[00:02] verterok: currently to_file expects str, if we change it with codecs.getwriter, it would now expect unicode [00:03] verterok: btw, why do you use xml-rpc in java plugin? why not talk directly with xmloputput? [00:04] RenatoSilva: I got lost :) [00:05] RenatoSilva: directly how? [00:05] verterok: currently to_file expects str, if we change it with codecs.getwriter, it would now expect unicode, and that py2 auto ascii encode/decode thing would get place, and would raise errors for non-ascii stuff... [00:05] verterok: I mean for example [00:06] verterok: you want to pull branch's log, then why not you run bzr locally and parse the XML response? why do you have to send an xml-rpc call to a local server, then get the response and parse it? [00:06] verterok: the server runs bzr locally anyway [00:06] RenatoSilva: mainly for performance reasons [00:07] RenatoSilva: the startup time in windows is terrible [00:07] verterok: you mean xml-rpc server keeps the executable in memory? [00:07] RenatoSilva: I'll take a look to the codecs.getwritter, don't have the code fresh enough in my head [00:08] RenatoSilva: right, only oone bzr process [00:08] *one [00:08] verterok: there isn't a way to do this in java? [00:08] verterok: if python does, java does it too ^^ [00:09] RenatoSilva: hmm...what? [00:09] verterok: the plugin could load bzr.exe into memory [00:09] RenatoSilva: we need to use bzrlib API [00:10] verterok: oh I see, bzr-java-lib... [00:10] RenatoSilva: I'm missing something, could you extend? :) [00:11] verterok: you mean start bzr.exe every time you need to run a command is too slow, so you put xml-rpc server in scene for keeping one executable in memory, and run the commands faster, right? [00:12] RenatoSilva: yeap, exactly [00:12] RenatoSilva: take a look to the different *Runner's in bzr-java-lib [00:13] verterok: I don't have the code right now... [00:13] RenatoSilva: the XMLRpc extends the CommandLine [00:14] verterok: my idea was, when the plugin is loaded, it loads the executable into memory.When it wants to run something, then it activates bzr-java-lib, and passes the cached executable, so that bzr-java-lib does not have to start the proccess. [00:14] verterok: both bzr-eclipse and bzr-java-lib would need changes in the code to achieve this [00:15] verterok: however I'm not sure if you can do this kind of thing in Java [00:15] RenatoSilva: how would you load the executable into memory and use it several times? [00:15] verterok: I'm guessing: if python can, then yes java can! [00:15] RenatoSilva: I think it could be possible with JNI...but you know...JNI [00:16] verterok: and how python does? [00:17] verterok: how does the xml-rpc server loads and keeps the executable in memory? [00:17] RenatoSilva: bzr-java-lib executes: bzr start-xmlrpc in a subprocess [00:17] verterok: also, maybe bazaar should have some kind of "daemon" mode [00:21] verterok: oh I see, the server is a bzr instance itself + xmlrpc plugin. When you send a command, the server is "inside bzr", so it does not need to run bzr.exe. I thought for a moment that run_bzr was running the process... [00:21] RenatoSilva: there is the bzr-serivce plugin [00:21] RenatoSilva: right [00:21] :) [00:22] verterok: nice [00:23] verterok: how does it work? you send raw arguments to a net socket? [00:23] RenatoSilva: the bzr-service plugin? or the xmlrpc service? [00:23] verterok: or do you have a formal protocol? [00:23] verterok: bzr-service [00:24] RenatoSilva: jam wrote it, I think it send the command args via the socket [00:25] verterok: hum, what do you think is simpler code, talking with xmlrpc or bzr directly? [00:25] RenatoSilva: bzr via bzrlib? [00:27] verterok: ok, but you're the bzr-java-lib developer too, right? :) [00:27] verterok: not talking about xmloutput anymore hehe [00:27] RenatoSilva: xmlrpc is more flexible in that you can talk to it from almost every language [00:27] one issue with xml though [00:27] verterok: every language? [00:27] not all bzr commit data can be put into xml now [00:28] lifeless: hi [00:28] lifeless: yeah, xml is maybe an unecessary middleware [00:28] RenatoSilva: bzrlib is only usable from python [00:28] hi poolie [00:29] what did you mean by "guess-good"? [00:29] that it had tried to help you ;) [00:29] lifeless: also, things get weird when you send xmloutput stuff through xml-rpc: xml inside xml... [00:30] and "it" is? [00:30] RenatoSilva: the current xmlrpc service is a sort of a hack :) [00:30] bzr-guess [00:30] 're bzr-guess helping you that is good to know' [00:30] verterok: sorry I read bzrlib as "bza-java-lib" [00:31] verterok: what I mean with " hum, what do you think is simpler code, talking with xmlrpc or bzr directly?" is [00:31] verterok: from java code [00:31] verterok: in bzr-java-lib [00:31] oh ok [00:33] RenatoSilva: currently there is no way to talk to bzrlib directly from Java, I'm trying to implement OSLocks in Java, in order to use Jython [00:33] verterok: maybe being able to directly send arguments to bzr, but having the process chached in memory, maybe it would be eassier to java code... [00:33] verterok: I'm talking about service plugin [00:34] RenatoSilva: xmlrpc is quite simple, you just send basic data types and receive the same (Strings, int, HashMap, Arrays) [00:35] verterok: imagine in java code something like Bazaar bzr = Bazaar.getServiceInstance(); then you bzr.run('log -r 10..12') ro so, instead of building an XML and send as xml-rpc [00:36] verterok: what would be easier, that's what I mean [00:36] s/ro/or [00:36] RenatoSilva: that's the way bzr-java-lib is sending the command :) [00:37] verterok: anyway, would the plugin work if bzr.exe is in another host and you don't have one locally? [00:37] RenatoSilva: but it's wrapped in IBazaarClient interface [00:38] jelmer: we preserve the sha1 cache between runs now [00:39] lifeless: quick call? [00:40] verterok: that's the way bzr-java-lib is sending the command ---> bzr-java-lib could implement the cached process then, using the service plugin, so that users won't need xml-rpc anymore to run bzr faster... [00:41] verterok: users could choose between xml-rpc or direct talk [00:42] poolie: sure [00:43] . [00:46] RenatoSilva: there is no cached process, it's just a python subprocess running a server [00:46] RenatoSilva: the advantage of xmlrpc is that's more standard vs a custom socket service [00:47] verterok: by cached process I just mean the service plugin [00:47] RenatoSilva: in any case, the big plan is to have a more flexible xmlrpc service, using methods instead of running bzr commands [00:48] RenatoSilva: and avoiding to send xml in xml ;) [00:48] ok :) [00:49] RenatoSilva: also I thinked a *lot* about using other protocol insted of xmlrpc, but the main issue are the dependencies [00:50] RenatoSilva: e.g: using google protocol buffers [00:50] RenatoSilva: the other option is to use bzr smart server [00:50] verterok: the protocol could be just command line arguments [00:51] verterok: Bazaar.getServiceinstance().run("--version") [00:51] RenatoSilva: yes, but we need a way to receive the information, and possibly streaming it ;) [00:51] verterok: through a socket [00:52] RenatoSilva: I mean the format of this info [00:52] verterok: you receive the response throught the socket, then you parse the xmloutput output [00:52] verterok: format = keep using xmloutput [00:53] RenatoSilva: with the current xmlrpc service in xmloutput (or with any other transport, e.g: socket) we need to send the whole info back [00:54] RenatoSilva: e.g: bzr log -v of bzr trunk is going to take a while, and it's a lot of info kept in memory (in the case of xmlrpc+xmloutput) [00:54] verterok: List logs = XMLLogParser.parse(Bazaar.getServiceInstance(BZR_PORT).run('xmllog')) [00:54] RenatoSilva: we could stream with a custom socket server [00:54] RenatoSilva: but why reinvent the wheel ;) [00:56] verterok: what are you planning to do to stream it? [00:56] RenatoSilva: also, using commands isn't such a good idea, in the case of bzr-java-lib it works, but it would be a lot better if we have custom service methods to call [00:57] verterok: ok a more computer-friendly protocol, cmd line is mostly user-friendly, I agree [00:58] RenatoSilva: I think that the first step could be to implement the method to replace the xmllog command, and customize the xmlrpc server to stream the response [00:58] verterok: doesn't bzr-service stream the response too? [01:05] verterok: I've heard it's mature...or would that be JRuby? [01:05] RenatoSilva: I'm started a little experiment, write a Java version of the OSLocks, to be used with Jython [01:06] verterok: what's OSLocks? [01:06] lifeless: do you think that ^ might work? :) [01:07] verterok: doesn't Jython gives you full access to python code from java? [01:07] verterok: what is OSLocks and why do you need it [01:07] RenatoSilva: bzr grabs a lock (read and/or write) on some files, it uses fctnl module or pywin32 to do that [01:08] RenatoSilva: as the code is C, Jython don't have access to those modules [01:08] RenatoSilva: Jython only can use pure python modules === tro|| is now known as tro [01:09] verterok: and you have something in python to access the C code, doesn't bzrlib do this for you? [01:09] verterok: shouldn't all bzr C code be wrapped to python for proper third-party access? [01:10] RenatoSilva: you can write python modules in C, but you can use those in Jython [01:10] RenatoSilva: yes, but Jython can't import the required modules [01:10] verterok: I see [01:11] verterok: then Jython is not a full python VM right? [01:11] RenatoSilva: it's [01:11] verterok: environment I mean [01:11] RenatoSilva: but it can't use C extensions [01:12] RenatoSilva: some C extension are already ported to Java [01:12] verterok: as it does not behave as a complete python vm (containing the c access module) [01:13] verterok: if python itself can use C extensions, then Jython should be able too maybe [01:13] RenatoSilva: because the "system" langauages are different, some python C extensions were ported to Java, so Jython can take advantage of it's speed [01:14] RenatoSilva: not really, there is some related work in progress in JRuby using JNA, but not in Jython...yet [01:14] verterok: ok [01:16] RenatoSilva: It's not a matter of having a "c access module". CPython extensions use the CPython C API, which a Java implementation of Python obviously doesn't have. [01:16] * verterok --> dinner, bbl [01:17] jelmer: lots of kudos to you, bzr-git imports working well now thanks to mwhudson's recent work to remember the sha1 cache [01:17] jelmer: can we do the kernel yet? [01:17] jelmer: mwhudson requested vlc which worked taking 13 or so hours, with 34k mainline revisions [01:19] wgrant: I see [01:20] * RenatoSilva --> go home [01:20] thanks everybody [01:20] thumper: Now that you've perfected bzr-git imports, can we get bzr-svn too? [01:21] verterok: sorry was on a call [01:21] verterok: whats the question in short? [01:21] wgrant: we are trying [01:21] wgrant: but it has issues with python 2.4 right now [01:21] wgrant: and our migration to python 2.5 took a hit with the developer working on it being sick for a week [01:21] thumper: Well, I see a solution to that that you need to implement in a few months... [01:21] Oh, so you're actually doing it> [01:21] wgrant: but it is very much in the "near future" [01:22] But why only 2.5? Because Hardy doesn't have 2.6? [01:22] thumper: https://code.edge.launchpad.net/~vcs-imports/wine/git-trunk still not happy :( [01:22] wgrant: because we need to migrate zope through 2.5 first [01:22] wgrant: we aren't jumping two versions at once [01:22] wgrant: because going 2.4 -> 2.5 -> 2.6 is going to be easier than 2.4 -> 2.6 [01:23] I guess so. [01:23] mwhudson: -> #launchpad [01:24] mwhudson: !assertion [01:28] lifeless: it's an opinion but a pretty firmly held one [01:29] mwhudson: yeah, I realise. [01:29] I hold the contrary position. Also fairly firmly ;) [01:31] well it's not either of us doing the work, so i guess it doesn't matter very much [01:43] thumper: were you going to or did you propose an lca miniconf? [01:50] poolie: thanks for reminding me, I'll talk to the people this afternoon [01:50] lifeless: np [01:51] jam, hello, if you're still here [01:52] lifeless: it wasn't a question actually :), I'll try to add an OSLock implementation for Jython, using java FileLock class [01:53] lifeless: is there any specific issues I should take care of? [01:53] lifeless: FileLock docs: http://java.sun.com/javase/6/docs/api/java/nio/channels/FileLock.html [01:54] verterok: not really [01:54] the main thing would be to test interactions with python bzr oslocking of the dirstate file [01:54] make sure they mutex properly [01:55] (read locks sharable, write locks exclude all readers and writers) [01:55] lifeless: ok, thanks for the tips :) [02:20] hi all, i have a repo that i made weird changes to, how can i just make one directory go back to the last comit? [02:20] bzr revert -r -2 directoryname [02:21] thanks lifeless [02:31] I found a class that tests PyQt models. I'm using it now in qbzr. [02:31] http://paste.ubuntu.com/213349/ [02:32] But it use assert(). How important is it to use TestCase.assert*()? [02:33] garyvdm: the main difference is that assert is skipped when run with python -O [02:33] so [02:33] we ban its use in the main codebase [02:33] in tests, it's ok [02:33] but not great [02:34] poolie: So do you think I should customize it? [02:34] hm [02:35] well, what are you going to do with it? [02:35] copy it into qbzr? [02:35] Yes - I have copied it into qbzr because it is not installed with pyqt === mario_ is now known as pygi [02:36] And I call it from qbzr's tests [02:37] i wouldn't stress about it [02:37] Ok [02:38] hope that helps :) [02:39] for inconclusive advice :) [02:39] lol [03:03] Hmm - I can get a new test to run. [03:04] This is qbzr/lib/tests/test_annotate.py : http://paste.ubuntu.com/213366/ [03:05] bzr selftest -s bp.qbzr.lib.tests.test_annotate = Ran 0 tests [03:06] *cant get a new test... [03:07] bzr selftest qbzr.*annotate [03:07] Ran 0 tests in 0.149s [03:07] its not being found then [03:07] hav you added the file to the qbzr test list? [03:08] Ah [03:08] No [03:08] Thank lifeless. [03:16] poolie: ping; a) hunger, b) you'll need time to get home for your call [03:16] lifeless: btw did you file a bug for the unstacking issue with inventories? [03:16] well [03:16] if you didn't it's ok, and i might just comment on 391411 [03:18] I hadn't no [03:18] I was going to, but your adding a comment would be great [03:18] I've isolated my machine turn offs [03:18] spiv, are you here? [03:18] if I hold the left side of my laptop and press firmly (key press firmness) on the top right, it turns off. [03:19] that's handy [03:19] much better than needing all that clicking through menus :) [03:20] ctrl-left click when the right edge is on my knee is enough === MTecknology is now known as \MT === \MT is now known as MT- [03:26] How hard is it to create a bzr repository? Is there any simple guide? [03:29] bzr init-repo DIRECTORY [03:29] but you may not mean what you think you mean [03:29] what do you want to achieve, in nontechnical terms [03:31] Me make pushy wordy stuff to metal box thingy and have wordy stuff in metal box thingy to get later :P [03:31] I had to have fun with that [03:31] :) [03:31] I want to create the repository on the server that was freshly installed with Ubuntu [03:33] poolie: I'm around for a bit now [03:34] jam, was just going to say hi [03:34] so "hi' [03:34] lifeless, leaving now [03:34] hi poolie [03:35] MT-: sure; but I mean - do you want to put a new project under bzr control? or package trees, or ..? [03:36] lifeless: I'll eventually want to put projects in it, I just want the repo available to do things with, fresh installation all the way to having groups of users that can edit specific branches [03:37] MT-: so, bzr doesn't really get organised that way :) [03:37] and the very question led me to suspect you were starting off oddly ;) [03:37] MT-: what VCS are you most familiar with? [03:38] lifeless: bzr, but using with launchpad [03:38] ok [03:38] so on launchpad each branch is in its own repo [03:38] I've setup svn before and had it running over http[ [03:38] oh [03:38] and bzr will create a repo automatically when you push if there isn't one [03:40] poolie: I am, yeah. [03:41] MT-: if you want http/s hosting, you can set that up now [03:41] I'd prefer only ssh [03:41] in which case you're already setup:) [03:42] not yet :P [03:42] I need to find the IP of the server so I can install bzr [03:46] MT-: oh, well thats entirely different :) [03:46] anyhow, use unix permissions to manage access [03:47] try to keep the permissions for an entire repo the same [03:47] (by which I mean *keep them the same*) [03:47] just create as many repos as you need [03:48] and for single-shot things, just use 'bzr push bzr+ssh://host/...foo' to push up a branch [03:48] Can I set where the repos get created? [03:48] when bzr automatically creates one, its inline with the branch [03:49] otherwise they are all manually created [03:54] lifeless: I just meant, can I tell bzr that whenever somebody tried to do bzr push bzr+ssh://server/blah, blah is created within /bzr [03:55] that would push to /blah [03:55] so its very easy, don't give them permission to write to / :) [03:56] what I mean is bzr+ssh://server/ == / on the filesystem (unless you've chrooted your ssh environment) [03:58] Another option is to configure a different bzr serve command line in your ~/.ssh/authorized_keys. [03:58] See e.g. contrib/bzr_ssh_path_limiter \ [03:58] spiv: bzr on the client won't run that though [03:58] spiv: isn't the result a permission error ? [03:59] lifeless: you can create a key that always runs a specified command no matter what the client asks for [03:59] I just found the IP [03:59] fooding [03:59] lifeless: (and that command can then inspect an env variable to get the command client asked for) [04:03] So what I need to do is create a bzr group, add myself to the group, then set the group to own the directory where branches can be created in [04:12] lifeless: does that sound like what you were saying? [04:17] it didn't like that [04:44] spiv, so, i'm ready when you are [05:11] poolie: ok [05:22] When I try to push, I'm getting this error - bzr: ERROR: Unsupported protocol for url "sftp://bzr@208.107.52.90/bzr/tmp": Unable to import paramiko (required for sftp support): No module named paramiko [05:22] lovely that I pasted my ip... [05:26] ok, I figured it out [05:29] YES! I pushed it - although it still needs proper permission setup [05:57] MT-: use bzr+ssh rather than sftp [05:57] MT-: its a lot faster [05:59] lifeless: I was just working off a guide online. I think I'm readdy to tackle loggerhead now but I'm not fully sure how [05:59] looks like it defaults to port 8080 [06:00] ssh seems to work exactly the same except as you said, much faster :) [06:02] I need to wait for the guy to get home and forward the port. We don't have a decent router setup yet. [06:32] ok, back into 391411 [06:47] lifeless: you still around? [06:47] by accounts [06:47] though I'm working on shrinking [06:50] lifeless: When I create a branch by pushing it, the branch is owner/group is me, so other users can't access it. How can I make it so the owner is bzr and the group is the top level directory of the files? If I just push a random branch the group is bzr for all files and if I push to a branch I set the group to devs the files group is devs. [06:50] hopefully that's easy to deal with. I'm probably thinking wrong too [06:51] if I do bzr initi-repo it's root:root, I imagine just changing that to bzr:devs would be easiest [06:52] looks like I should justuse mkdir instead of init-repo instead though, maybe [06:52] lifeless: filed bug 397286 from our call [06:52] Launchpad bug 397286 in bzr "fetch(find_ghosts=True) may not find all CHKMap pages" [High,Confirmed] https://launchpad.net/bugs/397286 [06:52] poolie: thanks [06:53] MT-: set the sticky bit [06:53] ? [06:55] MT-: unix permissions [06:56] I've never heard of that.. [06:56] hi all [06:56] hi [06:56] hey vila [06:56] actually, its not the sticky bit thats needed. [06:56] one sec [06:57] lifeless: +t ? [06:57] or apparently 1000 [07:00] lifeless: whenever you know, I'm all eyes :) [07:00] s/know/remember/ [07:04] set the group on the dir; oh, and the umask needs to be altered to preserve g+w in new subdirs [07:04] that should be all thats needed [07:04] looks like +s on the parent directory [07:05] hrm, that preserved the group [07:06] shouldn't that inheret user too? [07:07] or is that not possible? [07:14] you'd need everyone to be the same user for that to happen :P [07:15] that's ok then :P [07:18] I love how loggerhead is breaking... oh well [07:18] not important atm :P [07:19] lifeless: I'm confused though.. I add a file to the branch, push the branch, and then no files show up in it [07:20] I guess it mush just be a server thing. I pull it and it shows up. Where does bzr keep the files? [07:21] MT-: compressed in a database in .bzr/repository [07:21] oh [07:22] wow... I'm learning a LOT today [07:25] lifeless: thanks for all the help : [07:25] :)* [07:26] anytime [07:27] probably ina few minutes :P [07:34] EOD [07:34] You mean F? [07:34] oh - day [07:36] lifeless: does that mean you're gone or here for a few minutes? [07:36] night [07:57] I'm pretty sure bzr takes care of preserving the g+w if it's set on . I don't believe it tried to explicitly set the group though. [07:57] Depending on FS semantics, it'll inherit automatically all the time (BSD) or when the setgid bit is set on the dir (SysV) [07:59] I think bzr is working perfect now - but loggerhead is irritating me [08:23] vila: bzr-gtk release woo-hoo! [08:23] poolie: yeah ! First release ! :-/ [08:23] mm, why first release? [08:23] your first one? [08:23] Hopefully someone can populate the PPAs from there [08:23] poolie: yes :) [08:26] I'm trying to setup loggerhead. I have no idea what I'm doing... I installed it and I have no idea where to go from here. There doesn't seem to be any apache config created for it [08:26] MT-: Good idea ! You should file a bug asking for one [08:26] vila: idk if I screwed up or not [08:27] vila: Do you know what I'm supposed to do? [08:27] i think there's a readme file describing what to do [08:28] MT-: no, but I know that you shouldn't have such doubts as a newcomer, so filing a bug right now, before getting your hands dirty will help make future newcomers experience better :) [08:28] MT-: try pinging beuno / mwhudson as the usual suspects [08:29] vila: I already screwed with too much stuff. If I file a bug on it, It'll be on a fresh system. I did have it working for a short while before I did something stupid (following a random guide) [08:30] I'll try to flush everything about loggerhead off this sytem [08:31] vila: it's supposed to open port 8080 for it, right? [08:32] MT-: sorry, I never installed loggerhead myself so I can't really help here :-/ [08:32] oh [08:33] I kind of understood at one point that the http server was talking to the loggerhead one but even that bit may be outdated [08:34] poolie: can you save my sanity? [08:34] I fucked this up and idk how to get it back [08:35] ^ I think that sums it up. I broke it [08:35] you're not giving us much to go on there... [08:36] MT-: can you be more specific? [08:38] I tried following this guide - http://wiki.flexion.org/LoggerheadServer.html - Then tried aptitude install loggerhead over top of it. It kinda worked but it crashed some so I tried to finish the process. That wouldn't work so I went back the the loggerhead package. I realized how screwed up I made it so I tried to purge apache2 and loggerhead. I searched the system for files named loggerhead and removed them. I tried to them reinstall apache [08:39] actually, now I'm getting this error - sh: getcwd() failed: No such file or directory [08:42] Getting that error in what context? [08:42] root@carpo:/etc/apache2# /etc/init.d/loggerhead start [08:42] sh: getcwd() failed: No such file or directory [08:43] start and stop are the exact same message [08:43] have a question about API versioning [08:44] developers/api-versioning.html said: In the plugin foo exporting the API (in __init__.py): [08:44] version_info = (0, 0, 1, 'beta', 1) [08:44] api_version = (0, 0, 1) [08:44] what if my plugin don't have qpi_version declaration? [08:45] does bzr will use just version_info? [08:45] spiv: dpkg-reconfigure loggerhead and dpkg-reconfigure bzr-search show those messages too [08:45] even /etc/init.d/apache2 stop is doing it too.... [08:46] I have a knack for destroying a system I guess [08:47] MT-: if /etc/init.d/apache2 is doing that too then it sounds like you've somehow overwritten something pretty crucial in the system. Hm.. [08:47] MT-: or, [08:48] MT-: perhaps the shell you are invoking them from is running from a deleted directory? [08:48] MT-: if you do "cd /" before running those commands, does that help? [08:48] doesn't help [08:49] It's certainly a misfeature of python on debian/ubuntu that "python setup.py install" of random projects tends to clash badly with OS-installed files. [08:49] * MT- sucks at computers [08:49] bialix: I don't know, sorry :( [08:49] glyph: hey [08:49] bialix: no, [08:50] glyph: nice to see you in here :) [08:50] i think it should just assume nothing about what api you need [08:50] ie we don't know if you'll work with the current bzrlib or not [08:50] wait [08:50] it's about compatibility between plugins [08:51] e.g. bzr-explorer requires qbzr >= 0.11 [08:51] spiv can i ask you about test_unstack_fetches in branch_implementations [08:51] oh right, this is your api, not the required api [08:51] but qbzr has no api_version declaration in old versions [08:51] poolie: you can ask ;) [08:51] spiv, i think it ends up with the stacked branch being a local one [08:51] so now I'm checking qbzr version manually [08:51]