[00:42:27] --- kaj has become available [01:18:28] --- Russ has left: Disconnected [01:18:59] --- kaj has left [02:24:33] --- sxw has left [02:39:21] --- Simon Wilkinson has left [02:46:03] --- haba has left [03:45:56] --- Jeffrey Altman has left: Replaced by new connection [03:46:31] --- jaltman has left: Replaced by new connection [03:46:31] --- jaltman has become available [03:48:09] --- sxw has become available [03:50:48] --- haba has become available [05:05:04] --- pod has left [05:47:58] --- dwbotsch is now known as RedBear [05:47:58] --- RedBear is now known as dwbotsch [05:47:58] --- dwbotsch is now known as RedBear [05:48:15] --- RedBear has left [05:48:15] --- dwbotsch has become available [05:48:21] --- dwbotsch has left [05:48:58] --- RedBear has become available [05:49:39] ok... I was gonna test this new mac dmg [05:55:23] nope, not fixed [05:55:43] tho, is this the latest? AFS version: OpenAFS 1.5.72 built 2010-03-10 [05:55:58] well, wait, I take it back, sorta [05:56:52] I thought Derrick had built a special dmg with the fix for you to try? Are you running that, or one of the release builds/ [05:57:09] running the latest dmg I could find in his homedir [05:57:15] so, unless he hid it someplace else [05:58:19] no, definitely would not call this fixed... must be the wrong dmg [06:03:13] This the dropbox bug? [06:35:24] sure worked for me. a directory i had li on i could using finder drop files into. [06:35:33] shrug. you can try 1.5.73 when it's out [06:37:47] I wonder if we should make git builds of OpenAFS report the git hash as part of their version number. [06:38:39] works until you cherry-pick things in [06:38:47] considered tho. probably a decent idea [06:38:58] oh, wait... 2 different bugs [06:39:06] you were only fixing the drop, not the not showing files part [06:41:14] and.... finder is spinning at "Preparing to copy to 'incoming'" [06:41:49] it showed files to me right along [06:42:06] well, it's not here [06:42:12] it does seem to think for a while about preparing to copy [06:42:32] might be that 10.5 sucks. [06:42:37] 10.6 [06:42:50] *using Snow Leopard* [06:42:54] weird. well. i can confirm 10.6 did exactly what i expected. [06:43:09] ok... finally got a list of files in said folder [06:43:21] but still "Preparing to copy" [06:44:36] could it be being held up somehow by finder doing all this background stats of other parts of afs that it doesn't have r access to? [06:45:18] yes, it could. finder's idea of doing stuff in the background isn't really 100% in the background apparently [06:46:22] Hmmm. More stuff about IMA on Fedora 12. But I'm loathe to put a new cache file access mechanism in just to solve this, when we know that mechanism has locking issues. [06:46:26] are there any changes to the server side that might be helping the situation? [06:46:45] our servers are still at 1.4.10 [06:47:15] --- reuteras has left [06:47:56] --- reuteras has become available [06:54:00] i was testing against andrew's servers. probably older than 1.4.10 [06:55:05] copy: The operation can't be completed because you don't have permission to access some of the items [06:55:13] and I have a 0 size file in there [06:55:15] so, did not work [06:55:57] what's the acl [06:56:07] it works from the commandline just fine [06:56:33] Normal rights: cnfhosts li grp_all li system:administrators rlidwka admin rlidwka dwb rlidwka dwb7@cit.cornell.edu rlidwka [06:57:56] for kicks, find a directory you can make system:anyuser li and try it? i'm curious if there's something else at play [07:01:10] well... there's the slight problem of finder not refreshing what directories are there [07:03:08] perhaps finder is doing some work not using the uid of the user and therefore doesn't get tokens. [07:03:39] I'm doing this w/o tokens... being on our subnet should be sufficient [07:03:59] pts members cnfhosts Members of cnfhosts (id: -211) are: 128.253.198.0 128.253.11.0 10.253.198.0 [07:04:01] yeah. well. that's an fsevent. i have a patch that fixes it at the expense of shutting down hanging. something is leaking vnode refs. it's not me that i can tell [07:04:25] the problem is the fsevent generation is not exposed to kpi users (e.g. us) [07:04:38] apple knows. i am using a hack in the meantime but it may have side effects [07:05:13] if i could work aroun the vnode ref leak you could have finder refresh on callback break [07:05:28] the copy seems to work in a mostly empty subdirectory, both with and without system:anyuser li [07:05:40] tried first w... then took anyway the system:anyuser... copy still worked [07:05:46] --- deason has become available [07:05:54] Finder isn't setuid, so it's unlikely that it's doing things as not-user. [07:06:31] so... some relation, perhaps, to the quantity of stuff in the folder one is copying to? [07:06:58] about 210 existing items in my dropbox folder [07:07:51] there was stuff in my dropbox. [07:08:03] if you can isolate the issue further i can recreate [07:08:17] how much stuff was in your dropbox? [07:08:20] in any case, it does seemingly work more now than before? [07:08:34] maybe a dozen things. at the start, one thing. i kept dragging shit into it [07:13:41] and now it's working for the folder [07:14:07] ? [07:14:18] the copy worked... no preparing to copy [07:14:36] after all the stat()s are done and cached, perhaps? [07:14:46] which means the error the first time was probably a timeout? [07:14:49] it is still statting stuff someplace in my volume [07:17:37] it will do that forever, seemingly [07:17:49] --- pod has become available [07:54:41] er, I thought IMA started with 2.6.30? is it not a problem back then? [07:56:31] Yeh, I should have checked with git before typing. [07:56:36] It landed in the kernel with 2.6.30rc1 [07:56:56] I have a recollection that Fedora didn't enable it until 2.6.31, though, although that could be wrong. [08:10:15] --- cclausen has become available [08:17:34] --- sxw has left [08:31:37] --- jaltman has left: Replaced by new connection [08:31:38] --- jaltman has become available [08:40:00] --- meffie has become available [08:40:29] --- jaltman has left: Replaced by new connection [08:40:30] --- jaltman has become available [08:44:56] --- kaj has become available [09:03:28] this article. yeah. ++. http://lwn.net/Articles/370419/ [09:03:50] --- jaltman has left: Disconnected [09:04:02] --- jaltman has become available [09:36:28] --- reuteras has left [09:51:43] --- haba has left [10:11:37] --- mdionne has become available [10:12:11] --- sxw mobile has become available [10:13:27] --- sxw mobile has left: Lost connection [10:14:01] Not getting any gerrit mail today - related to the upgrade? [10:17:41] i bet it is. [10:18:13] hopefully next time simon's around he'll notice your comment [10:20:23] btw the web page needs to be updated for 1.4.12. The bits seem to be in gerrit already [10:23:14] uh. that means the export script failed [10:23:44] weird. rerunning it seems to have fixed. it will be there in a few seconds [10:24:20] fixed. sorry [10:24:53] (i did the release between somewhere and somewhere else Saturday; my neighborhood is the party-tastic neighborhood here so we ran away from the crazy drunk people) [10:31:50] don't see the new page yet, but that may well be caching somewhere [10:32:17] if you can, force reload. it loaded for me [10:32:51] i hate proxies... [10:33:45] didn't know if it was that. firefox often needs a hint to reload frames [10:35:09] Looks ok on my phone, so it's surely the local proxy [10:37:00] Ok [10:38:11] Meant to say,opening the frame by itself fixed it. Thanks. [10:41:07] yeah. "reload frame" in firefox usually fixes it [10:42:09] --- mdionne has left [10:43:24] I'm restarting without NAT ping since I have no time to debug that problem at the moment [10:47:03] --- mdionne has become available [11:08:52] --- mdionne has left [11:33:37] --- meffie has left [11:36:07] so what precludes you from making new calls on a conn marked RX_CONN_DESTROY_ME? it looks like nothing. [11:38:29] likewise, if a conn is for some reason RX_CONN_BUSY when its refCount hits 0, i suspect we don't dtrt (tho that'd be a bug anyway, i think) [11:45:38] The fact that you called rx_DestroyConnection and thus no longer have a legal reference to the connection? [11:49:39] *i* called rx_DestroyConnection. doesn't mean every thread has [11:53:45] --- Russ has become available [12:05:41] Yeah, so? _you_ no longer have a reference and aren't allowed to make new calls. If other threads still have references, they can make calls. [12:07:45] well, they *shouldn't* have a reference. the point is nothing is enforcing that [12:07:57] if they have a reference, refcount should not be 0 [12:13:06] --- haba has become available [12:13:39] --- meffie has become available [12:17:19] If refcount is zero, then either your code is buggy, or no one has a ref. Hence, no problem. [12:17:45] "your" code. it's the rx library. it's probably buggy [12:18:06] If they try to create calls without a reference, the "connection" might well be random garbage memory. There's really nothing you can do about it. [12:18:17] Er, yes; by "your code" I mean the rx library. [12:18:25] this isn't new. i'm laying odds openafs *already from day 1* had code that was in fact wrong. we know it was buggy. [12:18:50] sure we can. we can assert or something and actually find the real bugs [12:18:59] Yeah, but you've been a gatekeeper since day one, so it's all "your code" now, for purposes of blame. :-) [12:19:18] i blame lyle seaman :) [12:19:21] you can assert or something if refcount is 0. that won't find the bugs where they call you with garbage. [12:20:17] i am not going to touch it today. it's on my list for "at some point" [12:27:26] --- jaltman has left: Replaced by new connection [12:27:26] --- jaltman has become available [13:53:42] --- haba has left [14:01:50] damn. ok. i suspect i can make fsevents now. it won't be perfect. but i found my issue. *facepalm* [15:11:04] --- meffie has left [15:13:20] --- deason has left [15:19:03] --- haba has become available [15:25:34] It's time to do routine patching of git.openafs.org / gerrit.openafs.org and the host for this conference. When would be a good time? [15:32:32] how much time is required? [15:33:17] --- jaltman has left: Disconnected [15:33:56] The system will probably be down for less than 10 minutes, unless something goes wrong. It needs a reboot for the kernel upgrade, and there's also a new PostgreSQL server package that hopefully won't cause any problems, plus lots of other minor things. [15:45:01] --- mdionne has become available [15:49:51] no ksplice? :) [16:18:40] --- jaltman has become available [16:19:40] I suspect that now is as good a time as any for me [16:20:08] we have had gerrit down for a lot longer than 15 minutes and everyone has lived [16:22:07] Okay. I can do it now if there are no objections. [16:27:09] --- pod has left [17:12:45] --- jaltman has left: Disconnected [17:20:19] --- pod has become available [17:38:05] --- summatusmentis has become available [19:03:56] --- jaltman has become available [19:05:57] --- mdionne has left [19:18:40] * Russ was rather ambiguous about the last and hasn't actually done anything. I'm going to do the upgrade at 8pm PDT (in 40 minutes) unless anyone yells. [19:32:12] go nuts [19:32:33] --- Born Fool has become available [19:43:28] --- deason has become available [20:44:14] --- Born Fool has left [21:12:34] Or I'll forget completely and do it an hour and a half after that. Damn, I'm spacy today. [21:12:39] Anyway, doing the upgrade now. [21:12:48] woot [21:12:56] omg omg stop stop i need to. yeah, kidding [21:13:13] * Russ laughs. [21:13:22] I mention in part because I could still stop for wise-asses. :) [21:16:54] And rebooting now -- server will be back up shortly. Everyone will get kicked off this chat and need to rejoin (which most clients will do automatically). [21:19:58] --- LOGGING STARTED [21:19:58] --- Derrick Brashear has become available [21:20:02] --- Russ has become available [21:20:15] --- kula has become available [21:20:20] --- shadow@gmail.com/owl68DB543C has become available [21:20:31] --- Derrick Brashear has left [21:20:48] --- phalenor has become available [21:22:19] --- jaltman has become available [21:22:36] --- cudave has become available [21:23:45] --- summatusmentis has become available [21:24:08] Everything should be back up. [21:24:58] --- RedBear has become available [21:30:13] --- Born Fool has become available [22:24:29] --- Russ has left: Disconnected [23:02:36] --- Born Fool has left [23:12:56] --- reuteras has become available