa [07:28:43] i think the path forward is "try our modules", given the resources which are available to pursue this now, and that no one else has seen this [07:33:25] k [07:35:25] scheduled reboots for this afternoon [08:05:07] --- reuteras has left [09:10:24] --- dev-zero@jabber.org has left [09:12:51] --- moose has become available [10:15:32] actually, i just realized what the cache pinning needs (tho not easy to do). keep a running total of the size of pinned data and don't let it exceed the cache size, or, optionally based on user input, some percentage of it and ultimately a gui showing percentage cache use. the split cache code makes this more interesting but a diferentiator function other than the provided one would be better yet [10:17:33] --- dev-zero@jabber.org has become available [10:23:46] yep. [10:36:03] --- Russ has become available [11:33:00] --- dragos.tatulea has become available [11:33:42] --- dragos.tatulea has left [14:38:31] --- Simon Wilkinson has become available [14:40:36] Derrick - am now running on your kmods [14:42:23] --- moose has left [15:13:44] --- todd has become available [15:14:00] so... [15:14:09] how long does it take to get a CellServDB entry processed? [15:21:28] Back on May 29th I got ticket #101004 (no clue what this gets me other than reply history) [15:21:45] still don't see myself in the public CellServDB (freedaemon.com) [15:21:48] I'm sure that jhutz will publish a new one before 1.4.8 is released [15:22:11] since the CellServDB is bundled with the new release. [15:22:27] I see [15:22:43] I do note it was not generated since I sent my initial email [15:24:18] it has not been generated since 1.4.7 [15:26:21] --- shadow@gmail.com/owl27BF0A01 has left [15:26:39] --- Simon Wilkinson has left [15:42:06] I vaguely remember someone complaining that the AFS Perl module didn't work with current OpenAFS, but with a few minor patches, it passes all of its test cases. [15:42:16] Was there something fundamental broken, or did I just fix it? [15:42:22] I wish I could remember where I saw that mail. [15:46:16] wasn't the complaint something that was mentioned at the workshop? [15:47:00] Oh, hm, maybe I didn't go back far enough. [15:47:33] The com_err stuff definitely broke it, but that was an easy fix. [15:49:21] there is the complaint that the perl module doesn't compile on 64-bit amx [15:49:22] amd [15:49:29] Yeah, that I know about. [15:49:35] That's just because the libraries aren't built PIC. [15:49:46] Which actually they might be now. [15:50:20] Yup, indeed, they are. I wonder if it just works these days. [15:50:23] I'll try in a few. [15:50:41] Norbert Gruener seems to be the one with most of the complaints these days [15:52:43] I wonder if the perl module would build on windows [15:53:40] It uses a bunch of the other libraries that aren't in afsauthent and afsrpc. [15:53:58] they could be added [15:54:10] Admittedly, it does this in part to provide interfaces to stuff I don't care about, but that's probably the main hold-up. [15:54:24] I don't know how good the Perl XS stuff is at coping with Windows compilers either. [15:55:57] where are the sources? [15:58:41] http://www.mpa-garching.mpg.de/kwiki/nog/afsperl/ [16:03:33] Yeesh, my Debian packaging of this was ancient. [16:03:46] Haven't touched this since 2005. [16:04:06] making that build on Windows is goiing to take more than 20 minutes [16:04:23] Yup. [16:04:46] Oh, crap, and it was in Sam's Subversion repository. I thought I'd sucked everything out of that. [16:05:01] Right, I think I start over in Git. [16:08:28] Interesting. I wonder what in the Perl AFS module isn't working with 1.4.7, since I've been using 2.4.0 of the module with the libraries from 1.4.7 pretty heavily the past few weeks. [16:08:57] It won't compile unless you screw with the com_err bits, since there's no error_message symbol in 1.4.7. [16:09:17] Or rather, it will compile, but the dynamic object can't resolve error_mesage at load time. [16:09:44] Interesting. I don't remember futzing with that. [16:18:16] --- todd has left [17:46:28] --- summatusmentis has left [18:44:06] --- summatusmentis has become available [18:46:13] --- Russ has left: Disconnected [18:46:55] --- summatusmentis has left [19:04:44] --- summatusmentis has become available [19:05:44] --- summatusmentis has left [19:29:47] --- summatusmentis has become available [19:43:40] --- shadow@gmail.com/owl321D9145 has become available [21:02:24] --- summatusmentis has left [21:02:35] --- summatusmentis has become available [22:10:30] --- reuteras has become available [22:14:20] --- reuteras has left [22:21:26] --- Russ has become available [23:18:14] --- Rrred has left [23:18:57] --- dwbotsch has become available