[00:07] lifeless, you know what would be cool? Making bzr-stats have similar indexes to bzr-search. So to make it easy to get stats, and be able to get something line "how many lines X commiter has added" [00:11] (in a cheap way, not processing everything each time) [00:11] beuno: yah [00:35] hi people - are there any script to fetch svn:externals with bzr-svn? [00:37] pinax.hotcluboffrance.com uses a lot of external apps, I think I'll write a quick script to fetch each one with bzr... [00:53] beuno: pushed, mail sent [00:54] I'm off, at least for a few hours (birthday party), seeya later [00:58] Verterok, cool, have fun [00:59] beuno: thanks [01:34] Hi. I've pushed, with sftp, a branch to a web server and I can browse to it at http://www.studentrobotics.org/~stephen/jsdndide/local/ - what do I use to branch from this over http from another machine? I've tried every variant on branch with that URL I can think of! [01:38] engie: You seem to have a shared repository in use on that machine, which is not exposed via http [01:38] I'm guessing you have one in your home dir [01:39] and Apache is only serving things under public_html [01:39] So you need to "bzr init-repo public_html/" or possibly "public_html/jsdndide" before you do another "bzr push" [01:40] OK. Is this because the local branch is under an init-repo directory on my local machine? [01:40] engie: no, there is a repo on the remote machine [01:40] otherwise we would have created one with the branch [01:40] sorry, with the "push" [01:42] engie: It is a good thing to have a shared repo on the other machine, it just needs to be somewhere that people can access it through http [01:42] Ohh, looks like I accidentally pushed a .bzr up there when I was trying things out yesterday [01:42] That machine doesn't have bzr on it [01:43] engie: you don't need one, you can do "bzr init-repo sftp://host/~/public_html" === eMxyzptlk is now known as eMxyzptlk[away] [01:43] Is there any benefit to doing that? Does it share history between multiple branches? [01:46] Ahh brilliant, that's working, thanks [03:33] jelmer: so what do you think [03:33] RAOF: go to http://200.127.6.219:8080/bazaar/bzr/changes [03:34] RAOF: and type in the search box [03:40] lifeless: Is there really only one change there matching 'lambda'? [03:41] looks like [03:42] So, the find-as-you-type thing is good. And it's nice and fast. [03:42] The formatting seems a little off? [03:42] yes [03:42] being worked on [03:42] Hm, it also doesn't much like "refactor" as a search term. [03:42] hey, spare time, hacked up, new search engine etc [03:42] It flips between anything matching 're' and 'refactor' :) [03:43] beuno: http://200.127.6.219:8080/bazaar/bzr/revision/33?q=lambda doesn't search-complete? [03:43] RAOF: uhm, not sure what you mean about flipping [03:43] lifeless: Oh, I thought you were after comments/queries/criticisms. If you'd like unadulterated praise, I can provide that, too :) [03:43] RAOF: I'm not after either actually :) [03:44] RAOF: you can throw peanuts at SLUG on friday if you like [03:44] RAOF: this is just a follow up to the xesam comment in u-m [03:44] Right. [03:44] Aaah, I think I see the problem: Typing r..e..f..a..c..t..o..r gives "refactoring" as the sole hit. [03:44] which comment I can't even remember now [03:45] just that it made me think I should show you this [03:45] Right. Contrary to Amaranth, I think search can be useful. And I can immediately see how this could be nice and useful :) [03:45] yah, this is a tiny code base its searching - the bzr-search plugin itself [03:45] searching bzr's code base will give more suggstions: [03:46] :!bzr search -s refactor [03:46] Suggestions: [('refactor',), ('refactored',), ('refactoring',), ('refactorings',), ('refactors',)] [03:46] But typing 'refactor' really quickly will briefly bring up 'refactoring', and then bring up all matches for 'r' [03:46] RAOF: oh interesting. Thats because its a lot cheaper to search a smaller dataset I suspect [03:46] RAOF: so its getting out of order results [03:47] beuno: you should include in the response to a lookup, the string that was asked for [03:47] beuno: then the client can ignore responses that are not the same as the current search is showing (or perhaps just ones that were done off a shorter prefix) [03:48] s/smaller dataset/smaller number of terms/ [03:50] RAOF: there is a full command line client to bzr-search, if you prefer that [03:52] garh, fugliest site ever: http://python2.near-by.info/index.php?option=com_content&view=article&id=144&Itemid=111 [03:52] I was trying to find pure python b+/b-tree implementations [03:52] but that one sure seems locked away [04:05] lifeless, the javascript isn't included in the revision page. It's a bug :) [04:06] beuno: :P [04:06] and, it does wierd things because it looks for every character, no matter how fast you type [04:06] another know bug [04:07] I have to make the javascript stop propagating if you type before it sends out the request [04:08] right now it queues them, which was simpler at the time [04:08] beuno: not at all [04:08] beuno: my suggestion will make it appear better [04:08] say it asked for - in order - 'r', 're', 'ref', 'refa' [04:08] but the answers come back in the order 'ref', 'refa', 're', 'r' [04:09] thats why it looks wierd [04:09] but more specific searches complete faster, so this is actually expected [04:09] yeah, probably because [04:09] now, imagine that in each answer you include the search at the top of it [04:09] 'r' took longer to process [04:09] so the answer for r will have 'I am a search for 'r' [04:09] and so on [04:10] then the client can do this: [04:10] if the current contents of the widget does not start with the response, ignore it [04:11] if the contents do start with the response, and no drop down is shown, store the thing the response is for in a variable and show the drop down [04:11] yes, that's one way to do it. But that would hammer the servers pretty bad on a large scale [04:11] if a drop down is already shown, and the response is for a shorter query than the drop down is, discard the response [04:12] beuno: this is not a replacement for asking less often; its needed anyway (because under high load this can happen anyway) [04:12] lifeless, right, that makes sense [04:12] (oh, and if the user deletes terms, obviously reset this variable or something [04:12] so it should be a combination of both [04:13] yes [04:13] ok, so *more* javascript to write :) [04:13] :) [04:14] but it's great that we're debugging this so much, it may get in shape to go into trunk sooner [04:14] I'd like to add a few optional plugins to LH [04:14] this being one of the strongest recommended ones [04:16] also, I have plans to get the integrated into bzr-gtk, if no one does it first [04:16] it's even better as a desktop application ;) [04:19] perhaps jelmer will now hes seen it work :) [04:19] beuno: perhaps these glitches (like revision page missing the javascript should be bugs ? [04:20] lifeless, I'd have to create a project for that. Do you think it's worth having a seperate project for it? [04:20] nah just on loggerhead [04:21] oh, sure, that sounds like a good idea [04:22] I'll file the revisions one (and annotate, and the rest) if you file the search order :) [04:22] k [04:23] done [04:24] me too [04:24] File "/home/jml/.bazaar/plugins/svn/core.py", line 25, in [04:24] from bzrlib.plugins.svn.client import get_config [04:24] ImportError: No module named client [04:25] I got that when I tried to branch a svn branch with bzr [04:25] is there a client.py file ? [04:26] there's a client.c file. Apparently I need to build bzr-svn. [04:27] jelmer: ^ [04:29] * jml installs a bunch of stuff [04:30] ok. that version of bzr-svn is not going to help me much. [05:01] now [05:01] Peng: around? [05:03] jml: ping [05:03] lifeless: pong [05:03] loggerhead is down [05:03] do you know if we moved it off vostok yet ? [05:05] lifeless: no I don't. [05:05] dang [05:06] do you know the new machine name it would have moved too ? [05:07] lifeless, AFAIK, the new version of LH will be deployed on wendsday, so maybe that's when they're moving it? [05:07] beuno: these things are unrelated [05:07] beuno: deployments are routine; moving to a new machine is reconfiguration, testing etc [05:08] mwhudson was talking about doing it driday [05:08] lifeless: no, I don't. [05:08] anyhow, It's 5am in London, I'll wait till 7am and ring james [05:10] jml: have you played with kjbuckets ever? [05:10] http://gadfly.sourceforge.net/kjbuckets.html [05:10] lifeless: years ago [05:10] thats more than I; impressions? [05:10] http://bazaar-vcs.org/RobertCollins#head-b2dc9ae46f1860c46f79682db455c3512152d148 [05:11] lifeless: memories more like it. in the end I did my own graph classes. [05:12] lifeless: I recall the API as being a little cumbersome. [05:12] k, thats sufficient to tip me away [05:12] needing to compile would be a burden anyhow [05:14] scary: http://www.pythonweb.org/engine/ [05:16] lifeless: pong [05:16] Peng: do you have that magic to get the new LH running at a subfolder? [05:17] lifeless: Subfolder? [05:17] yeah, so it can fit in an existing site [05:17] mwhudson showed you how last week [05:17] lifeless: Oh. [05:19] lifeless: Here's my serve-branches.py: http://paste.pocoo.org/show/75647/ [05:19] Peng: thanks! [05:19] lifeless: (Then I just used lighty's mod_proxy/mod_proxy_core.) [05:20] Peng: yeah, I have squid for that :P [05:20] :) [05:35] loggerhead is back [05:35] lifeless, are you deploying the new LH in squid-cache.org? [05:36] beuno: not just yet, I'm going to deploy a demo copy at my house [05:41] lifeless, re: bug #242046 [05:41] Launchpad bug 242046 in bzr "ppa feisty bzr and bzrtools not installable" [Undecided,New] https://launchpad.net/bugs/242046 [05:41] is bzr uninstalable too? [05:41] I used PPAs copy feature, and have had to re-upload without it akready for gutsy [05:42] beuno: bzr and bzrtools - both [05:42] * beuno grumbles [05:42] I'll re-upload now [05:42] so the copy tool is not really useful... [05:43] (or I'm using it wrong, which is more probable) [05:44] hrm, that's odd. I *did* re-upload bzr for feisty [05:45] not bzrtools though [05:45] well [05:45] the problem is the dependency on python-central [05:49] buildlog seems to think the right version was used: http://launchpadlibrarian.net/14569048/buildlog_ubuntu-feisty-i386.bzr_1.5.0-1~bazaar1~feisty1_FULLYBUILT.txt.gz [05:51] ok [05:51] may just be bzrtools;) [05:51] let me uninstall that and upgrade bzr [05:51] I'm fixing bzrtools now [05:51] ok bzr is fine [05:52] cool, I'll upload bzrtools in a minute [05:52] sorry about that [05:55] np [05:59] * beuno uploads and waits for the build deamons to wake up [06:01] Bug 241938?! [06:01] Launchpad bug 241938 in bzr "MemoryError on "bzr ci" for 1396 Megabytes file" [Undecided,New] https://launchpad.net/bugs/241938 [06:03] Peng: yah, we should be able to handle such files [06:04] theres nothing in the disk format preventing handling them [06:04] How much work would it take not to read entire files into RAM? [06:06] Peng: for simple storage and tree building, not a huge amount, though doing it without affecting performance for normal files is important [06:08] Peng: for delta storage, diffs, merge its much tricker [06:10] delta storage depends on diffs [06:10] merge could just write .BASE, .OTHER, .THIS and give up [06:10] but three-way merge etc is nice to offer, and that depends on diff [06:11] in a high entropy file the line count is about 1/256 the file length, a 1.4G file is 5.6million lines long [06:12] if we hash the file to (say) sha1, its still 100MB per instance of the file (20*5.6million) [06:12] which is pretty high [06:13] and easily exploited to cause DOS by having many more \n's. [06:21] lifeless, re-uploaded bzrtools, should be fixed, but can you double check? [06:25] beuno: its good now [06:26] :) [07:19] lifeless: hmm, I'll give that a try some time [07:19] lifeless: the loggerhead frontend certainly is very fast! [07:20] jml: you need to run make these days (or ./setup.py build_ext --inplace) [07:20] jelmer: revno 42 should work nicely for you [07:20] jelmer: Is bzr-svn supposed to have any old-style classes? [07:21] Peng: nope [07:21] It has 11. :P [07:21] is there anything in particular bad about old style classes? [07:22] or is it just a stylistic thing? [07:22] For me, it's stylistic. [07:26] jelmer: is there a small svn repo around you run live tests against ? [07:26] jelmer: in particular one that bzr-svn supports :P [07:28] I generally use svn://svn.samba.org/smb-build/trunk [07:29] lifeless: if you've got any repositories in particular that don't work, please file a bug.. [07:29] jelmer: do you support repository.weave_store.get_weave ? [07:30] lifeless: nope, don't support any weave implementation at all [07:30] not as provider, that is [07:30] jelmer: are you working on it? (will be needed for stacking) [07:30] jelmer: fwiw, I have a branch up at http://bzr.mattnordhoff.com/bzr/bzr-svn/trivial/new-style-classes that fixes all of the old-style classes I could find. [07:31] lifeless: I'm planning to work on it, haven't started yet [07:31] jelmer: and can you point me at the bits of bzr-svn that would give access to the svn backing data? [07:31] lifeless: Trying to finish my own python bindings atm [07:31] lifeless: what do you mean with backing data specifically? [07:32] lifeless: *svn python bindings [07:32] Peng: thanks, I'll have a look at merging that [07:32] :) [07:34] jelmer: I'd like to make bzr-search index svn repos [07:35] jelmer: all it needs is a non-colocated branch object supporting the repository graph and versionedfile apis [07:36] Peng: merged, thanks [07:37] lifeless: Getting the versionedfile APIs to work is a fair amount of work [07:37] lifeless: the graph code is all there [07:39] eww [07:39] bzr: ERROR: libsvn._core.SubversionException: ("Can't find a temporary directory", 20014) [07:40] jelmer: ^ hints? [07:40] jelmer: I did '/home/robertc/bin/bzr', 'checkout', '--lightweight', 'svn://svn.samba.org/smb-build/trunk', 'sbm-build-test' [07:44] 4.G free on / [07:44] 198 files in /tmp/ [07:46] hmm, no - what version of bzr-svn is this? [07:46] svn --version [07:46] you may want to try with the latest release just to be sure [07:46] svn, version 1.4.6 (r28521) [07:46] compiled Mar 11 2008, 08:53:49 [07:46] I'm running hardy [07:47] happens just running svn co [07:50] also happens running svn co you mean? [07:51] svn: Can't find a temporary directory [07:51] I mean [07:51] $ svn co svn://svn.samba.org/smb-build/trunk [07:51] svn: Can't find a temporary directory [07:51] time for strace :-) [07:52] garh [07:52] I was hoping you had seen fuckage before :) [07:53] I vaguely recall something like this [07:53] but don't remember the specifics :-( [07:53] Set TMP or whatever, just to check? [07:58] server side fuckage [07:58] http://rafb.net/p/sOwYPM49.html [08:00] the first line, I think, is the key [08:00] read(3, "( success ( ) ) ( success ( 36:e8204dac-47fa-0310-9482-a4e8a03ab89e 29:svn://svn.samba.org/smb-build"..., 4096) = 105 [08:01] sorry, line 5 [08:01] read(3, "( success ( ( ) 0: ) ) ( failure ( ( 20014 32:Can\'t find a temporary directory 77:/home/tretkowski/s"..., 4096) = 170 [08:01] success (failure!) [08:01] hardcoehmm, chroot broken perhaps :-( [08:02] hardcoded paths? [08:02] well, I'm not tretkowski :P [08:02] bzr-svn work with gcode yet ? [08:02] yeah, always has afaik [08:03] k [08:03] I'll hunt for a small project there ;P [08:03] you just need the svn+ hack to work around the fact that bzr aborts as soon as it gets a permission denied [08:05] Oh. [08:06] Hmm. [08:06] With the 0.4 branch, trying to branch http://ack.googlecode.com/svn/trunk/ (with or without "svn+") exits after a second with a return code of 11. [08:07] lifeless: last time I checked bzr-svn was unable to checkout some code from gcode : http://code.google.com/p/waf/ [08:13] asabil: even with the svn+ hack? [08:14] jelmer: hack ? I was using either http:// or svn+ssh:// [08:15] asabil: I don't remember seeing a bug report about that :-) [08:15] asabil: I meant svn+http:// [08:16] I don't think I tried svn+http [08:16] jelmer: what needs to happen to fix the bug? [08:17] lifeless: bzr needs to be somewhat more persistent when probing for repository formats [08:17] rather than aborting the minute it gets something that's not NotBranchError [08:18] jelmer: why? what raises the permission denied ? [08:18] lifeless: trying to open .bzr/branch/format raises a 403 [08:18] rather than a 404? [08:19] yes, afair that was the problem [08:19] hmmm [08:19] so [08:19] I don't think the driver loop should catch more things [08:20] the probe functions should catch whatever other things that neither indicate an error nor a found branch [08:20] I don't think the Bzr branch format catching 403 is very nice though [08:21] if something's got to do it, rather have it be the driver loop [08:23] well [08:23] how can the driver loop tell that 403 means 404? [08:25] have we filed a bug with googlecode that it gives 403 when a 404 is appropriate? [08:25] (I agree that the Bzr branch format catching 403 is not nice). But I also think that the driver catching 403 is not nice [08:26] because, if you had nested branches, typeA and typeB [08:26] ah, and the other reason the svn+ stuff exists is to work around the ssl certificate stuff [08:26] and there is a permission problem on typeB, typeA can be detected incorrectly [08:29] what will address the ssl certiicate stuff? [08:31] there's an ancient bug report open about bzr dealing with self signed certificates [08:38] k, I have a lightweight checkout [08:38] so I can fiddle [08:39] no promises, but if I get something shaping up I'll give you a branch [08:39] now for dinner and games though :) [08:40] cool [08:40] have a nice weekend [08:40] * jelmer goes and removes the last dependencies on the official svn python bindings [08:45] oooh [08:45] congrats jelmer [08:45] they finally did 1.5.0, but still.... [08:45] Would your bindings be usable by other projects? [08:46] Peng: in theory, yes [08:47] however, they're not very well documented and the set I converted is limited to what bzr-svn needs [08:47] that's good enough [08:49] if there are other projects interested in using these bindings as well, I'd happily provide them as a separate package [08:53] I'd rather wait for such projects to ask first though, rather than putting all of the work in now to find out nobody cares [09:27] jelmer: you still support 1.4 w/patches? [09:27] jelmer: (hardy only has that ...) [09:29] jelmer: how do I get a repository id out of a bzr-svn branch object? [09:30] lifeless: yeah, patches aren't even necessary anymore [09:30] lifeless: branch.repository.uuid [09:46] jelmer: and the relative path to the branch? [09:46] jelmer: I'm going to use uuid/branch-path-here/ as a lookaside index path [09:50] lifeless: branch.get_branch_path() [10:08] hi AfC [10:09] lifeless: Good evening [10:14] Is there a command [line client command] which lists all the revision properties / metadata of a given revision? [10:14] I don't think there is today [10:14] some of the data can be pretty big and is not human consumable [10:15] Huh [10:15] gcommit stores per-file commit messages as revision properties [10:16] (oh yeah? What consumes it?) [10:16] I think gannotate or something [10:16] jam will know [10:16] (not important) === yacc_ is now known as yacc [10:33] AfC: speaking of locale, ulrich drepper rejected the latin patch for libc [10:33] AfC: so Ubuntu is carrying it locally :) [10:34] jelmer: ok, I've got to the point where File "/home/robertc/.bazaar/plugins/svn/repository.py", line 332, in get_inventory_weave [10:34] raise NotImplementedError(self.get_inventory_weave) [10:34] is triggered [10:34] jelmer: its kinda slow to get that far :( [10:48] lifeless: Ulrich has always been a hard case. [10:48] AfC: oh yes :) [10:49] he applied his trademark tact and diplomacy [10:49] Maybe if you were to become a Benedictine monk or something. That might add credibility (you know, since you have none). [10:49] :) [10:49] wasn't my credentials, its was the locals [10:53] 'just another made up locale' wa the prhase I think [10:55] Yeah, I can see that. Latin is a lot like Klingon, after all. [10:56] :P [11:01] ok, its working now for commits [11:01] jelmer: [11:04] jelmer: revno 43, if you'd like to tell me what I'm doing wrong that makes it so slow, that would be cool. [11:40] jelmer: ping [11:42] jelmer: where do you want bugs filed, ctrl-C was having trouble stopping 'bzr search test' on http://waf.googlecode.com/svn/trunk [11:46] * Jc2k is tempted to forcefully upgrade parents to ubuntu [11:47] lifeless: will bzr-search work against bzr-mirror.gnome.org :) [11:52] Jc2k: yes, bzr-mirror is native bzr [11:52] Jc2k: I recommend fresh indices with revno 42 [11:53] cool i might give that a spin later [11:59] if you do, you could also try the loggerhead branch that supports it :) [12:00] presumably thats now all shiny and wsgi-ify-arific [12:00] in which case, take me to ur url :) [12:00] yes [12:01] https://code.edge.launchpad.net/loggerhead [12:01] second from top [12:07] this ok for bzr-search: cd ~/.bazaar/plugins && bzr branch lp:bzr-search search && cd search && bzr revert -r40 [12:07] ? [12:07] no need to revert [12:09] 40 was wrong anyway, i meant 42 [12:09] 43 is fine too [12:09] but < 42 will use a different index logic that gives poorer results on searches [12:10] i see [12:10] rar, the search branch of loggerhead hard depends on a bzr-search that is globally installed [12:11] well [12:11] file a bug [12:11] and show me the error [12:11] surely you can fiddle BZR_PLUGINS_PATH ? [12:11] Or run LH as a user with it in his ~/.bazaar/plugins? [12:12] Jc2k: what user does loggerhead run as [12:14] lifeless: at the moment, as me, but in a second im going to move it to x469yq [12:15] pastebin thee rrror [12:17] Jc2k: it should not depend on a global install of bzr-search [12:17] Jc2k: if you can pastebin the error I can advise [12:18] Jc2k: for different accounts, just put bzr-search in the ~/.bazaar/plugins/search dir in the relevant account [12:18] lifeless: sorry, laptop just ran out of power [12:18] lifeless: thats where it is [12:19] this is going to be painful.. *tries to remember copy and paste in putty* [12:21] lifeless: line 20 of loggerhead/search.py [12:21] from bzrlib.plugins.search import errors [12:21] No module named search [12:21] ok, do this [12:21] python -c 'import bzrlib.plugins.search' [12:22] same error [12:22] ok [12:23] for me: [12:23] ah [12:23] got it [12:23] :P [12:24] Does Loggerhead use any global bzr settings for anything, like your username or something? [12:24] Peng: well, it doesn't do commits [12:24] Yeah, but.. [12:24] Peng: but yes, it does load the global settings [12:24] Do any of them affect it? [12:24] dunno [12:25] My only ones are email, editor and LP user. [12:25] I can't imagine any of those affecting it [12:25] Anyway, I'm moving it to its own user, and wondering if I should copy bazaar.conf over. [12:25] Jc2k: what was wrong? [12:26] * Jc2k blushes [12:26] fix was [12:26] Jc2k: also, currently indices are per-branch, not per-repository (because search shouldn't return commits from a different branch), so you probably only want to index trunk; a full text index is a pretty large fraction of the history in the first place [12:26]