[00:00] lifeless: CAn you BAN announcer? Pretty please? [00:04] Rev 3273: (jam) Fix moving directories to root nodes. john@arbash [00:05] Rev 3274: (Neil Martinsen-Burrell) Explain version-info --custom in the User ian.clatworthy@canonical.com [00:05] its saying channel not synced [00:05] when I try to ban [00:05] * awilkins is not familiar enough with IRC to know what this means [00:05] neither am I [00:05] I am hoping someone else where will be ;) [00:06] lifeless: op me [00:06] It's because there's a netsplit going on, I think [00:06] * awilkins has a cunning plan.... /ignore announcer [00:08] elmo: what was the issue ? [00:08] lifeless: no idea; it was a WAG that it'd work for me [00:08] I'm spethial ;-) [00:08] heh. thanks [00:09] Rev 3273: (jam) Fix moving directories to root nodes. john@arbash [00:10] Rev 3274: (Neil Martinsen-Burrell) Explain version-info --custom in the User ian.clatworthy@canonical.com [00:10] elmo: it would be nice if it had worked for you :) [00:12] i knew the ban syntax was funky, but that's really strange [00:13] awilkins: I don't see anything there about allowing c variables as arguments to extension type constructors [00:13] jelmer: The bit that caught my eye was "So you can use extension types to wrap arbitrary C data structures and provide a Python-like interface to them." [00:14] I was thinking that meant you could pass a c-struct to a parameter that demanded an extension type based on that c-struct ? [00:15] awilkins: probably, but I'm trying to work out how :-) [00:15] What are the types? [00:15] not really relevant, but svn_ra_reporter2_t, apr_pool_t, void * [00:16] void *? One of the dreaded "batons"? [00:16] yup [00:16] that's not really relevant here though [00:16] I just need to pass pointers, no need for interpretation [00:16] by pyrex [00:17] anyone know who is running announcer? [00:17] Well, a pointer is a platform-sized integer [00:18] jelmer: you can't allow C ariables directly because extension types are constructable by python [00:18] jelmer: the constructor has to follow python ABI, which means getting python objects and unpacking appropriately [00:19] lifeless: statik is running announcer apparently [00:19] jelmer: Is this relevant? http://www.obviousobscurity.org/?page_id=13 [00:19] lifeless: Ah, crap [00:19] statik: ping [00:19] Just under "Example 1" "wherever you can pass a void * you can pass a python object" [00:19] jelmer: care to give me a quick snapshot on your problem [00:19] lifeless: It's possible in SWIG, it just makes the class impossible to construct from python itself [00:20] jelmer: that means it has no constructor :) [00:20] it has a constructor, but no python users can ever construct the arguments to call that constructor properly :-) [00:20] jelmer: one way you could meaningfully do this is to have a no-args constructor, then a C initialisation interface [00:21] jelmer: swig has issues though ;) [00:21] lifeless: Basically, I have a pyrex function which calls out to a C function that returns 2 pointers [00:21] I would like to return a Python object from my pyrex function that uses those two pointers [00:22] lol searching for 'pyrex wrap' on google is not sfw [00:23] * awilkins isn't at work *click* [00:23] :-) [00:24] lifeless: yeah, pyrex owns swig big time if you only need Python bindings [00:24] The next release of bzr-svn will likely no longer use python-subversion but its own pyrex code [00:25] jelmer: you can define __new__ with C args I think [00:26] http://ldots.org/pyrex-guide/5-python-wrapper.html#class [00:26] e.g. def __new__(self, reporter, pool) [00:27] oh hmm [00:27] no [00:27] jelmer: the easiest and cleanest way [00:27] it'll complain that it's not possible to use cdef for special methods [00:27] would be a trivial type wrap (e.g. an incomplete type definition) [00:27] and without cdef it's not possible to have c type arguments [00:27] what are the types of the two pointers [00:28] one is a typedeffed struct and the other is a void pointer [00:28] so [00:28] ctypedef typedeffedstruct [00:29] ctypedef ctypedef baton void * [00:29] cdef class foo: [00:30] def __init__(self, struct, baton): [00:30] self.struct = struct [00:30] self.baton = baton [00:30] cdef typedeffedstruct struct [00:31] cdef baton baton [00:31] obviously reorder but roughly that should work I think [00:31] that breaks [00:32] pastebin it? [00:32] because it complains in the __init__ that struct is being converted from a Python object to a ctypedeffedstruct [00:32] thunk it [00:32] def __init__.. [00:33] thunk it? [00:33] cdef typedeffedstruct c_struct [00:33] self.struct = c_struct [00:33] emr [00:33] c_struct = passed_in_struct [00:34] totally unclear I know. 3 lines; a typed variable, assignment to that [00:34] then local assignment [00:34] if that doesn't work, give me a second [00:35] * awilkins thinks 00:21 < jelmer> lifeless: Basically, I have a pyrex function which calls out to a C function that returns 2 pointers [00:35] 00:21 < jelmer> I would like to return a Python object from my pyrex function that uses those two pointers [00:35] oops [00:35] sorry [00:38] lifeless: nope, I don't seem to be able to fool it that way either :-( [00:38] jelmer: hi [00:38] jml: Hey === jamesh_ is now known as jamesh [00:39] Did you look at http://www.obviousobscurity.org/?page_id=13 , because I thought that looked promising [00:39] jelmer: I was looking at your subunit patches: https://bugs.edge.launchpad.net/subunit/+bug/178362 and https://bugs.edge.launchpad.net/subunit/+bug/178361 [00:39] Launchpad bug 178361 in subunit "[MERGE] Split by newline rather than by any whitespace character." [Low,Fix committed] [00:39] jelmer: but I couldn't find the actual patches. [00:41] I think I originally sent them to robert and later forwarded them to launchpad before it had support for attachments [00:41] I'll see if I still have them somewhere [00:43] jelmer: can you pastebin your experiment ? [00:44] Or just post a branch of it :-) [00:44] well [00:44] right now, fastest to pastebin ;) [00:45] I presume it's svn_ra_do_diff ? [00:46] jelmer: anyway, if you post them on LP, I'll try to review them [00:47] Poor jelmer, being pulled three ways [00:48] lifeless: The branch is at http://people.samba.org/bzr/jelmer/bzr-svn/pyrex [00:48] meh [00:48] ok [00:48] I'll see if I can trim it down to something that's easier to test and play with [00:49] jml: sure, I'll have a look [00:50] jml: Hmm, it's not possible yet to create a branch in launchpad yet using "bzr push lp:..." ? [00:51] ah, only if I specify a homedir explicitly [00:52] Yup, no squiggless [00:52] jelmer: huh what? [00:53] "bzr push lp:subunit/cparent" complains about "invalid series" [00:53] but "bzr push lp:~jelmer/subunit/cparent" works ok [00:53] oh [00:53] what would you rather it do? [00:54] Tell me new branches can only be created if I specify an explicit owner, perhaps [00:54] it's a minor thing though [00:54] I tend to only use launchpad for mirrorring my branches rather than hosting them [00:55] hey is it somehow possible to ignore a file which is under version control? let me explain why I need this: [00:55] jelmer: error messages are pretty important [00:55] whe have a website in bzr, a forum - its in its untouched status so everybody can install it [00:56] but the changes that occur by installing it dont neccessarily need to be submitted [00:56] thus I'd like to ignore some of those files - is this possible? [00:56] jelmer: please file a bug [00:57] jelmer: also, bzr register-branch --product subunit, now uses the public location so just-works [00:57] lifeless: yes, but this is really a minor thing compared to "bzr register-branch" which gives a traceback for pretty much everything [00:58] (branch already registered, invalid product, authentication, ..) [00:58] milian: bzr help ignore [00:58] jelmer: file bugs on all of those :) [00:58] jelmer: feedback -> improvements [00:58] awilkins: I don't htink that will do what milian wants [00:58] milian: you basically want a file which noone can edit [00:59] lifeless: I'm not sure what milian wants [00:59] milian: I don't see how you could edit it yourself subsequent to doing this [00:59] milian: what I suggest you do, is have the file called something like 'filename.example' [00:59] lifeless: I know :-) I've already filed bugs on some of them earlier [00:59] milian: then folk can copy it to filename, to get an unversioned local copy [00:59] milian: How about template files which you copy to "install" the thing? [01:00] milian: You could even write a small script that does the copying [01:00] Darn, lifeless got there first [01:00] hmmm complicated [01:00] but well, lets see what I can come up with [01:00] thank you guys [01:02] milian: or you could just 'bzr revert filename' after merging from someone to ignore any changes they made [01:03] The template idea coupled with an ignore list for the real config files has the advantage of not requiring constant vigilance [01:06] ah blast it - now I've accidently commited some files I dont want - how do I revert that commit without losing my changes? backing up my files and reverting? [01:06] I haven't pushed yet - so the commit should be local [01:07] or is bzr revert --forget-merges what I want? [01:07] bzr uncommit [01:07] bzr uncommit [01:07] oh darn - so simple :D great [01:07] thank you again [01:08] hello again [01:09] poolie: hi [01:09] poolie: did you see leslies 'register now urgent urgent' mail ? [01:10] just checking my gmail now so just saw it [01:13] statik: ping, announcer banned, it wouldn't shutup [01:17] lifeless: not yet, will look [01:17] lifeless: i just saw that hardy went into beta freeze [01:17] i would really like to get the baz rename done before it's too late [01:17] was it beuno looking after it? [01:18] yes, was getting the patch up [01:18] some serious headache though [01:18] baz fails to build [01:18] api changes in libneon [01:18] poolie: oh, so unless you applied for bzr to be an organisation, we've missed soc. [01:18] I didn't put us in. [01:19] i have not [01:19] right, that makes deciding on soc projects easy [01:19] right [01:19] tbh i don't think it gives a good return on the amount of time it requires, compared to helping people in the project directly [01:20] we can sponsor people directly if they'd like to do similar things [01:20] I don't htink there is much soc specific overhead myself [01:20] poolie, baz doesn't build in either lenny or feisty > [01:20] the questionaires etc are pretty quick to do [01:20] so the work is done for the transition [01:20] but since we can't get it to build, we can't upload [01:22] beuno: so the old binary can stay in the archive? how is that allowed? [01:23] lifeless: there's time in applying, dealing with student applications, reporting to them [01:23] it's non-trivial [01:23] poolie, it won't in Debian, and, in Ubuntu, it's been passed on in binary. I have _no_ idea how/why [01:23] jml: Still there? [01:24] jelmer: always [01:24] it will get removed from Debian if it's not fixed, as it's a release critical bug [01:24] jml: The changes for those bugs are in the branches I just registed on launchpad [01:24] jelmer: ok. [01:24] jml: My trunk branch contains the SubunitTestRunner change [01:24] and the "jelmer" branch contains a bunch of other changes [01:24] poolie: there is, I'm comparing that time to the time to get contributions directly; we still have to decide whats good, work with folk contributing code, assess how they are going if we are doing a bounty [01:24] poolie: I don't think the /incremental/ overhead is significant [01:25] poolie: anyhow, its all moot [01:25] it is moot [01:25] approximately zero of the students last year committed long-term improvements [01:26] that doesn't mean there's no prospect of doing better [01:27] the main advantage i think is if it brings people in who would not previously have thought of working on it, and who do actually get right into it [01:27] the potential is there... [01:28] phanatic got into bazaar through the SoC [01:28] Samba got one committer because of the SoC [01:29] I know wine has similar results [01:29] all 1 in 6 students actually staying around [01:29] poolie: I think we should do bounties for the folk that were interested from the sprint [01:29] poolie: if we can get funding for that [01:30] that's true [01:30] we can [01:30] poolie, and you already have the topics to choose from organized in: http://bazaar-vcs.org/SprintLondonMarch08/Brainstorms :) [01:30] so we can fund jelmer, phanatik, guillermo, [01:33] cool [01:35] so, poolie, lifeless, what do you want to do about baz? [01:36] excuse my ignorance, but if it ftbfs how can it go into a new distribution release? [01:36] well, I don't know how Ubuntu is doing it. They seem to just pass on the binary [01:36] beuno: you're still in london?+ [01:37] Debian will remove it before relase [01:37] poolie, no, Madrid. My flight was cheaper if I returned the 20th, so I'm waiting here until then [01:37] beuno: i want to not have hardy release with 'install bazaar' not giving you baz [01:37] i mean, bzr [01:38] poolie, even at the expense of removing baz? [01:38] neither me or james_w could get it to build in debian or ubuntu, so we can't upload the changed binary name for transitioning [01:41] is there any reasoning as to what options are retrieved using their own function on BranchConfig or using get_user_option() ? [01:41] http://weblog.latte.ca/blake/tech/bzr/Looms - nice :) [01:42] jelmer: get_user_option has a basic lookup [01:42] jelmer: custom methods are used for things that don't fit into that [01:42] lifeless: nice. [01:43] lifeless: also http://thread.gmane.org/gmane.comp.package-management.vcs-pkg/11 [01:44] lifeless: thanks [01:44] james_w: great post' [01:44] james_w: now how about a patch to make merge actually merge the looms :) [01:45] lifeless: erm, yeah... [01:45] I might do some looms tomorrow. [01:46] I've got my own little bzr sprint going on. [01:46] Going to work is for people who never want to get anything done :) [01:49] james_w: :> [01:50] * beuno points to http://blog.orebokech.com/2008/03/emacs-in-bzr-initial-impressions.html [01:58] hello yacc [02:00] poolie: hi. [02:10] beuno: so it really looks like log in the main problem atm; mwh has a patch to speed it up for very long history [02:10] (when he's not busy killing x) [02:11] poolie, ah, cool. We should really make an effort to make sure "bzr is painfully slow" doesn't get stuck in people's heads [02:15] indeed [02:17] jml: could you followup about that, if you haven't done so already? [02:17] * beuno is off to bed [02:21] anyone know how to stop vim from forcing you to the first column when you type # to start a comment? [02:22] james_w: hm, yes, somehow [02:22] poolie/lifeless: Any chance one of you could set "child_submit_to = bazaar@lists.canonical.com" in bzr.dev's .bzr/branch/branch.conf ? [02:22] * poolie looks [02:22] jelmer: i think only lifeless can write there [02:24] james_w: i remember hitting that myself [02:24] maybe 'filetype indent on' ? [02:25] I screwed up a merge. commit from one machine is the mainline is at rev 40. changes from second are recorded as rev 38.1.1. how do I merge rev 40 and 38.1.1 into a new revsion? [02:25] or set nosmartindent nocindent, probably in an autocmd for python files [02:26] piedoggie: are those numbers from the same branch? [02:26] y [02:26] hm [02:26] actually, can you explain more? it sounds like you've already merged them [02:26] you don't like the way you merged it? [02:27] all the changes (38.1.1) have not made it to head. [02:27] because they conflicted and you rejected them? [02:27] it is like I created a branch [02:28] the changes should have been in conflict but I never got the merge message [02:28] poolie: it seems like :set cindent might be it. [02:28] rather nocin [02:28] er.. conflict message [02:29] piedoggie: [02:29] poolie: hmm, permissions [02:29] possibly easiest is to do 'merge -c 38.1.1', which will redo all those changes, then you can pick what you want [02:29] lifeless: ? [02:29] on that branch [02:29] k let me try that [02:30] -c not -r? [02:30] yes [02:30] poolie: rsync/bzr/bzr.dev$ chmod g+w .bzr/branch/* [02:30] poolie: you should be able to edit [02:30] poolie: sorry about wrong permissions [02:30] if the changes you want are not in that particular revision, but rather something it merged, then you might need to pick out the particular one and use -c on that [02:30] poolie: thanks for the pointers. [02:31] you're welcome [02:31] *yawn* [02:31] 5am is the wrong time to wake [02:31] how can I show all the diffs? diff -c 38.1.1?? [02:32] yes [02:33] bbiab [02:33] this may not be a good thing :-) but bzr works in a way that feels natural to my mind [02:33] :) [02:44] * yacc wonders if bzr is not a better client for svn repos than svn even now, ... [02:45] * RAOF has found it to be so. [02:46] But that's because I don't use svn properly (ie: I try to do branch-merge, which is almost exactly half supported by svn :P) [03:54] i have a question: can i use bzr for multiple projects which are in multiple directories (not all under one obvious parent directory)? [03:55] Potato-[CE]: Sure, you could but you have to examine whether there's advantage in keeping each project under the same repository. [03:56] As long as you house all projects within some central directory, though, then yes, you can keep all of them in a single repository. [03:57] well, i think i need to start over with my directory layout. I have one project just started, i don't mind erasing the data...what's the best way to change the repository directory? [03:58] just delete .bzr? [04:00] Potato-[CE]: That would do it. [04:02] okay, i see that here too "If you accidently put the wrong tree under version control, simply delete the .bzr directory." [04:02] thanks for the information, i appreciate it. :-) [04:11] So I'm trying to check out a svn repo with bzr-svn, and I can't tell if it's in a loop or not. [04:11] I've got the message: [04:11] using experimental bzr-svn mappings; output may change between revisions [04:12] So, I think I've mentioned this before, but did you know that the ssh process started by Bazaar hangs around doing things if the Bazaar process is SIGINTed? [04:12] and a progress bar keeps filling up. [04:13] oops, nevermind, there it goes. [04:13] I guess it just had tons of branches to check out. [04:13] *heart* bzr-svn. [05:03] jam; ping [05:30] New bug: #202063 in bzr-gtk "gannotate display of uncommitted lines is unobvious" [Undecided,New] https://launchpad.net/bugs/202063 [05:49] does Bazaar use Twisted Conch as its ssh client in Windows? [06:41] jml: not as far as I Know [06:43] lifeless: it's just that there's some things that say "Twisted" on the Bazaar / Launchpad showmedo video [06:50] interesting [07:06] lifeless: hi, want a brief call? [08:07] ubotu: ping? [08:07] Sorry, I don't know anything about ping? - try searching on http://ubotu.ubuntu-nl.org/factoids.cgi [08:07] Well, that's a good enough response. [08:20] New bug: #202079 in trac-bzr ""Invalid url supplied to transport" in sub directories" [Undecided,New] https://launchpad.net/bugs/202079 [08:31] I've got some strange behaviour with Hardy's bzr: I'm branching Do (with lp:do, so bzr+ssh), and when I do this into a directory which *isn't* a shared repository, "bzr info" in the branch directory gives Standalone tree (format: dirstate). [08:32] When I branch it into a directory under a default-format (pack-0.92) repository, and then run bzr info in the branch directory I get: Repository tree (format: unnamed) [08:33] Yeah. [08:33] The name like "pack-0.92" refers to the repository type, branch type, and working tree type. [08:33] When I branch it into a repository with rich-root format, the branch errors out with: http://www.paste2.org/p/15882 === tro|| is now known as tro [08:34] When you branch it into the repo, it's using pack-0.92's repo type but dirstate's branch type, so bzr doesn't know what to call it. [08:34] Right. So that should probably be handled by printing both repo & branch type. [08:34] Indeed. [08:35] info -v will show all of them. [08:35] Ah, right. [08:36] Is there already a bug for this? [08:37] Shrug. [08:37] Wait, a bug about what? [08:38] That information or the error you pasted? [08:38] That information. [08:38] Oh. [08:38] Shrug. [08:38] Heh. It's a small problem, but it'd be nice if it displayed both :) [08:38] Any ideas about the error?+ [08:39] Eh [08:39] I dunno. [08:39] Heh [08:39] But, rich-roots and non-rich-roots don't really go together. [08:39] Huh. [08:39] I'm surprised you didn't get a compatibility error. Maybe bzr's smarter than that. [08:47] jml: plink appears to be the best choice for bzr+ssh:// on windows [08:47] What's wrong with paramiko? [08:47] Had a chap in here a couple of days ago that couldn't get it to work, but plink worked fine for him [08:48] I suppose I could upload a key to LP and try both [08:48] What problem was he having? [08:48] Password prompt? [08:48] I thin it was dropping the stream in the middle of establishment [08:49] He was using an agent [08:51] * awilkins uploads a key [08:51] Oh, huh. [08:54] DId you buy your network card from the same guy that sold jelmer his :-) [08:55] Hahaha. [08:55] Actually, the Ethernet port on my 2-year-old motherboard failed, so I'm using my 5-year-old NIC... [08:55] But I'm tinkering with a new IRC client. [08:56] I'm quite happy with irssi running on my router atm. [08:56] Goodbye Konversation. [08:56] I'm due for another bout of screwing with irssi to try and resolve annoyances... === frsk_ is now known as frsk [08:57] Heh. [08:57] I've used irssi a little, but I'll stick with my GUI clients.. [08:57] The part where it (sometimes) eats part of wrapped lines is especially bothersome. [08:58] Running irssi on the router has the advantage that my office IT services can't block it. [08:58] Peng_: weechat [08:58] The big problem with GUI clients is that they don't work well inside screen :p [08:58] asabil: ? [08:58] weechat is a quite nice client [08:58] Ah. [08:59] Oh, non-GUI. [08:59] still very usable [08:59] Connection reset by peer? What the hell? I hit the quit button! === Peng_ is now known as Peng [08:59] Well that's a pile of crap then [09:01] New bug: #202083 in bzr "bzr info confusing when branch format does not match repository format" [Undecided,New] https://launchpad.net/bugs/202083 [09:43] So once you upload an SSH key to launchpad, is there a way to test it's working without creating branches? [09:44] If you have to create one, there's staging and junk. [09:44] Meh? [09:45] Branches pushed to lp://staging/ or something are deleted daily. You can also register random branches to the +junk project. [09:45] Aha [09:45] Wait, if you're using bzr+ssh, any operation would require a password or SSH key.. You could just "bzr info" something. [09:46] Mmkay [09:46] So, just replace the http with bzr+ssh ? [09:46] Yeah. [09:46] And add your username [09:47] After running bzr launchpad-login, lp: URLs will automatically use bzr+ssh too. [09:48] so my username would be adrian-wilkins (that's what it says in my key edit window after the "~") [09:52] Ooh, doesn't work well from powershell [09:53] Ok, does the same thing from cmd.exe tooo [09:53] THe SSH stuff works with plink though.. now for paramiko [09:55] bzr: ERROR: Unrecognised value for BZR_SSH environment variable: paramiko [10:06] Dumb-ass question : do you need to install pycrypto and paramiko for python 2.5 on windows? [10:06] * awilkins is finding it hard to locate a pycrypto binary for py2.5 on win32 [10:06] They're both still necessary with 2.5, but they're likely provided with the bzr installer. [10:07] THe standalone one or the python one? [10:07] * awilkins is using the python one [10:07] Aha [10:08] * awilkins intalls the damn dependencies [10:08] http://bazaar-vcs.org/BzrWin32Installer#bzr-dependencies ? [10:08] Yeees [10:08] That might not be the Python-based installer. [10:08] I have no idea how it all works. [10:08] I've always used putty/plink [10:09] THe Python flavoured one has none of the extra deps in it [10:09] Ah. [10:12] Paramiko seems to work better than plink [10:13] Plink gives you this "unable to read from standard input" error after it finishes [10:14] It's happy enough using Pageant for keys though (puttys ssh-agent) [10:14] Yay [10:42] Man, progress bars have gotten unprogressive with packs :| === cpro1 is now known as cprov [10:53] I always just _love_ seeing progress bars that clearly have no idea of how long things take :-| [10:53] Even better are the ones that go backwards [10:53] I'd much rather see a spinner, or an "estimate bar" [10:53] Maybe make the progress bar out of "?" instead of "=" [10:53] Or "~" [10:53] "Approximate progress bar" [10:57]