[00:17] jelmer: oh ... that doesn't seem to have helped. even with sbronson@ in the https:// url, it still finds "sbronson" as the user to authenticate as ... [00:23] jelmer: also, apparantly subversion always defaults username to the user's account name? [00:23] maybe you should allow multiple authentication attempts ... [00:48] I'm having a bit of difficulty doing a bzr get through sftp. I have a repository at example at come and use this command: bzr get sftp://user@example.com/path/to/repo. But I'm getting a EOF during negotiation error. [00:48] I'm on windows and have installed paramiko and pycrypto. [00:50] Would it be a bad idea just to download the repo from example.com using sftp? Could I still merge between them in that case? [00:57] hmm [00:58] Hmm? [01:01] lifeless: you there? [01:02] * lifeless checks [01:02] yup, I seem to be [01:04] lifeless: hi. Anything else you wanted me to do about that bzr check failing stuff? [01:04] oh uhm, what bug number was it [01:05] 356028 [01:05] bug 356028 [01:05] Launchpad bug 356028 in bzr "bzr check fails with KeyError" [Undecided,New] https://launchpad.net/bugs/356028 [01:07] kingos: no, sorry I had lost it from my queue; I've assigned it to me [01:12] lifeless: thanks. [03:37] thunderbolt: example.com - really? [03:37] Yes. ;-) [03:37] test taht you can just do plain sftp [03:37] your own private example.com? [03:38] Thanks mlh_, I can plain sftp to it. [03:38] Yeah, it's actually a domain for my work, it's been changed to example.com to protect the guilty :) [03:39] me <-- not bzr expert but [03:39] you can rsync I believe or scp -pr it all and have it work [03:39] Cool! [03:39] Thanks mlh_. [03:40] Now the only problem is pushing it back, I guess I could make a patch/bundle. [03:40] * thunderbolt was using bzr for deployment of some django onto a website. [03:40] perhaps try bzr+svn scheme instead of sftp [03:41] bzr+svn ?!?!? [03:41] bzr+svn? Do you mean bzr+sftp? I tried that with similar results :( [03:41] I mean bzr+ssh [03:41] Ah. [03:41] right. [03:42] * thunderbolt had no luck with that, either :( [03:42] jelmer: I just got the weirdest error. I tried to branch a foreign Subversion branch on machine A (which doesn't have the branch). bzr errored out telling me the branch had no revision. Yet I pushed to that branch on machine B not one minute ago >.> [03:42] Oh dear. I'll have to pass to 2nd level support then :-) [03:43] * thunderbolt chuckles [03:43] using for deployment is widely frowned upon [03:43] but everybody (including me) still does it [03:44] It is? [03:44] Why is it frowned upon? [03:44] I think you should have a make/scons/maven/... target dir then rsync that target dir to the deployed area [03:44] because things like .bzr don't belong on a website [03:45] there's a chance of inadvertantly exposing too much info [03:45] mlh_: Ah, in my setup .bzr lives in a directory that isn't web accessable. [03:46] so you think :-) [03:46] It's along with all my Django config stuff (including database passwords) [03:46] So if .bzr is accessable, so is my database, so I'm screwed. [04:13] there is also a 'bzr upload' plugin [04:13] mlh_: ^ thunderbolt ^ [04:14] which is expressly for web deployment [04:14] lifeless: Ah, cool! [04:14] Neat, kinda like rsync. [04:23] seems like my checkout got stuck :( [04:24] * sidnei tries once more === ja1 is now known as jam [04:25] jam: hi [04:25] hi lifeless [04:25] just made it into San Jose [04:25] I figured I might try to ping rockstar [04:28] jam, hi. [04:28] jam, are you in the hotel? [04:28] hey rockstar, how and where did you want to meet up tomorrow? [04:28] rockstar: yeah [04:29] jam, um, shall we meet down in the breakfast area for breakfast tomorrow, say, 9ish? [04:29] sounds good [04:29] jam, I'll probably want you to review my notes for the BoF tomorrow night. [04:30] certainly [04:32] jam: I'm analysing the index insertion issue [04:32] lifeless: you may want to check out lp:~jameinel/bzr/btree_with_bloom [04:32] (just pushed up after I got off the plane) [04:32] I haven't done any benchmarking, etc. [04:32] jam: thanks; I have one here :) [04:33] jam: I'll be letting you do the tuning I suspect. I just wanted to put some math around it. [04:36] lifeless: so I think branching *from* a repo with 500k nodes isn't terrible, because we make large requests (50k roots, 200k internal, 200k leaf) [04:36] and that code sorts the request, etc [04:37] the issue I was seeing was that *insert* was doing it one at a time [04:39] yes [04:39] its NlogN disk reads [04:40] which is better than N^2 but not that much [04:40] its superlinear and as its read-and-parse operations per key, it sucks [04:42] e.g. [04:43] 95% of time is 1.54M queries (this test data has some duplicates) [04:43] 3% of time is 1.49M insertions [04:43] 1.38% is the final read-and-serialise to the finished index. [04:51] jam: one thing I observed was that bloom probing rehashed [04:51] jam: we need to have a __contains__ that allows passing in the md5 [04:51] lifeless: we *can*, but I might point you to BloomMurmur [04:51] which is about 5-10x faster [04:52] I'll leave that in your hands [04:52] my mail will be finished in < 10 minutes I think [04:52] so in my testing [04:53] ''.join() was 160ms [04:53] str in BloomMurmur [04:53] was 500ms [04:53] for ~ 100k keys [04:53] sorry 24k*20 = 500k keys [04:54] I was going to look at allowing tuples, so we don't have to do a string join [04:54] anyway, at that point, I wasn't sure if it was specifically worthwhile to cache [04:54] versus the complexity [04:55] jam: but you could just put them in a dict! [04:55] * SamB_irssi teases [04:55] SamB_irssi: :) [04:55] SamB_irssi: we're talking about the implementation of a dict on disk :) [04:55] right, the whole point here is to use a lightweight memory object that lets us drop the actual keys, etc to disk [04:56] (yeah, my "suggestion" *was* intended as utter point-defeating nonsense ;-) [04:58] rockstar: I was looking over the schedule, and it seems the first sessions start at 8:30, do you want to meet up a bit earlier? [04:59] * rockstar looks at the schedule again. [04:59] jam, oh dear. Yea, how 'bout 730 or so? [04:59] rockstar: sure [04:59] http://www.mysqlconf.com/mysql2009/public/schedule/grid [04:59] btw [05:00] ah, I was reading Monday it looks like 8:30 is keynotes [05:00] but still good to go to [05:00] jam, yea, I have that bookmarked. For some reason, my brain skipped the keynotes. [05:02] jam, maybe we ought to skip out on some talks and go visit this: http://www.mysqlconf.com/mysql2009/public/schedule/detail/7791 [05:03] rockstar: yeah, I think we will [05:03] jam, it seems to kinda go all day long. [05:03] yep [05:03] "a community organized event designed to share and improve the essential skills required to participate in collaborative, free and open online projects" [05:06] "bug reporting", "grepping", and "patching"? [05:06] jam: mail sent [05:15] Peng_: fwiw, i like the idea of a single LoggerheadConfig object being created in serve-branches and being passed around the various wsgi app instances [05:16] (and i still haven't reviewed your branches properly) [05:16] mwhudson: Okay, cool. :) [05:16] mwhudson: is my search fix landed? [05:16] lifeless: yes, i think so [05:16] good [05:16] now, to get it working on bzr-playground.gnome.org [05:16] also, that needs to start pulling git :( [05:24] What options do I need to pass to bzr init to test the latest branch format? And will LP support it? [05:24] BasicOSX: do you mean the development6-rich-root format? [05:24] lifeless: yes [05:24] --development6-rich-root [05:24] is what I thought [05:24] -format=development6-rich-root [05:24] lp won't support it [05:25] and its a one-way migration, stuff you commit in there can't be moved to regular formats [05:25] LP won't support it? Why not? [05:25] Peng_: lp is running 1.13 [05:25] won't support it "now" is probably what he means :-) [05:25] right [05:25] Oh. I misunderstood. [05:27] :) [05:27] Peng: it's ok, I spent the whole day being misunderstood. Endless debate over Free (speech) and free (beer) and the GPLv1, GPLv2, and GPLV3 [05:27] BasicOSX: at a conference? [05:27] and zomg gpl1?!?!?! [05:28] lifeless: /blush in my WoW guild :-P related to Curse/WoWI blocking WoWMatrix from downloading Addon source code [05:28] heh [05:28] curse are very odd [05:29] waaay of topic here, but 3 of the addons I use are GPL and I stated that the GPL requires, if request, access to the source code [05:29] BasicOSX: #ubuntu-wow :P [05:30] to anyone, and those addons are only on curse, so in effect they are not compliant with with the GPL then religion kicked in and everything go ugly :-) [05:30] heh [05:30] endless rounds of emacs vs vi [05:30] windows vs linux [05:31] GPL vs BSD [05:31] so they don't have to allow anyone without the binary access, but blocking by client is arguably a violation yes [05:31] open source vs Free software vs commercial software [06:19] s/commercial/proprietary [06:23] Cue endless debate over *that*. :D [06:54] not really, commercial just means 'for money' :) [07:01] lifeless: if they distribute binaries without source, they need to provide source access to anyone on request. [07:01] if everyone who had access to the binaries also had access to the source at the same time, then they could deny requests from third parties [07:04] they wouldn't be required to support any particular method of source distribution, as long as they use "a medium customarily used for software interchange" [07:39] jamesh: its more subtle than that [07:40] jamesh: and not very interesting [08:10] I don't know when it started, but I have some branches that when I commit changes to them locally, they automatically want to commit to the remote (bzr+ssh) repository and this causes some delay for me when committing that I would rather not at times [08:10] how do I turn that off? ;-) [08:10] bzr unbind [08:11] thanks === quicksil1er is now known as quicksilver [08:22] jfroy: hi [08:22] jfroy: this seems similar to the problem you were reporting earlier [08:40] jelmer: hey [08:41] I filed a bug to report it === davidstrauss_ is now known as davidstrauss [09:16] jelmer: Ah I see, yes the 2 bugs may be duplicates. [09:16] Didn't catch my eye at fist because the exception and error message are different. [09:46] hi folks [09:48] is it possible to create a PHP comment block on top of a PHP file via Bazaar, with the whoami name + last commit? [09:49] we seem to be having regular issues where we get "too many concurrent connections" - it's gotten to a point where one branch is unusable. [09:50] is there any reason that this could happen on a regular basis [09:50] Lambo_: bzr has nothing like cvs's $Id$, if that's what you're talking about [09:51] bob2, i just want bzr to insert the last commit in each file [09:51] Lambo_: nope [09:59] this just doesnt seem to work. and it's getting extremely frustrating [10:04] to lp? [10:04] Mez: didn't we debug this last week? [10:04] bob2: we have filters now that can do $id$ if people want [10:06] lifeless: yes, we were, and we've just worked out the issue. [10:06] Zen's "magic filtering" [10:06] hi guys, stumbled into #354036 again [10:07] I added the traceback to the ticket [10:07] which doesn't like us using SSH, and kills the connection if we send across a certain control char [10:07] Mez: !!! [10:07] lifeless: after changing the route on the box, it worked straight away [10:10] lifeless: such a horrid issue :( [10:10] Mez: might be worth a faq entry on answers.launchpad.net/bzr for this [10:11] lifeless: really? I think it's just our setup to be honest [10:11] Mez: well, up to you. Be nice if someone else runs into it to have an easy answer ;) [10:11] anyhow, ciao, /wave everyone [10:12] "change ISP" isn't a valid answer [11:01] hi all [11:01] is it normal that bzr locks when using bzr send with gmail ? [11:30] asabil: did someone answer? [11:32] poolie: Nope. [11:32] You didn't miss anything. [11:44] asabil: no, it's not normal [11:44] what happens if you hit Ctrl-C? [11:52] poolie: nothing happens when I hit ctrl-C [11:52] it's still hung? [11:53] but later I get a error: (110, 'Connection timed out') [11:53] like this? https://bugs.launchpad.net/bugs/364462 [11:53] Ubuntu bug 364462 in bzr-email ""connection timed out" causes full stack trace to be dumped" [Undecided,New] [11:53] the smtp server address seems correct to me: smtp.gmail.com [11:55] poolie: well it is the same issue, but unrelated to the bzr-email plugin [11:55] which I don't have installed [11:55] right [11:55] i've just commented that i think it's actually a core issue [11:58] thanks [12:21] asabil, i'd guess there was just some transient outage on your network causing it to be unreachable [12:21] or indeed it's not impossible there was a glitch at google [12:21] unless it's reproducible [12:22] poolie: to me it seems like gmail only accepts smtp over ssl [13:30] Am I completely misunderstanding what bzr missing does? I run it between two branches where one is massively ahead of another, and it says "Branches are up to date." [13:30] Anyone going to the Jaunty release party in London? [13:31] I'd expect it to show me committed revisions that are present in one branch but not the other [13:32] Arse, that's Thursday [13:39] VSpike: it should list both sides; are you running bzr mising other-url? [13:39] VSpike: that is what it's supposed to do [13:40] hi poolie [13:40] gnight all [13:40] hi lifeless [13:40] Ah. Perhaps I see the problem. Both branches were bound to the same parent, but one had not had bzr update run [13:40] So bzr st showed a very old revisions [13:41] but doing the bzr missing seemed to do a bzr update at the same time, which sort of makes sense. === kkubasik_ is now known as kkubasik [13:51] VSpike: I don't see how that makes sense [13:51] are you trying to call bzr missing with two arguments? [13:51] * SamB doesn't have any idea what that would do [13:52] Nope.. I mean I was trying to compare two branches, so I cd'd into one and did "bzr st", then "bzr missing ". It said they were up to dat.e [13:53] So I cd'd into other and did "bzr log | head" and "bzr missing " and it said up to date - yet the head revisions were totally different [13:53] sorry, in the above replace "bzr st" with "bzr log | head" [13:54] But when I cd'd back into , and did "bzr log | head" again, it showed the same head. And bzr info revealed they were both checkouts of the same branch, a third one. [13:55] jelmer: ping [13:57] VSpike: ah. checkouts! right. [13:57] I forgot about heavyweight checkouts. I ran into some strange mergy issues with them ... [13:57] ... and haven't used 'em since [14:00] I started using them because I'm terrible at remembering to commit, let alone push :) [14:00] ah [14:01] well, I'm more interested in not losing changes to unintended "merge"s [14:01] maybe I should instead figure out how to reproduce the issue and report it ... [14:18] jelmer: shouldn't the subvertpy docs be online ? [14:20] jelmer: also, the "epydoc" target doesn't seem to work right in an unbuilt subvertpy source tree [14:20] hi. how do i set the parent branch? [14:21] you can use pull --remember [14:21] or you could edit the .bzr/branch/branch.conf file [14:22] can i use some short-names for other remote branches? [14:22] huf: with the bookmark plugin, bm: [14:22] huf: next to the builtin :parent, :public, :submit etc [14:22] LarstiQ: and where do you get it ? [14:23] where can i find docs on those builtins? [14:23] and what each is for, and when each is chosen as a default [14:23] SamB: without look at http://bazaar-vcs.org/BzrPlugins I'd guess lp:bzr-bookmark [14:23] jam, hi? [14:27] cool, i think i got it [14:27] thanks [14:29] LarstiQ: I don't suppose you can bookmark just a URL prefix? [14:29] you can [14:29] then you can use bm:foo/path/to/branch [14:29] neato [14:29] where foo is e.g bzr+ssh://example.com/path/to/repo [14:30] or it could be svn+https://dosemu.svn.sourceforge.net/svnroot/dosemu, no? [14:30] yes [14:31] of course, I still haven't gotten authentication to work for me ... [14:31] what happened to the bzr versions? what came after 1.5? [14:31] presumably 1.6? [14:32] oh, 14... i cant read. [14:32] then .7, .8, .9, .10 ... [14:32] yes yes, i see now [14:33] * SamB wishes the NFS at his school's cs department was faster ... [14:35] * SamB wonders why SVN doesn't check out files in alphabetical order [14:38] SamB: It does it in inode order [14:38] (SVN FS inode, not normal-world fs inode) [14:38] awilkins: but the inodes don't exist yet [14:38] oh [14:38] some kind of progress indicator would be nice ... [15:00] SamB: yeah, the docs could be online, it's just too much effort to keep them up to date [15:03] jelmer: buildbot? cron job? [15:09] SamB: No python on the machine that my homepage is on [15:10] SamB: It also seems like it's not too much effort for whoever is going to use subvertpy to build it [15:13] I dunno. I find building docs to be easy, but copying and pasting the URL into Firefox a pain. :D [15:46] SamB: I relatively frequently do things like `bzr missing :parent/../foo` [15:47] Hey, that works? [15:47] I totally didn't know that. === sabdfl2 is now known as sabdfl [17:16] jelmer: well, maybe you should at least debianize the docs ? [17:17] because I don't currently build subvertpy myself on my own machine, and the one at school lacks all the doc-building tools ... [17:17] (I also don't think I have room for the build-deps on my personal machine) [17:17] SamB_irssi: any chance you can file a wishlist bug? [17:17] ah, sure [17:18] * SamB_irssi wonders if there is something wrong with Debian's subversion ... [17:20] SamB_irssi: why is that? [17:20] jelmer: is there an up-to-date ppa for subvertpy? [17:20] SamB_irssi: no [17:21] jelmer: oh, well, it doesn't seem to cache my auth credentials for that repository I've been having so much trouble with ... [17:21] SamB_irssi: svn itself (not bzr involved) ? [17:21] even with the options explicitly set [17:21] yes, SVN [17:21] though I admit I haven't actually done a commit [17:21] just revision property manipulations [17:22] do you perhaps have subversion configured to not store password / [17:22] s? [17:23] jelmer: not entirely, but I'm trying not to! [17:23] and I can't see how to make subversion mumble aloud what it's doing ... [17:24] SamB: strace? ;) [17:25] LarstiQ: strace doesn't tell me what configuration settings subversion has determined ... [17:25] or what it's basing it's decisions on ... [17:26] jelmer: oh, I guess you want me to file a debian bug huh ... [17:27] SamB_irssi: yeah [17:27] almost filed it in launchpad ;-) [18:06] jelmer: oh, did you catch that I still haven't managed to commit to sf.net via bzr-svn? [18:14] * Kinnison yays as his answer to problem 67 on projecteuler.net is correct [18:34] jelmer: any thoughts on the branch problem? It [18:34] *It's kind of a bummer not being able to get that branch out of svn, especially since it started as a bzr branch :p [18:53] hi. I want to use bzr for our project, but due to internal policies, we have to keep all our code inside a svn repo. In order to cope with that, I thought of using a post-commit hook, that would just push the changes to the svn repo. In that way, every time someone commits to our bzr main branch, it gets pushed to the central svn repo [18:54] how can I do this? (what is the best way to achieve this?) [18:54] from my research I see I can write hooks for branches, but not for the whole repo... (i am using a shared repo) [18:54] ricardokirkner, is it just you who wants to use bzr? [18:54] do I have to write a hook for each branch in the repo, or is there a way to make this easier? [18:55] pygi, no.. its a team of people [18:55] otherwise I would just use bzr-svn [18:55] I want to use bzr.. just push to svn as a side-effect [18:56] hmmm [18:57] jelmer: I've noticed a message about refusing a basic authentication challange now ... [18:58] ricardokirkner, you can set global hook for a user [18:58] would that work? [19:00] mhhh [19:01] You could just use a bzr central branch, and cron pushing it to svn Every So Often(tm)... [19:01] I dont think so, as that would require that every team member commits with the same user [19:01] If everybody's going to actually use bzr, that could work.... [19:01] fullermd, that might work... [19:01] it's not 'clean' ,. but it might work [19:01] thx [19:02] * fullermd specializes in 'not clean' 8-} [19:02] :-) [19:10] how difficult would it be to implement a "poor man's PQM" with a pre-commit hook on a shared branch? [19:11] phinze, are you using Launchpad? [19:12] no i'm in a dev group at a university [19:12] trying to see if there's a way to get PQM-like functionality without significantly changing the workflow of the developers [19:13] phinze, well, I guess it wouldn't be too hard. Tarmac is more lightweight than PQM, but is (currently) specific to Launchpad. [19:13] ahh that's why you ask [19:13] Tarmac doesn't use a hook though. [19:13] ricardokirkner: does the policy state how soon the code must be in the svn repo? [19:13] rockstar: you think it would be better to look into modifying tarmac? [19:13] * phinze knows nothing about tarmac [19:13] phinze, bzr branch lp:tarmac [19:15] phinze, it's undergone some big changes since 0.1, but I don't see a direct way currently to have it be non-launchpad specific. If you do, I'd be glad to review and merge it. [19:15] phinze: james_w has recently worked on a poor man's PQM [19:17] LarstiQ: thanks, james_w: ping? [19:19] rockstar: cool, rooting around in it now... so it's based off the Launchpad concept of " Branch Merge Proposals"? [19:19] phinze, yes. [19:20] phinze, actually, the best way to do a poor man's PQM with Tarmac would probably be to create a new TarmacScript [19:21] yeah that makes sense [19:22] okay here's a dumber question about hooks [19:22] given a pre-commit hook in a BzrBranch... is there a way of failing the commit? [19:24] * LarstiQ would assume raising an exception [19:24] or is the hook only allowed to spin off side effects and not get in the way [19:25] * LarstiQ looks at the docs [19:25] * phinze is looking as well, and not finding... [19:25] phinze: http://paste.ubuntu.com/155496/ [19:26] james_w: you come in and answer all the questions! :) [19:26] raise errors.TipChangeRejected("The testsuite failed") [19:27] phinze: it can raise any exception [19:27] phinze: hmm, mention of this should be more clear in `bzr help hooks` [19:27] that one says "I didn't go wrong, I want to cleanly stop the operation please", which means bzr formats it differently [19:28] it's specific to the tip_change hooks though [19:28] I don't think it works for pre_commit [19:28] gotchya [19:28] well tip_change is good enough in our case [19:28] we're using append_revisions_only on the shared branches anyways [19:29] errors.HookFailed [19:29] might still not be entirely correct, given its docstring [19:30] in fact, tip_change is probably more correct [19:32] well this definitely gives me some stuff to play with... rockstar, LarstiQ, james_w: thanks [19:32] phinze: if it suits you we could turn the above in a project and work to make it useful to more people [19:33] it was a quick hack the other day after I get fed up of committing broken stuff myself :-) [19:33] haha that's exactly what brings me here with the interest [19:33] i would definitely be interested in that [19:33] * phinze is still in the process of making the snippet work ATM [19:33] my snippet? [19:34] yeah, just shuffling some branches around here [19:34] install it as ~/.bazaar/plugins/testrunner.py [19:35] then edit ~/.bazaar/locations.conf to ahve [19:35] [/path/to/branch] [19:35] ... hmm so it needs to be installed under the user's home dir who will be doing the tip change? [19:35] pre_change_branch_tip_test_command = ./run-tests [19:35] any way to install it system-wide...? [19:36] phinze: sure, bzrlib/plugins [19:36] this will be going on a shared server to protect a branch shared by many users using bzr+ssh [19:36] you can set it to a parent directory, and the same command will be run for all users under that directory [19:36] cool [19:36] phinze: "bzr version | grep bzrlib" [19:36] and install it in the plugins/ directory there [19:36] you will need to tweak it to use branch.conf as well in that case [19:36] otherwise each user would have to edit their locations.cond [19:36] gotchya [19:41] command = config._get_best_value("pre_change_branch_tip_test_command") # ...ok? [19:42] james_w: ^^ [19:44] phinze: I think so === ja1 is now known as jam [20:42] phinze: how's it working for you? [20:43] james_w: working great! [20:43] cool [20:51] hmm ok [20:51]