[00:00:41] --- kaj has become available [00:03:49] --- dwbotsch has left [00:05:30] --- phalenor has left [00:07:19] --- dwbotsch has become available [00:11:59] --- Born Fool has left [00:14:50] --- phalenor has become available [01:11:47] --- steven.jenkins has left [01:16:45] --- steven.jenkins has become available [01:29:17] --- abo has left [01:29:52] --- abo has become available [01:31:58] --- Simon Wilkinson has become available [01:42:21] --- steven.jenkins has left [01:42:36] --- steven.jenkins has become available [02:08:16] --- steven.jenkins has left [02:08:57] --- steven.jenkins has become available [02:27:22] --- Simon Wilkinson has left [02:49:44] --- steven.jenkins has left [02:50:09] --- steven.jenkins has become available [03:55:08] --- steven.jenkins has left [03:55:23] --- steven.jenkins has become available [03:59:44] --- Simon Wilkinson has become available [04:12:04] --- JSund__ is now known as JSund [04:12:04] --- JSund is now known as JSund__ [04:12:04] --- JSund__ is now known as JSund [04:12:13] --- JSund has left [04:12:14] --- JSund__ has become available [04:12:25] --- JSund__ is now known as JSund [04:12:25] --- JSund is now known as JSund__ [04:12:25] --- JSund__ is now known as JSund [04:16:43] --- JSund is now known as JSund__ [04:16:43] --- JSund__ is now known as JSund [04:16:52] --- JSund has left [04:16:53] --- JSund__ has become available [04:16:59] --- JSund__ has left: leaving [04:17:53] --- JSund has become available [04:26:46] --- rod has become available [04:40:43] --- Simon Wilkinson has left [05:04:42] I'm not sure I understand the objection to having a hypen in the file name. personally I think the best name would be fileserver-da [06:23:04] ick [06:59:54] --- mdionne has become available [07:04:45] pushing to gerrit is an order (or two) of magnitude faster, I assume as a result of the git gc [07:07:40] with enable-fuse-client now the default, configure just bails out if you don't have the right fuse bits. In my case I was missing fuse-devel [07:24:42] --- dwbotsch has left [07:25:00] --- cudave has become available [07:26:15] --- deason has become available [07:40:25] --- reuteras has left [08:32:40] > with enable-fuse-client now the default, configure just bails out which means the PKG stuff blows. sigh. [08:34:48] there's a function that bails if it doesn't exist, and another one that doesn't bail if it doesn't exist [08:35:05] just need to use the other one, depending on if an explicit --enable was passed or not? [08:35:08] --- Russ has become available [08:35:55] actually, i thought just setting the "action if not found" would fix it, seeing what the function is, now [08:36:11] and now it will just try to build. no enable needed. [08:37:36] I know; I just thought passing --enable would cause it bail out if it doesn't find it would be nice [08:37:59] (defaults to autodetect, pass an explicit enable or disable to force) [08:43:15] that would involve more than yes/no defaulting to yes. how does --enable differ from the yes we already had? [08:43:36] i suppose i could do the wretched thing i did with inode/namei/"default" [08:43:47] See how the krb5 macros handle it. [08:44:08] If --with-krb5 is specified, they abort if Kerberos isn't found; otherwise, they just continue on if it's not found and don't build stuff. [08:44:27] we already pass an explicit var=yes in the "if not specified" case... I thought it'd just be var=auto or something [08:44:34] or var= [08:44:57] Actually, that's not a great example, not because it doesn't work but because it's too complicated to follow. [08:45:01] (Mine.) [08:45:05] --- meffie has left [08:45:07] apparently on rhel6 beta there's debug stuff in the kernel enabled by default. that's fine. but it wants to use KBUILD_MODNAME in a macro it uses. i can define it, but i thought kbuild was supposed to define it for us? [08:45:38] the namei example is better. i remember doing it, and what i did [08:45:49] Here, I'll propose a patch in Gerrit to do what I think should be done with FUSE. [08:48:26] russ: ok [08:48:56] --- meffie has become available [08:59:10] so the fuse change from russ when it lands, the KBUILD_MODNAME thing... other than documentation changes and the register thing, that's it, right? [08:59:23] (e.g. no one knows of anything else) [08:59:37] --- kaj has left [09:01:06] I'd like to fix VSALVAGING getting sent over the wire... but don't need to wait on that [09:02:14] yeah, that's a bug fix, easy to pull in whenever [09:02:34] --- Kevin Sumner has left [09:02:53] --- Kevin Sumner has become available [09:12:36] russ: if you haven't done it yet, 2418? if so, i can abandon [09:24:25] I've done it already, and was just testing all the iterations. [09:24:53] if you push i can test. mine failed one of the tests [09:25:02] tho i just fixed it [09:29:19] tho i just realized how to make it much simpler. and tested, and pushed [09:30:12] 2419. [09:31:19] i think mine is simpler logic-wise, but yours is cleaner. [09:31:59] I can rewrite mine to not repeat the enable fuse stanza by changing the variable to yes if not set and FUSE is present, if that would make it simpler. [09:32:13] (an if/else in the not found case, instead of two branches both wanting to PKG_CONFIG) [09:32:43] I don't think we should warn if there's no FUSE and --enable wasn't given, actually. [09:34:06] Yeah, that's cleaner. [09:35:00] Just pushed a new version to 2419. [09:35:50] With a typo. [09:35:51] Sigh. [09:36:09] well, version 1 worked. i'll try version 3 [09:37:04] in 2 hours i will have rhel6 beta. so i can't look at the KBUILD_MODNAME issue til then, sadly [09:38:50] ok, well, merged [09:40:52] you need devel on it, or just testing? I have an rhel6 beta vm, and I thought someone else in here had one, too [09:41:14] Hm, I dropped lpia support from the latest openafs packages because I thought Ubuntu dropped that experiment, but I just got a bunch of build failure messages for lpia builds. [09:42:46] basically, i want to know with the default kernel why (apparently) KBUILD_MODNAME woul be needed but not set [09:43:07] like, pr_debug (via a few other macroes) needs it, but kbuild should set it. but apparently it doesn't [09:43:42] The ktime test is bogus on 64-bit, yay. Building to test a patch now. [09:43:48] oops [09:44:15] -1 != 0xffffffff if you have a 64-bit long. [09:44:36] "oops" [09:45:48] --- Kevin Sumner has left [09:45:57] --- meffie has left [09:46:25] --- Kevin Sumner has become available [09:52:31] worst case, MakefileProto.LINUX.in gets -D"KBUILD_MODNAME=KBUILD_STR(${LINUX_MODULE_NAME})" presumably [09:53:39] --- meffie has become available [10:03:40] --- kaj has become available [10:04:22] --- rod has left: Disconnected [10:14:47] first doc change for dafs fileserver name changes. comment omn 2425 and i will take it into account before i do the salvager and volserver (unless someone beats me to it) [10:16:56] hmm, we want separate manpages for each of these? [10:18:07] There are tradeoffs either way. [10:18:37] If we're going to maintain them both in parallel for a while, synchronizing the changes to anything that isn't specific to one implementation or the other is going to get a bit annoying. [10:18:52] But the wording is more complicated if we document both in the same man page. [10:19:17] Having separate man pages also makes it easy to just throw one of them away when we drop the old binary, but I'm not sure how soon we'll do that. [10:19:59] as proposed, the dafileserver page just documents the dafs-specific options; not all of them [10:20:24] Ah, hm, no, I don't think that's a good idea. [10:20:29] and for non-fileserver... the other binaries differ very little in terms of options and such; I don't see much benefit in separate pages for those [10:20:49] for fileserver/dafileserver... I don't know; there's a lot of different stuff [10:21:03] Although reading the page, it does read fairly well. [10:21:26] Ponder ponder. I'm torn. [10:21:33] I think it's probably better to document them both in the same man page, though. [10:21:34] the dafileserver manpage describes only what is dafleserver specific. everything else is in the fileserver man page [10:22:41] preprocessing the man pages sounds nicer every day :) [10:22:48] the other option would be just to install the old page as fileserver.8 and dafileserver.8, maybe? [10:22:49] i dunno [10:23:38] Me either. [10:23:47] I added that comment to Gerrit. Maybe other people will weigh in. [10:34:05] --- shadow@gmail.com/owl9DEC967B has left [10:45:50] --- shadow@gmail.com/owl12A6B1A3 has become available [11:03:01] --- pod has left [11:18:42] --- jaltman has left: Replaced by new connection [11:18:43] --- jaltman has become available [11:21:18] --- pod has become available [11:21:26] --- kula has become available [11:34:29] --- JSund has left [11:34:58] --- JSund has become available [12:05:15] --- jaltman has left: Replaced by new connection [12:05:16] --- jaltman has become available [12:05:53] --- jaltman has left: Disconnected [12:07:17] --- jaltman has become available [12:09:43] --- jaltman has left: Replaced by new connection [12:09:44] --- jaltman has become available [12:15:16] --- jaltman has left: Replaced by new connection [12:15:17] --- jaltman has become available [12:19:04] --- jaltman has left: Replaced by new connection [12:19:05] --- jaltman has become available [12:23:49] --- mattjsm has become available [12:36:48] --- jaltman has left: Replaced by new connection [12:36:49] --- jaltman has become available [12:42:38] --- jaltman has left: Replaced by new connection [12:42:39] --- jaltman has become available [12:45:43] --- kaj has left [12:45:45] --- jaltman has left: Replaced by new connection [12:45:46] --- jaltman has become available [12:49:09] --- kaj has become available [12:56:53] --- jaltman has left: Disconnected [12:57:04] --- jaltman has become available [13:03:05] --- jaltman has left: Disconnected [13:03:13] --- jaltman has become available [13:33:35] --- Jeffrey Altman has become available [14:05:29] --- kaj has left [14:16:29] KBUILD_MODNAME: the problem seems to be that kbuild does not set it when an object will be part of several modules. which is our case for files linked in both libafs and afspag [14:17:50] --- kaj has become available [14:19:28] --- mattjsm has left [14:30:27] --- kaj has left [14:37:47] --- Kevin Sumner has left [14:38:15] --- Kevin Sumner has become available [14:39:47] --- shadow@gmail.com/owl12A6B1A3 has left [14:40:46] --- meffie has left [14:50:47] --- meffie has become available [15:06:44] --- Simon Wilkinson has become available [15:08:58] I suspect we should just give up on afspag [15:09:10] It's not like it works on current Linux, anyway. [15:09:38] Derrick: I went through RT to find tickets with patches, I didn't quite finish the set, but I'll send you my findings so far. [15:11:14] Summary is: We should probably consider 45867, 88966 to fix our default RC file on Linux. [15:11:34] I think we need 104732 so that a ptserver with supergroups enabled doesn't blow up on a 64 bit machine. [15:11:55] 56530 looks plausible - it depends on whether we think its still necessary. [15:12:17] --- meffie has left [15:12:42] --- Simon Wilkinson has left [15:18:03] --- Jeffrey Altman has left: Replaced by new connection [15:18:04] --- Jeffrey Altman has become available [15:22:18] --- meffie has become available [15:27:38] --- Jeffrey Altman has left: Replaced by new connection [15:27:40] --- Jeffrey Altman has become available [15:27:45] --- Jeffrey Altman has left [15:30:45] --- Jeffrey Altman has become available [15:30:49] --- Jeffrey Altman has left [15:32:27] --- Simon Wilkinson has become available [15:35:53] --- Jeffrey Altman has become available [15:40:08] --- Jeffrey Altman has left: Replaced by new connection [15:40:08] --- Jeffrey Altman has become available [15:40:15] --- Jeffrey Altman has left [15:41:12] --- Jeffrey Altman has become available [15:41:20] --- Jeffrey Altman has left [15:43:58] --- Simon Wilkinson has left [15:54:27] --- deason has left [16:06:48] --- jaltman has left: Replaced by new connection [16:06:49] --- jaltman has become available [16:08:43] --- jaltman has left: Replaced by new connection [16:08:48] --- jaltman has become available [16:21:21] --- shadow has become available [16:22:33] Got the email; I'll check it out shortly. Currently at a benefit for the Brashear Association. (yes, they're a relation; no, that's not why I'm here) [16:23:15] And marc: so you think it's reasonable to just set KBUILD_MODNAME? [16:23:36] --- shadow has left [16:28:20] --- shadow has become available [16:28:30] --- shadow has left [16:30:10] --- shadow has become available [16:36:43] --- shadow has left [16:51:51] --- jaltman has left: Disconnected [16:53:10] --- Jeffrey Altman has become available [16:53:49] --- Jeffrey Altman has left: Replaced by new connection [16:53:50] --- Jeffrey Altman has become available [16:55:00] --- Jeffrey Altman has left: Replaced by new connection [16:55:01] --- Jeffrey Altman has become available [16:55:28] --- Jeffrey Altman has left [17:03:11] --- jaltman has become available [17:09:15] --- shadow@gmail.com/owl56357369 has become available [17:20:50] --- kaj has become available [17:22:19] --- jaltman has left: Disconnected [17:30:50] --- jaltman has become available [18:01:10] --- jaltman has left: Disconnected [18:24:35] --- steven.jenkins has left [18:26:17] it may be reasonable, although it gives duplicate definition warnings because we don't always need it. but it's not enough to get things to compile cleanly [18:27:31] --- steven.jenkins has become available [18:28:37] we could, i suppose, #ifndef #define it before including errqueue.h. [18:28:42] how does it fail? [18:28:48] (my download still hasn't finished) [18:28:51] there's more breakage from the kbuild makefiles because modname is either not set or set to multiple values [18:29:51] ok. so if we just set it before including errqueue.h, in the source for rx_knet.c? [18:32:57] hmm, ok for KBUILD_MODNAME. But we also need to deal with the other error from the value of DEBUG_HASH and DEBUG_HASH2 [18:35:05] I don't know what the value is used for. The values of those variables are defined using a script. with the bad value for modname, they get defined to a "Usage: " message... [18:40:59] just tested defining KBUILD_MODNAME, DEBUG_HASH and DEBUG_HASH2 just before including errqueue.h, and that builds OK [18:51:32] i assume they can be set-ifndef? [18:51:43] ifndef-define [18:52:33] for DEBUG_HASHx they are defined, but have a bad value [18:53:12] ugh. so ifndef KBUILD_MODNAME, #undef the other 2 and define all 3? [18:55:01] that works, but I'm not sure what the effect of setting bogus values for those might be. they are used by the "dynamic debug" feature. they are 2 different types of hash on the module name [18:56:11] --- Kevin Sumner has left [18:56:15] can we eject dynamic debug somehow? i mean, i know this will be turned off for rhel6 final, but we should try to deal [18:56:44] looks like it's turned on in the regular config, not just the debug one [18:56:50] ugh [18:57:14] --- Kevin Sumner has become available [19:10:02] for a given module name we could run the hash utility by hand, get the right value and hardcode it in [19:11:20] --- clc31 has become available [19:11:32] that's probably reasonable [19:12:18] so for "libafs", #define DEBUG_HASH 54, #define DEBUG_HASH2 3 [19:13:40] afspag, libafs and openafs would be the 3 cases [19:13:48] i guess afspag won't matter [19:15:12] afspag is the thing that does syscall interception and PAG management on NFS clients? What's wrong with it? [19:15:19] --- jaltman has become available [19:15:20] nothing. [19:15:36] it doesn't use rx_knet.c so it won't need this change [19:17:10] rx_knet.o is listed in its object files.. otherwise we wouldn't have this problem because it would have a single parent module [19:17:46] openafs: DEBUG_HASH: 17 DEBUG_HASH2: 14 [19:18:19] I would rather that if I read fileserver(8) I get to know everything I need to know about running the non-dafs fileserver, and if I read dviced(8) or whatever, I get to know everything I need to know about running the dafs fileserver, including things that are common to both. That does mean a lot of common text, which at some point calls for a way to have common text, but I think it's important that each program have a complete man page. [19:18:35] afspag: DEBUG_HASH: 23, DEBUG_HASH2: 44 [19:19:04] and of course, that was from hours ago; that's what I get for going back in time to catch up. Still, it's what i think. [19:19:52] in afspag's object files? looking... [19:20:28] yes, objects that are not used by afspag don't have the KBUILD_MODNAME issue [19:20:40] hm. yup. [19:20:51] I think it does use rx_knet.c. It intercepts AFS system calls and handles them via rmtsys. It needs rx for that [19:23:12] hm. this is going to be a mess to do [19:23:28] perhaps we could name the object files differently for afspag.ko? [19:23:53] yes, renaming the object files would probably work [19:24:38] aaaand, i have an iso. [19:24:39] .. and would be a cleaner solution, if it's not too hard to do in the makefiles [19:26:22] Makefile.common.in bloat but dead simple [19:26:49] a second set of names, for AFSPAGOBJS and a second set of .o targets [19:34:55] just add a suffix or prefix I guess. you could just do the problem files - only rx_knet? [19:36:12] i figured might as well do em all, to future-proof us. i am working on it [19:38:39] Ok, I tried just using rx_knet_afspag.o for AFSPAGOBJS, added a rule for it, and that builds fine [19:40:32] eh. know what? that's probably simplest. we can change again if we need to [19:42:16] and we do it for afs_user.c [19:42:49] I can push what I tested [19:44:24] should it be rx_knet_afspag, or rx_pag_knet, or do we care [19:44:43] see if 2435 looks right [19:45:06] (rx_pag_knet to match what happens to afs_pag_user) [19:45:12] that looks fine [19:45:59] and the log message is correct? [19:46:37] going backwards: jhutz, documentation, you should comment in gerrit. i... i wonder if we can usefully preprocess the pod files to accomplist that without having 2 copies of the same text [19:46:47] marc, sorry s/log/commit/ [19:48:12] What should I be comparing ExpirationTime to to know when a callback has expired? (as in, RXAFS_FetchStatus returns an AFSCallBack, which has an ExpirationTime) The ExprationTime numbers are low (like 14400-ish), so are they relative to a certain time? [19:48:23] i figured.. :) the message looks Ok to me. [19:49:42] chaz, uh... the fileserver returns the value as "expires N seconds from now", for whatever now [19:49:55] so we use that to make an expire time we store: tvc->cbExpires = CallBack->ExpirationTime + osi_Time() [19:50:11] where osi_Time is "now" [19:50:38] note that this means we assume it's a low enough latency connection that that value works [19:51:50] okay, cool. i'm working through the libafscp stuff -- looks like that part will need some work. it's storing the returned ExpirationTime and then comparing it to time() to see if it's expired...obviously won't work. [19:52:29] seems like no [19:52:45] store ExpirationTime + time() and then, golden [19:53:59] --- mdionne has left [19:57:13] something like that, yes. it's storing the whole AFSCallBack at the moment. to prevent confusion in the future, may go with storing the "now" as well -- 'cause if we mess with the stored cb info we might, down the road, pull it back out of the hat and not realize it's been modified. [19:58:04] You also assume that the system clock isn’t stepped, I guess? [19:59:00] asked that in the gerrit incident. hopefully the pod2man author will comment :) [20:04:40] i guess that's the working assumption, yes [20:31:19] thanks Derrick -- looks like that bit is working now [20:49:49] Hm, in this freebsd status report, do I want to ask people to submit detailed bug reports to port-freebsd@openafs.org ? [20:50:21] you could. we're getting there but there are still issues. a novice would probably be in over their head. [20:51:01] Yeah, we only want detailed reports (/from technical people), though I'm not sure if I can word things so as to make that clear. [21:10:46] --- clc31 has left [21:21:48] --- Russ has left: Disconnected [21:56:04] --- reuteras has become available [23:09:05] --- kaj has left