[00:08:53] >Simon: where did you read that Microsoft require %Ip [00:09:08] Well, it's what you said when you implemented AFS_FMT_PTR for Windows [00:09:32] > I is only valid with the integer types [00:11:24] In which case, our Windows definition of PTR_FMT is wrong. [00:15:26] If Windows really does just use %p, then we can probably kill the macro (unless we want to still have the option to cater for platforms with braindead printf() ) [00:39:44] --- Russ has left: Disconnected [01:56:51] --- haba has left [02:29:20] --- haba has become available [03:00:18] --- phalenor has left [03:31:26] --- phalenor has become available [04:57:10] --- abo has left [05:32:49] --- Jeffrey Altman has left: Replaced by new connection [05:32:50] --- Jeffrey Altman has become available [05:39:08] --- abo has become available [05:54:52] --- sxw has become available [06:08:46] --- haba has left [06:10:03] --- haba has become available [06:16:52] --- sxw has left [06:22:56] --- cclausen has become available [06:40:19] Is some brain cell here that rembers in whar version the salvager-does-not-die patch was put in? [06:57:30] --- stevenjenkins has left: Lost connection [06:58:33] --- summatusmentis has left [07:01:43] --- summatusmentis has become available [07:03:04] I think a brain cell led me to "salvager avoids corrupting length of directory objects. (111585, 107767) " in 1.4.8. [07:55:02] --- cclausen has left [08:01:29] --- Russ has become available [08:05:05] --- reuteras has left [08:13:09] --- deason has become available [08:20:56] --- stevenjenkins has become available [08:21:01] --- deason has left [08:21:11] --- deason has become available [09:01:23] --- sxw has become available [09:10:20] simon, do you know anything about the windows codebase? [09:11:23] haba - are you referring to the chdir() patch so that the cores don't get littered all over the filesystem? or something else? [09:11:51] i think he meant something else. [10:04:57] --- haba has left [10:12:09] Jake, what do you need? [10:13:52] --- sxw has left [10:20:27] --- mmeffie has become available [12:20:44] --- mmeffie has left [12:29:21] hm, has anyone ever had gerrit just... not load? [12:29:36] all I'm seeing is "Loading Gerrit Code Review ..." [12:29:38] It did it to me on my laptop this afternoon. [12:29:51] But I couldn't repeat it. [12:33:18] Speed hounds might be interest in ... http://homepages.inf.ed.ac.uk/sxw/2d-read.png [12:35:04] the javascript error console dealie says http://gerrit.openafs.org/gerrit/gerrit.nocache.js?content=somesha1 couldn't load... hmm, does it usually? [12:35:25] trying it with or without the query string does indeed 404 [12:35:51] Gerrit still works for me. [12:36:02] (and that URL doesn't) [12:36:06] some browsers handle that better than others [12:36:21] best practice is don't load it if it doesn't exist [12:36:25] Hmmm. But having visited that URL, gerrit now fails to reload. Stranger and stranger. [12:37:15] Quiting and restarting Safari fixes my problem, though. [12:37:20] --- edgester has become available [12:37:31] My suspicion is that Safari is doing something it shouldn't. Is your problem on a Mac too? [12:37:48] Hi guys, I'm back for more punishment [12:37:52] it is, but it fails with all opera, firefox, and safari [12:38:01] And permanetly fails? [12:38:29] after it broke in opera, I tried in firefox and it worked for a little while, then that stopped working [12:38:32] Hmmm. shift-reload breaks me now. Let me take a look and see what the server log is saying. [12:38:39] then I tried safari, but that didn't work even on the firsrt try [12:38:42] I have an odd problem, ls /vicep? shows .vol files that do not show up in vos listvol [12:40:40] that's not odd. several things could cause that [12:40:59] This is related to my vos rel problems [12:40:59] obvious one: a header with no actual volume contents on the disk [12:41:13] it probably is related, yes [12:41:14] I had to fix a different volume this morning [12:41:17] deason: Should now be fixed [12:41:31] the files are 76 bytes [12:41:44] ok. so? [12:42:03] The files are dated Aug -3-4, 2006. [12:42:15] I'm wondering what the cause was, and how to I clean all of these up [12:42:53] rather, how do I identify similar files and clean them up [12:43:25] got full salvage logs? [12:43:32] or ability to do a full salvage? [12:43:40] full salvage++ [12:43:44] or vos zap.. [12:43:59] vos zap is potentially dangerous unless you're sure there shouldn't be data there [12:44:05] vos zap can often clean up the majority of those, but you may encounter some that need a salvage. [12:44:08] true, very true. [12:44:10] like, skipping right to "hey, let's nuke data" is questionable [12:44:18] I can't salvage right now [12:44:32] Simon: hmm, still happening to me, but let me try clearing a few things... [12:44:33] should salvage clean all of those up, though? [12:44:33] how many of these do you have? (10s, 100s, 1000s?) [12:44:36] well, it's hard to say what's not real just randomly, for inode. for namei this would be easy [12:44:54] 7 on one server. I don't konw how many on other servers [12:44:57] ah. [12:45:21] manually salvage per volid might solve your problem. might not, though, depending on what caused those to not be online. [12:45:32] --- deason has left [12:45:39] --- agoode has left [12:45:51] and salvaging one vol at a time won't shut down your fileserver (of course, if you make a typo and accidentally salvage the whole partition or server, that's a different story) [12:46:29] hold on, I might have zapped the wrong thing [12:46:36] deason: Hmmm. I think our gerrit war may be missing a file. [12:46:53] --- abo has left [12:47:00] --- abo has become available [12:48:28] why are you zapping anything? [12:53:02] me? [12:53:25] I mistakenly zapped trying to clean out things [12:54:35] --- deason has become available [12:54:39] Hmmm. gerrit go boom. [12:54:57] okay, yeah, I'm pretty sure a 503 error isn't a problem on my end, heh [12:55:38] It'll be back just as soon as the apache proxy module stops caching the fact it went away ... [12:57:39] --- shadow@gmail.com/owl1A15EC91 has left [12:59:09] also fwiw, the android gerrit site does have a /gerrit/gerrit.nocache.js [12:59:16] Yeh. So do we. [13:00:24] ugh, I killed a vos restore because it was taking too long, now the volume is busy. How do I make it unbusy? [13:01:52] --- haba has become available [13:04:38] edgester: If the volume is locked, vos unlock, if there is still an active transaction, vos endtrans (which only arla vos can do) or wait for timeout. [13:04:59] Gah. Russ isn't around. We'll just have to wait ... [13:06:23] Okay. We're back. [13:06:49] --- dev-zero@jabber.org has left [13:06:49] I don't know entirely what had happened there, but jetty had decided to stop serving the entire '/Gerrit/gerrit/' directory. [13:08:27] haba: thanks [13:20:28] edgester: Linked this one static for you: /afs/stacken.kth.se/home/haba/bin/i686-unknown-linux/vos.arla.static It has the endtrans. Your mileage may vary. [13:20:40] Back -- you need me to restart Apache? [13:20:46] Or did it figure it out? [13:20:54] * Russ was playing volleyball. [13:20:55] It worked it out. [13:21:08] I wanted to assist it with its thinking, but it got there in the end ... [13:21:29] You should be able to do a sudo /etc/init.d/apache2 restart on that system when you need to, btw. [13:25:35] Oh excellent. Thanks. [13:25:39] Can I do the same with jetty? [13:25:58] Yup! [13:33:30] --- agoode has become available [13:37:18] haba: thanks. I got the transaction stopped and restored the volume that I zapped [13:37:28] 3 windows client crashed so far [13:37:47] what do you mean by "crashed"? [13:37:58] service hung [13:38:17] --- mmeffie has become available [13:38:25] what do you mean by "service hung"? [13:39:20] RW volume was zapped from under them. I vos dumped the RO vol and vos restored to the R/W on the same server with the same volid, vos rel failed when I tried, then I vos moved to a different server, removed one replica, added another and had a successful vos release [13:40:39] were they actively writing to the volume when it went away? [13:41:18] "fs memdump" the clients and send me the afsd.dmp files [13:41:50] where by "send" I mean place in afsd [13:41:53] in afs [13:43:40] I don't think they were writing [13:44:14] I forgot to add that I aborted the first vos restore, had it timeout, then had a successful vos restore before the failed vos rel [13:46:25] the file server is either reporting volume offline or volume does not exist. I'm not sure how to interpret "service hung" with that information. Can you describe what the client is failing to do? or doing? [13:46:49] I wasn't at the console of the clients when it happened [13:49:23] --- dev-zero@jabber.org has become available [13:56:57] --- edgester has left [14:01:27] --- mmeffie has left [14:33:26] --- shadow@gmail.com/owlBFB92CD4 has become available [15:05:48] --- edgester has become available [15:10:39] There's a lot of good stuff in Marcus's patch, and the summary of changes is very useful. Unfortunately the patch itself is one big monolithic hunk. [15:14:35] Its a work style that fails miserably in the open source space [15:15:05] --- dev-zero@jabber.org has left [15:15:32] Yeh. But now I have to decide if it's worth picking stuff out of that patch. [15:15:56] --- deason has left [15:15:57] In particular, if the fix for the des keytype stuff works, then it's probably easier to do that than to work it out for myself. [15:16:11] the question is not if it is worth it but where does someone find the time to do so [15:16:21] I'm afraid of trying to fix the DES keytype stuff. [15:16:34] It has a well-known history of breaking on obscure platforms if one tries to eliminate the warnings. [15:16:43] agreed [15:16:49] --- deason has become available [15:16:49] Not the stuff within the DES directory. [15:16:56] That's voodoo and should be left well alone. [15:16:59] Oh, something other than what I was thinking about. [15:17:37] But the fact that we have multiple different types for a des key, and use them interchangably. Sometimes as they are. Sometimes as a pointer, occasionally as a pointer to a pointer, and when we're feeling really evil, as an array. [15:17:48] They all point to the same object, but boy does the compiler scream. [15:32:15] given the amount of time he has spent analyzing the results, he could have spent that time breaking up the patches and submitting them [15:32:52] the ctime issue should have a fix in RT [15:33:49] the afsconf_GetCellInfo side-effect to lower case the cellname is intentional. cellnames are case-insensitive [15:34:22] Perhaps we should stick that list in the wiki, and review it there? [15:35:56] ctime issue should have been posted to gerrit already by me (but has not); sorry about that [15:36:09] assuming I'm thinking of the right thing [15:36:18] I think you are [15:39:53] all of the interesting surprises are items that are either misunderstandings, we have fixes for in RT or can be fixed with one line changes and submitted to gerrit [15:41:36] I have a tool that can be used to remove all of the trailing blanks [15:41:53] I don't think we should make formatting changes in the middle of a release series. [15:41:57] and convert from tabs to spaces (if we want) [15:42:00] Agreed with Simon. [15:42:25] I agree with the idea of doing it, though, however iff we add a Git commit hook somewhere or some similar mechanism that ensures we don't reintroduce whitespace problems. [15:42:27] definitely not for 1.4 [15:42:31] And if we do it at the start of 1.6. [15:42:38] Er, 1.7. [15:42:48] Yeh. That sounds reasonable. [15:42:50] Hm. Although that does cause problems for pullups. [15:42:53] do we have gitweb setup? [15:42:58] git.openafs.org [15:43:00] We really want to do it in the stable branch at the same time. [15:43:04] Indeed. [15:43:04] Or something. Not sure. [15:43:13] We could do it in HEAD immediately before branching 1.5 [15:43:23] 1.6 you mean? [15:43:28] thanks, I confused gitweb with gerrit [15:43:29] Yes, sorry. [15:43:33] Yeah, that makes sense. [15:43:33] either tabs -> spaces or the other way around... most files have that irritating mix that is just awful [15:43:39] * Russ is all in favor of changing all tabs to spaces. [15:43:46] Not sure. [15:43:49] you don't like tabs? [15:43:55] I like our current formatting. [15:43:55] * Russ loathes tabs. [15:44:05] tabs are 8 in some folks editors and 4 in others [15:44:13] So get a better editor? [15:44:21] that's why they are good, you can adjust the tabstop yourself [15:44:32] The problem is that you can't with the OpenAFS code. [15:44:34] * Russ thinks that's silly, but it's bikeshed painting. [15:44:36] so people don't fight over how large an indentation is in the source [15:44:50] So I'm not going to argue it. :) [15:44:58] Yeh, but we do 4 spaces, tab, tab+4 spaces, 2 tabs, and so o. [15:44:59] not that we have that problem at all... quite the opposite, really [15:45:00] Just register one vote for changing all tabs to spaces, and if I get outvoted, that's fine. [15:45:02] that doesn't work if your style doesn't use full tab stops for indentation [15:45:11] Simon: Yeah, we're using default Emacs tabination. [15:45:16] If your tabs aren't 8 characters your code doesn't indent properly. [15:45:19] Which is sort of the worst of both worlds. [15:45:26] --- stevenjenkins has left [15:45:35] However, I think we should consider just how badly we will break everyone with a local tree if we made the change. [15:46:03] I would personally like tabs, but just spaces would still be a hell of a lot better than now, I think all would agree [15:46:31] Personally, I don't think that the benefits in changing are worth the pain of doing so. [15:47:07] http://www.jwz.org/doc/tabs-vs-spaces.html is basically exactly my opinion. [15:47:32] --- abo has left [15:48:18] 1.4.11 release submitted to freshmeat [15:48:20] --- abo has become available [15:48:35] However, jwz doesn't address the idea of having the code appear differently to different people. [15:48:47] Which I guess I understand, although it seems kind of bizarre to me personally. [15:48:55] --- stevenjenkins has become available [15:49:19] Bah. I guess if 1.4.11 is released, people are going to want to know how to do pullups ... [15:49:28] I probably just advocate for it because I like 3-space tabstops, and I know nobody else wants that :) [15:50:06] * Russ used to use 2 but got used to 4 a long time ago, since that's what everyone seems to use except Linus. [15:50:30] Linus uses 8 intentionally to make editing deeply nested code obnoxious, which is an interesting approach. [15:50:46] that is a good post. For Kermit we had a rule, no ASCII-9 in the files because we couldn't control the editors that were present on the hundreds of OSes that the software had to be edited on. [15:51:43] I prefer 4 with context sensitive 2. For example, labels are outdented by 2 from the current indent level. [15:52:04] Yeah. I can do either that style or pure 4. [15:52:28] I tend to use pure 4 just because it matches Emacs BSD style and I don't have to customize for the context-sensitive 2. [15:52:31] would it be possible to put in modelines in files, to specify whatever behavior? [15:52:38] since 1.4.11 is out, does that mean CVS is officially dead? [15:52:39] I was thinking about that. [15:52:43] Although that's also hitting a lot of files. [15:52:49] But if we're detabifying, we're hitting them all anyway. [15:52:55] It's a bit harder to do the modeline automatically. [15:53:07] modelines won't work for editors that do not support them. [15:53:08] and if we don't detabify, this would be less invasive than changing all that whitespace [15:53:21] yeah, it only helps for Emacs and vim, basically. [15:53:22] I don't think hitting a lot of files is a problem. Generally, the change will be in the first line, and patch (git, whatever) will fuzz around it. [15:53:30] Changing all the whitespace, not so much. [15:53:37] The tricky part is getting the stuff that isn't *.c or *.hy. [15:53:39] *.h. [15:54:04] Well, patch -l is your friend. [15:54:14] But yeah, it's going to really screw any pending patches. [15:54:18] Or private trees. [15:54:21] --- agoode has left [15:54:21] --- abo has left [15:54:22] with modelines, it would seem less important to hit all at once [15:54:23] --- haba has left [15:54:30] that's less likely to break everything [15:54:37] Yes. [15:54:38] --- abo has become available [15:54:40] its not going to help anyone that works on windows [15:54:42] --- agoode has become available [15:54:47] Is there anything that helps on Windows? [15:54:53] Can we set a global policy in a project file, for instance? [15:54:56] besides removing the tabs, no [15:55:08] What's the default Windows IDE behavior with this? [15:55:14] okay, so 'remove the tabs and specify the modeline' generall works for all? [15:55:23] --- haba has become available [15:55:24] both in terms of tab stops and interpretation of ASCII 9? [15:55:55] While we're touching every file, do we want to do anything about our copyright notice? [15:56:03] Or planning to do so in the future, I should say. [15:56:04] --- stevenjenkins has left [15:56:25] That reminds me. We really should think about sorting the rx snafu. [15:56:28] its a shame we can't change the copyright once every line in a file has been touched [15:56:38] * Russ hehs. Yeah. [15:56:48] which rx snafu [15:56:51] ? [15:56:55] Although we have no legal framework to hold the copyright yet. [15:57:00] are you referring to just... rx itself? [15:57:07] The RX contribution from IBM wasn't under the IPL. [15:57:22] We put it under the IPL when we cleaned up the licensing in one of the earlier commits. [15:57:24] oh, wrt copyright, not just "rx sucks" [15:57:26] that we should fix. didn't you generate a list of the contributors? [15:57:29] The RX code was originally public domain. [15:57:46] I can give you a list of everyone who has touched one of those files, yes. [15:57:52] I suspect we don't always have accurate author information, but thankfully since AUTHOR was in the template, it should be relatively close. [15:57:52] I give my permission for a relicense of my contributions [15:58:02] Me too, of course. [15:58:06] Me too! [15:58:08] I'm sure Tom will [15:58:10] and Derrick [15:58:41] The problem is, the stuff is still very much a derived work. Like all the man pages, for instance. [15:58:46] relicense under.... [15:58:58] We can only relicense clean rewrites. [15:59:08] --- stevenjenkins has become available [15:59:12] Legally, we even need to not refer to the previous code for it to not be a derivative work. [15:59:18] Indeed. Or things that are a separate 'module' and therefore clearly not a derived work. [15:59:37] I was wondering more whether we just wanted to say something other than Copyright 2000 IBM, not relicensing which is a different can of worms. [15:59:47] I was under the impression that that was unclear. That what 'derivation' implied depended very much on the judge in the particular case. [15:59:49] Since there's certainly material that's copyrighted more recently than 2000. [16:00:05] Not that it matters in Berne countries, really. [16:00:35] Plus, by the time OpenAFS is 90 years old, I'm sure the copyright duration will be 214 years. [16:00:45] --- abo has left [16:01:02] Depends on the juristiction... [16:01:42] --- abo has become available [16:03:16] > /me is all in favor of changing all tabs to spaces. me too [16:03:24] the only relicensing we can do is for things that were public domain when we received them [16:04:21] * Jeffrey Altman heading out. will be back before midnight EDT [16:08:59] > the change will be in the first line, I believe emacs and vim both also support modelines in the last line, which is more useful for some kinds of files. [16:11:07] Just emailed openafs-devel a pointer to my 'make reads faster' patches, which more than double read speed on Linux for some applications. Comments welcome. [16:11:28] Does it point into gerrit? [16:12:09] Nope. The changes aren't ready for the tree yet, so they aren't in gerrit. [16:12:37] (I guess we really need a 'pre-gerrit' where things can be placed for comment, but without the expectation they'll make it in) [16:12:47] Yeah, that'd be cool. [16:13:00] That could just be a place we let people push things, no? [16:13:15] Be nice to have all the pretty review stuff that Gerrit has. [16:13:30] Indeed. I was thinking that just as I wrestled with git format-patch. [16:13:32] I thought that was the point of appending stuff already in Gerrit? [16:13:33] simon: as in, another gerrit instance, or just the per-dev repos or something that was mentioned before? [16:13:51] * Russ isn't a big fan of per-dev repositories. [16:14:04] I really dislike the Github model of Git workflow. [16:14:09] or just another repo for all, whichever [16:14:10] If we want pretty review stuff, then we'd really need another gerrit. [16:14:31] I just meant "another gerrit" or "git repo(s)" [16:14:50] I just don't want to blur gerrit's purpose. I think it's really important that everyone understands that pushing to gerrit means "here's my change, it's ready for the world" [16:14:52] My concern about per-dev repositories is basically that they're going to diverge and never get rebased and then become a merge nightmare. But yeah, your point there is good, sorry. Didn't mean to go off on a rant when you were talking about other things. :) [16:15:00] I think per-dev repositories are overused. OTOH, Linux used to have a model where in addition to the main tree, there were several developers who maintained ongoing trees in which they baked things before eventually pushing them to the main tree, and having repos for those is useful. [16:15:36] Yeah, if we had enough people where we had Linus's merge problem and needed people to roll up changes to push to him, I could see it. I doubt we're going to have that problem, though, really. [16:15:37] is having another gerrit really that much different than using the same gerrit on different branches? [16:15:48] or is there no mechanism for separating branches to that degree? [16:16:02] --- agoode has left [16:16:08] So, a second gerrit with less strict policies about review and who can push, but which pushes into the "real" gerrit? [16:16:12] Oh, branches can be completely separate. You can even have completely different projects on different branches. [16:16:32] --- agoode has become available [16:16:53] No, I think it would be a completely separate instance. With the same review policies (anyone can review), and that wouldn't submit anywhere. [16:17:13] But I'm not sure how worthwhile it is. [16:17:41] For example, at this stage, I'm not after line by line review. I'd like people to read about the general ideas, and see if they think they're reasonable, and I'm providing the code as an illustration of that. [16:17:47] Or do we really want gerrit to gain a feature, such that I can push things that are not ready, have them clearly marked that way (thus not wasting people's time), and then flip a switch when I think it's ready for review and gatekeeper attention. [16:18:06] I'm not sure I see the difference between a separate instance and the same instance on a different project/branch/whatever.... [16:18:09] --- stevenjenkins has left [16:18:25] jhutz: Perhaps, yes. [16:18:54] I think we probably need to think a bit more about what problem we are trying to solve, and perhaps get some more experience with the tools we have. [16:18:59] deason: The tricky part with a separate branch is that you want that branch to stay very in sync with master, or patches based on the current master will upload a bunch of cruft to the other Gerrit view to resync its repo with master. [16:19:02] deason: It's easier to have a clone that automatically tracks the 'main' repository than it has to have a branch to track another branch. [16:20:12] could you have gerrit have another project that tracks the same branch, but is not allowed to push onto it? [16:20:35] We could, yes. [16:21:01] but a switch that jhutz mentioned sounds better... I'm just thinking aloud [16:21:19] But I think what jhutz says is correct. We should work out what the problem we're trying to solve is, and get more experience (and more users) for the tools we've already got before we start adding something new. [16:22:06] --- stevenjenkins has become available [16:22:16] well, we still have the issue right now of where to put non-ready patches [16:22:35] we've got the list, RT, or everyone's specifical personal spot [16:22:37] --- abo has left [16:22:44] The list, I think. [16:23:05] Or at least, the list is where discussion of the ideas around these unfinished patches should take place. [16:23:21] RT is the wrong place for that, as the number of people who can contribute is limited. [16:23:25] --- abo has become available [16:23:45] hmm, so what is RT's place? just "thing X broke"? [16:24:11] Yes. [16:25:42] okay [16:26:23] I think it's a place for users to say 'thing X broke', for developers to say 'I've noticed that X is broken' [16:26:42] these purposes should probably be explained on -devel [16:26:49] It's a great place to record intent - 'X is broken ... I'm working on a fix' [16:26:57] (to avoid overlapping effort) [16:27:32] Well, I'm just stating personal opinion. I guess we need to work out collectively what the New World Order looks like. [16:28:42] For example, to take the ZFS thread on openafs-devel [16:29:18] We should have a bug that says "Fix the partition lock". And whoever it is that's working on doing so should add something to the bug that says "I'm looking into this". [16:29:32] Otherwise, we might well end up with two people doing the same work, at the same time. [16:29:37] -info, but yes, point taken [16:29:47] That really sounds like a good use for RT. [16:29:50] Since you can set owners of tickets. [16:30:05] However, that has the few people can comment problem. [16:30:11] We should fix RT. [16:30:31] Where some values of fix may be 'replace' [16:35:26] RT's place is bug tracking gerrit's place is submission of code for review and inclusion in OpenAFS the mailing list and this room are the right venues for design and discussion. But please don't post huge patches to the list. [16:37:03] > We should fix RT. There are two kinds of broken here. One is performance, which will be improved dramatically by deleting all the crap and moderating incoming mail. The other is access control policy, which can just be changed if people decide what it should be instead. [16:37:51] Yes. I think the 'what it should be instead' is the fundamental issue. I'd be happy with anyone being able to do pretty much anything, in the same way as bugzilla.mozilla.org is configured. [16:38:00] Hrm. Maybe I missed this, but what ties my account in gerrit to an email address? Is that verified in some way? [16:38:34] Your initial OpenID ties it in in the first place. I think we trust what the IdP says, though. [16:39:04] After that, if you want to change your email address to something else, you have to click through a challenge response thing by email. [16:39:32] I wonder when the web spammers will discover openid [16:40:07] When it becomes significantly more widespread, I suspect. [16:40:28] We'll be able to work around that easily by requiring someone confirm new OpenID accounts, I suspect. [16:41:23] Yes. We could easily ramp down the permission set that you get by just registering with the site if it became a problem. [16:41:24] Perhaps. See, I'm thinking about how it might be useful to provide an openid-authenticated interface to autocreate RT accounts. The biggest barrier to commenting today is you have to get someone to create you an account. [16:41:57] The 'deltas/edit' application already uses OpenID, so I've got some (recent) knowledge of writing that ... [16:42:52] --- deason has left [16:43:25] --- deason has become available [17:02:33] --- mdionne has become available [17:08:55] would someone mind giving feedback to /afs/tproa.net/users/s/u/summatusmentis/patches_gsoc2009/rx_GetLocalPeers.diff ? [17:21:04] the purpose is to export rx stats so as to allow ranking based on said stats [17:21:22] the ranking stuff will come [18:32:32] summatusmentis: function names usually start on their own line for the implementation, like afs_int32 rx_GetLocalPeers(stuff) { [18:32:57] just stylistic reasons? [18:33:07] I'm also unsure why you return everything in network order, but I don't know about how this is used... [18:33:46] I guess it's just the style; it's made it easier for me to grep for where a function is actually defined at times [18:33:52] though I don't know if that's it's purpose [18:38:36] the purpose of this function is only to expose the rx stats, I don't believe network order is necessary [18:38:53] expose the rx stats to the cache manager* [18:51:05] Yeah, usually it's so grep ^function does what you want. [18:59:47] --- edgester has left: Replaced by new connection [20:06:49] --- mdionne has left [20:11:34] --- dev-zero@jabber.org has become available [20:29:07] --- dev-zero@jabber.org has left [20:29:36] * Russ gets a WebAuth release out and goes home. [20:29:39] --- Russ has left: Disconnected [20:30:42] have a good evening [20:37:56] --- dev-zero@jabber.org has become available [20:48:53] --- Russ has become available [21:40:17] --- dev-zero@jabber.org has left [22:00:54] You would think that someone would have written a Docbook source reformatter that would wrap text at a sane column. [22:01:08] Indeed, according to the Debian archive, multiple people have. Except none of them work. [22:01:08] someone probably did, and it's long forgotten [22:01:36] * Russ would really like to reformat all of the Docbook manuals sanely, since the really long lines are hard to review in Gerrit. [22:01:59] yes, gerrit hates to line wrap or to give you scrollbars, apparnetly [22:02:59] when I was editing the docbook I searched for something and came up empty handed [22:03:33] I could just go do it in Emacs. [22:03:41] I wonder how well Emacs's multiparagraph fill will do. [22:03:50] It may be that sgml-mode is smart enough to just fix it. [22:05:45] Hm, yup, fill-nonuniform-paragraphs pretty much just works. [22:06:00] Any objections to me submitting a pure reformatting patch? [22:06:37] do it [22:06:44] It doesn't clean up the tagging, but it will at least make the lines shorter. [22:07:29] --- dev-zero@jabber.org has become available [22:09:17] Ah, no, that doesn't quite work either. [22:09:18] Sigh. [22:09:39] It's close, but it messes up program listings and some tag constructs. [22:12:28] --- deason has left [22:20:02] that is the problem I had. We need a tool that understands the syntax enough to skip the sections we wish to have pre-formatted [22:20:21] --- Jeffrey Altman has left [22:38:55] venus/fstrace.c is a mess. [22:40:37] for ugly, look at kdump [22:41:09] I tihnk we have a complete embedded copy of the NLS message catalog stuff. [22:41:13] From who knows where. [22:41:49] all behind #if defined(AFS_OSF_ENV) && !defined(AFS_OSF20_ENV) [22:42:00] am I reading that right? [22:42:08] Does that mean "Digital UNIX before 2.0?" [22:42:16] it means osf pre 2.0 needed it [22:42:25] Can I delete it? [22:42:32] in the head? sure [22:42:33] 'cause that's, uh, old. [22:43:45] Also, this function just makes me happy. [22:43:48] char * rmalloc(int n) /*---- n: the number of bytes to be malloc'ed ----*/ { char *t; t = (char *)malloc(n); if (!t) printf("Failed to get mem\n"); return (t); } [22:47:52] Oh, hey, look, fstrace has its own copy of the AFS system call code. [22:47:54] Awesome. [22:48:01] 'cause we needed more of those. [22:54:21] --- Jeffrey Altman has become available [23:03:51] Why is fstrace not linked with libsys to use its afs_syscall implementation? [23:04:00] beats me [23:04:21] Wait, no, it is linked with libsys. But it still has its own afs_syscall. [23:04:23] * Russ is so confused. [23:05:01] when i am less busy i will look. i didn't think fstrace horribly sucked [23:06:46] It's not a big deal, and not something to spend a lot of time on. I'll figure it out and get it untangled. [23:07:49] Ah, I think it's because it's not doing pioctls and is instead doing its own separate system call. [23:09:25] Ah, yes, indeed. [23:09:34] And libsys doesn't export afs_syscall. [23:09:45] So it's reimplementing the same logic that's in lpioctl. [23:16:38] Russ, are you still up? [23:20:36] Yup! [23:20:52] Finishing finding a clean way of fixing the install path for the fstrace message catalog. [23:20:54] --- reuteras has become available [23:21:29] as you may have guessed I'm going through all of Marcus' interesting surprises. [23:22:13] if you become idle and wish to help. let me know. they are mostly trivial items [23:23:01] Okay. I was going to at least review the ones you already submitted, but possibly not until tomorrow. [23:23:09] I should go to bed before long. [23:23:13] me too [23:23:37] but I want to show how trivial it is to just report the bugs when you find them [23:28:11] Yeah.