[00:00] I don't think a good two-way CVS plugin is easily doable [00:00] * LarstiQ is inclined to agree with jelmer [00:00] the easiest solution is probably to just build on something like cvs2svn, but you would have to make sure that you don't change the output or make the file ids / revision ids/ depend on the changes in the content [00:00] jelmer: dead code, i've removed it locally [00:01] Jc2k, ah, ok [00:03] garyvdm: woohoo to the qbzr threadless-throbber branch landing :) [00:03] jelmer: can you have a look at r126 - change to find_missing_objects so it uses get_refs instead of heads. thats enough to fix git-serve for git clone [00:03] (just pushed again) [00:03] markh: I'm happy to. Just worried about regressions. [00:04] jelmer: (something doesnt feel right about it) [00:05] I'll create new binaries locally so I test it with my "real" bzr [00:05] I must find that resources problem first... [00:12] Jc2k, looks alright [00:12] merged [00:13] wicked [00:55] jelmer: any thoughts on dealing with the incoming push? i was thinking about having a FakeGitRepo class that operated on in memory heads and a single pack, and then dealing with it similar to the stuff in svn-import [00:58] jelmer: i guess the only difference between FakeGitRepo and LocalGitRepository is at the dulwich level anyway... [00:59] (its a .git less repo with one pack and the heads are just a dict..) [01:01] jelmer: so i could add a Repo class in dulwich that just took a pack and some heads and otherwise didnt touch the disk, then tweak LocalGitRepository in bzr-git so it can just be pasted a dulwich repo object directly, then in the server code create a LocalGitRepository with an in memory fake virtual rpeo think and copy revisions with InterGitRepository [01:02] and i think use pull to update teh tips? [01:02] pasted/passed [01:03] * Jc2k heads to bed before my spelling accuracy falls any lower [01:03] Jc2k, still there? [01:03] jelmer: just about [01:04] Jc2k, hmm [01:04] Jc2k, did you see Repository.add_pack() [01:05] jelmer: are you suggesting i actually create a valid git repo for each incoming push? [01:06] Jc2k, not sure I follow [01:06] whjat would you need the fake repository for? [01:06] how does add_pack help with handling an incoming push to bazaar? [01:07] ah, you were talking about an incoming push to bazaar [01:07] that makes more sense [01:10] * Jc2k might just create a git repo for each push to get something working, and then refactor [01:12] anyone know the syntax to pass a username through bzr-svn [01:12] dort, in the url [01:12] when checking out from an http:// repo (svn) [01:13] http://user@ [01:13] gives me a segfault after prompting for the password twice [01:13] dort, what version? [01:13] i just installed in my debian sid (didn't check yet) [01:14] Bazaar (bzr) 1.5 [01:14] in that case you would want the version from experimental [01:14] what bzr-svn ? [01:14] ii bzr-svn 0.4.10-2 [01:15] is there a better choice in experimental? [01:15] dort, yeah, you want the version from experimental [01:15] ok, thx much [01:16] if that still segfaults, please file a bug [01:19] hmm [01:19] O watta goo Siam [01:19] as i install the new packages (or try to) I realize i've filled up my hard disk [01:19] * dort sulks away to fix this [01:19] sorry, maybe that's what caused the segfault in the first place :-[ [01:21] sleepytime [01:22] working now, thx much jelmer [01:22] sweet dreams === timchen1` is now known as nasloc__ === Spaz is now known as BrickOfShit === BrickOfShit is now known as Kittens [07:02] hi all [10:38] hey Vila [10:38] jelmer: hi ! === Mario__ is now known as pygi === asac_ is now known as asac [12:14] jelmer: ping [12:14] Jc2k, pong [12:16] morning :) [12:16] i simplified apply_pack a bit (to use object_store.add_pack) [12:16] alas its now all kinds of slow [12:17] (well, it might even be stuck in a looop...) [12:25] Jc2k, hi :-) [12:25] Jc2k, ah, cool [13:12] Jc2k, back [13:16] * Jc2k waves [13:39] jelmer: iterentries seems to be going on forever :\ [14:15] Jc2k, yeah, the generating of the index is slow as hell at the moment [14:15] Jc2k, I think it's because of the repeated mmaps [14:15] jelmer: this is a definite loop [14:16] jelmer: traced it back to read_zlib [14:17] jelmer: for some reason add = data[base:base+1024] isn't fetching any data, so fed doesnt increase, so it tries the same chunk again, which is still "", and unused_data is still "".. so it goes on forever [14:18] ahh [14:18] jelmer: its read 2639 of an object of size 4117, any its about 20 or so objects away from the end of the pack [14:18] so im not sure whats going on [14:24] interesting [14:24] what happens if you manually try to feed that data string? [14:29] *tests* [14:32] jelmer: if i feed the pack manuall to PackData.iterobjects, it seems to be fine [14:33] (and the same pack was working before i moved to add_pack && comit.. [14:33] ) [14:35] flushing thing. [14:36] jelmer: gah! for some reason i thought commit closed the fd. [14:37] pushed. [15:53] checking [15:56] Jc2k, still there? [15:57] jelmer: i wasnt, but i just got back :] [15:58] jelmer: whats up? [15:59] Jc2k, just merged + pushed your branch [15:59] fetches from local git repositories and from remote git repositories work now [16:00] sweet [19:38] jelmer: ping [20:02] Jc2k, pong [20:04] jelmer: hey. it seems like im able to push revisions :] [20:05] Jc2k, w00t! [20:05] though its an evil hack [20:05] I'm trying to clone samba using bzr at the moment :-) [20:05] and its made me think using InterGitRepository is needless [20:05] i really should just use import_git_objects directly [20:06] which im going to try, but first i need to spawn branches [20:06] trying to rob some bzr-svn for this.. [20:07] Jc2k, yeah, I don't think InterGitRepository is necessary either for the server side of things [20:07] it just duplicates the work-out-what-to-copy step and then calls import_git_objects [20:08] geh [20:08] jelmer: is there a super easy way to just spawn a branch and set it to a given revision id? [20:09] Jc2k, a bzr branch you mean? [20:09] yeah [20:09] Jc2k, Yeah, once you have a BzrDir, just call "bzrdir.create_branch()" [20:09] and then branch.generate_revision_history(revid) [20:10] does bzrdir.create_branch need the directory to be present? [20:10] yeah [20:10] it also needs the .bzr directory to be present [20:11] theres a large chunk of bzr-svn code for this (get_dir, transport_makedirs, and some other utils) [20:11] Though you ca nprobably use BzrDir.create() or something to create the .bzr directory [20:11] ah, ok [20:11] Jc2k: It may be nice to move those utility functions into bzrlib [20:11] just so we don't need any copies [20:11] thats what i was thinking, but maybe BzrDir.create is better? [20:12] Jc2k, What I usually do is first move it to bzr-foreign, and then later submit it for inclusion in bzr.dev [21:02] jelmer: so import_git_objects. atm i guess the order of the object yielding function doesnt matter. will it> [21:09] Jc2k, no, it doesn't matter [21:10] It won't matter in the future either I thijnk [21:10] (with apologies for my bad spelling tonight, blame the beer) [21:10] Jc2k: The order in which remote hosts order it would be related to the delta order [21:15] jelmer: brilliant. in that case im going to see if i can tweak dulwich to support iterobjects directly off the wire [21:15] jelmer: at the very least, i dont need the index [21:16] If I have a branch that I want to basically remove all the contents out of it and replace it with new contents.. how would I do that? [21:16] smoothice, bzr pull --overwrite ? [21:16] Jc2k, yeah, you don't need the index indeed [21:17] jelmer: ok thanks === fawek_ is now known as fawek [21:57] * jelmer is having a hard time trying to map a git index file to a bzr working tree [21:57] it seems the two concepts don't map very well [21:57] perhaps support for (git-like) indexes has to be added to bazaar first [21:58] jelmer: is a primary aim for bzr-git to let people use bzr on a local git clone? [21:59] Jc2k, depends on what you mean by primary aim :-) but if at all possible, it would be nice to support it [21:59] :-) [21:59] at least personally, I would very much like to be able to use bzr on native local git repositories [22:01] jelmer: so i find myself needing a PackData.iterobjects that returns me ShaFiles [22:01] Jc2k, Use Pack.iterobjects() [22:01] Jc2k, you may want to generate an index first [22:01] jelmer: but if i tweak iterentries i dont need the index files... [22:02] Jc2k, you do, because you need to resolve deltas [22:02] oh, crap. [22:02] maybe you're just lucky at the moment that you don't hit any ref delta's, but once you do, you really want that index [22:02] perhaps we can allow the index to be in-memory at some point [22:03] but I don't think that will help significantly in terms of performance [22:11] ok, pushed what i have so far. now using import_git_objects directly, and accessing pack directly (with index) instead of going via Repo. [22:11] ah, cool [22:12] * jelmer is looking at implementing the index [22:12] its a bit messy, le sigh. === spm_ is now known as spm [22:42] ah well === davidstrauss_ is now known as davidstrauss [23:43] can someone with pycurl installed try "bzr info https://code.launchpad.net/bzr" please? [23:52] james_w: works for me [23:52] markh: thanks [23:52] markh: I assume you are not on Ubuntu :-) [23:53] :)