[06:21] spiv: hey, you might like to know your quiescing code for bzr+ssh bazaar.launchpad.net is now live in HA mode. [06:23] just in time :) [07:12] anyone up for a review? https://code.launchpad.net/~mbp/bzr/test-errors/+merge/64485 [07:55] poolie: sure, I'll have a look [07:56] ah hi jelmer [07:56] hi there [07:56] seems weekend is not a godo time to catch you ;-) [07:57] hi Noldorin [07:57] hi poolie [07:57] Noldorin: Usually it is, but I spent 27 hours of this one on trains [07:58] ah i see [07:58] soudns fun. :-P [07:59] jelmer, well if you got my email, i replied to your messages...had actually already submitted the bug report on friday when speaking to you :-) [08:37] jelmer, you gone again?? [08:42] Noldorin: no, sorry, but going through mail backlog at the moment [08:42] Noldorin: I'll follow up to the bug [08:44] jelmer, ok, please :-) [08:44] jelmer, this is very imporant to me, so i will be happy to help you with this. have some free time either now or tomorrow night (for you) [08:48] Noldorin: unfortunately it just requires a lot of time debugging [08:48] Noldorin: I can provide you with some hints if you would like to have a look at it yourself [08:49] jelmer, certainly... [08:49] jelmer, obviously leave it on your own task list, but it doesn't hurt for me to have a go as well. :-) [08:49] as i understand, the issue is a missing item in the Git database...don't know anything more [08:51] Noldorin: one of the bzr-git routines doesn't properly generate an entry for a particular src directory in the _revision_to_objects method [08:51] it seems to happen specifically with r47 of ircdotnet, but I'm not sure why (which change in that revision is problematic) [08:52] a good first step might be to try to push r46 into a git repository, and confirm that that works [08:52] interesting [08:52] and then see if you can reproduce the issue with a subset of the changes in r47 [08:52] jelmer, what do you mean "particular src directory"? [08:53] Noldorin: see the comments on the bug report [08:54] jelmer, link please? for some reaons i am not getting notifications :-( [08:57] Noldorin: Sorry, I don't have it handy here. You should have the bug# though, as you filed the bug. [08:58] hah i wouldn't assume [08:58] i will have to search for it... [08:58] i rely on LP to update me. oh well [09:00] got it [09:08] jelmer, what does Triaged mean? [09:11] it means accepted by a developer [09:11] ok heh [09:12] such an obscure term in normal English ;-) [09:12] heh, http://markmail.org/message/jt2psgctve3riact#query:+page:1+mid:jt2psgctve3riact+state:results [09:13] poolie: great to hear about HA for codehosting, looking forward to seeing jam & spiv's forking server code land! [09:13] isn't it great [09:17] lifeless: ha, thanks :) [09:26] jelmer, yes pushing as r46 works... will investigate the issue tomorrow; thanks! [09:27] good night for now [12:16] loom is so close to being great [12:52] anyone seen mgz around recently? [12:57] jml: I think I talked to him last week [13:30] hi everyone. i'd like to try some bzr hacking. if anyone has any suggestions on a good starter bug for exploring the code base, let me know. otherwise i'll pick the first "easy" bug that makes sense to me. [13:37] GRiD: I think the bugs tagged "easy" are indeed the best place to start [13:46] jelmer, ok thanks [13:59] Is there anyone here to answer my bzr-svn problems? [14:05] Silence.... [14:06] dreamkxd: can you be more specific? [14:07] dreamkxd: what is your question? [14:14] Well, I want to modify bzr-svn to handle multiple branch in an single path, where to start? [14:15] Or in other words, I want to ignore some content in svn branch. how to do that. [14:20] dreamkxd: those two things seem like different things [14:20] dreamkxd: what do you mean with multiple branches in a single path? [14:20] I think they can be handle in the same way. [14:20] ah, you're luoyonggan [14:20] Yes. [14:20] For example there is an path /branches/branchA/BranchB [14:21] ?? [14:22] right [14:22] at the moment you would have a different layout for interpreting that layout differently [14:22] so, the files under BranchB should be fully ignore by branchA [14:22] yes. that's right, that's why such an condition appeared. [14:23] dreamkxd: that's not really possible without changing the mapping format, and breaking all existing revision ids [14:23] So, that's an impossible task? [14:23] almost? [14:24] dreamkxd: yes, and I don't really see the value of branches that exclude certain data. I don't think it's a natural way to work, because "svn co" works with full subtrees. [14:24] That's the problem, svn co works with full subtrees. [14:25] but bzr is dvcs. So it's didn't accept redundant data in the repo, or the repo will be very large. [14:25] that's the value of bzr-svn. and bzr-svn should be able handle such an condition. [14:26] dreamkxd: The problem with excluding certain files is that it creates different history [14:26] ergo, different revision ids that are incompatible with checkouts that do have all the files [14:27] Excluding things like big files seems like a better idea for just the working tree (which bzr already supports) or by using ghosts. [14:27] Sorry, I meant to say: [14:28] It seems like a better idea to just limit what files are created in the working tree (which bzr already supports) or by using ghosts. [14:28] dreamkxd: I don't have any objections to support for considering multiple things in a path branches [14:28] to specify the file name? [14:29] dreamkxd: but I don't think considering /branches/branchA/branchB a branch means that /branches/branchA would no longer include the contents of BranchB [14:29] I am just considerating the fact condition. [14:30] dreamkxd: what do you mean? [14:30] because at some time, and because the dirty part of svn, some one like to copy the full project B as an subdirectory of project A. [14:31] so when convert, we should exclude project B when converting project A. [14:31] dreamkxd: that makes it impossible to create sensible revision ids [14:32] That means /ProjetA/branches/ProjectA/ProjectBForEasyBuild copy from /ProjectB [14:32] no, I need only one thing, that's ignore all files under directory /ProjetA/branches/ProjectA/ProjectB [14:33] we won't create rev id for projectB, because it's virtual, won't convert when we converting ProjectA [14:35] dreamkxd: sorry, but that means we change the contents of an existing revision (because we delete data from there) [14:35] dreamkxd: and that requires a format bump [14:36] dreamkxd: this would also be an issue for svn users of ProjectA [14:36] so is that really all that common? [14:36] common? [14:37] do you means that the data from bzr on an path is different from svn's path? [14:39] dreamkxd: that too [14:39] that's means the bzr's branch /ProjetA/branches/ProjectA/ 's content is different to svn's /ProjetA/branches/ProjectA/ path? because ProjectB is removed from bzr? but bzr still have? [14:39] So what's they other problems? That will affecting bzr-svn's codebase a lot? [14:40] dreamkxd: I think this should be entirely different of your autolayout work, it's a lot larger project (and I don't think it's a good idea) [14:41] Er... [14:42] dreamkxd: you can't simply change what data bzr-svn imports for a particular revision-id; excluding a directory that older versions of bzr-svn do import will break. [14:42] dreamkxd: this is wholly different from layouts [14:42] dreamkxd: unless I'm misunderstanding what you're proposing [14:42] well, may patch won't affecting other layout's work. [14:43] I means, there current layout works as-is. [14:43] and only when using layout auto, we handle such a conflict. [14:44] dreamkxd: the layouts don't have any influence on the revision-id; the revision-id depends just on the repository UUID, revnum and path [14:44] dreamkxd: your layout can't adjust history, it can only help in what primary paths get imported as a branch [14:45] adjust history? what's that means? [14:46] dreamkxd: if you'd like to support multiple branches for a single path (e.g. interpreting both /branches/branchA and /branches/branchA/branchB as a branch) that should mostly be a matter of patching RepositoryLayout.parse() to return multiple results rather than just one [14:46] dreamkxd: however, even if you do that and somebody does "bzr branch /branches/branchA" they will get a branchA bzr branch that has a "branchB" directory in it. [14:47] dreamkxd: does that make sense? [14:48] Er.. That's doesn't make sense. [14:48] except the orginal branchB and /branches/branchA is the same. [14:49] But that's almost impossible. [14:50] dreamkxd: Sorry, I'm having a really hard time understanding what you're saying. [14:50] orginal of( branchB and /branches/branchA ) [14:50] i means /prefix/branches/branchA copied form /prefix/trunk, and /prefix/branches/branchA/branchB both copied from /prefix/trunk. [14:51] in that case, "branchB" would still be a directory if you check out "/prefix/branches/branchA" [14:52] so bzr branch /branches/branchA checkout withou direcotry "branchB" at most case. [14:52] yes, that's right. [14:52] dreamkxd: it will checkout *with* branchB [14:53] it possible checkout *with* branchB [14:53] and the possiblity of possible is very low. in most case, it's not. [14:53] no, it will always checkout with branchB. If you want to make the presence of branchB optional, you should work on per-file ghost support in bzr core first. [14:54] do you means geting directory branchB as an ghost, then the problem will be resolved? [14:55] dreamkxd: a ghost is basically something that bzr knows about but doesn't existing the repository [14:55] dreamkxd: a ghost is basically something that bzr knows about but which it does not actually have [14:55] indeed, i have no knowledge about ghost at all. [14:55] dreamkxd: it currently supports ghost revisions (e.g. holes in the revision graph) and in principle there are ghost files as well, though they aren't really supported at the moment. [14:57] So where to start? I means in the current codebase? [14:57] dreamxkd: in bzr-core. It would probably be very hard to implement them, but it's the right place to do it. [14:58] dreamkxd: It also will be hard to implement what you're proposing in bzr-svn, especially considering performance. [14:58] not in bzr-svn? [14:59] dreamkxd: well, it would require support from bzr-svn too, but that's a fraction of the work required in bzr. [14:59] I have to say, it's makes me afraid. [15:00] well, i have a simple hack to avoid this problem, just directly ignore branchA, getting it to be death, even lost some data, but at least it works. [15:02] btw, didn't bzr ghost works like svn-externals? [15:05] dreamkxd: getting to directly ignore branch*B* you mean? [15:06] dreamkxd: ghosts are different from externals; svn-externals are a lot like bzr nested trees. [15:06] although I guess in some regard ghosts can be seen as absent by-value nested trees [15:06] svn-externals are like by-reference nested trees [15:07] no, I means ignore branchA. [15:09] only perserve the last(tailing) one, so that's didn't need to ignore it's subdirectories. [15:09] for example /branches/branchA/branchB. [15:09] we won't create /branches/branchA, but branchB. [15:10] ah, ok. fair enough [15:15] I meet antoher problem under bzr-svn, when an branch X exist before, but then it's deleted, it won't be converted by bzr-svn, am I right? [15:17] dreamkxd: if you're referring to "bzr svn-import", it will only be converted if you specify --all [15:18] Ok. I see it. === jelmer_ is now known as jelmer__ === beuno is now known as beuno-lunch === beuno-lunch is now known as beuno [18:00] hi jelmer [18:01] hi Noldorin_ [18:01] jelmer, so i'm going to check into the issue now a bit :-) [18:01] jelmer, pushin r46 indeed works... btw how can i dpush to another remote branch? [18:07] Noldorin_: how do you mean? [18:07] Noldorin_: fwiw, you can also push to a local git repo [18:07] that might make testing easier [18:07] jelmer, exactly that...what's unclear? [18:07] ok sure [18:13] Noldorin: if you want to push to a different branch, just specify a different URL [18:13] jelmer. how exactly does the URL change? [18:14] Noldorin_: if you want to push to /tmp/some/repo, use "bzr dpush /tmp/some/repo" [18:14] wel... [18:14] git+ssh://git@github.com/alexreg/ircdotnet.git [18:15] jelmer, that pushes to master branch...how do i change that url for other branches? [18:16] Noldorin_: ah [18:16] Noldorin_: it's not possible to specify colocated branches to push to yet, that's one of the things I'm working on [18:18] jelmer, ah i see... Git/Hg have collocated branches but not Bzr right? [18:18] jelmer, will this feature be ready any time soon? :-) [18:19] Noldorin_: There is a plugin that provides it for regular bzr branches (lp:bzr-colo) [18:19] The work I'm doing is about adding it to the core, so it can also be used for other things like the bzr-hg and bzr-git plugins [18:19] 20:19 Noldorin_: There is a plugin that provides it for regular bzr branches (lp:bzr-colo) [18:21] jelmer, ah ok. do let me know when it's implemented please! [18:22] Noldorin_: Subscribe to bug #380871 [18:22] Launchpad bug 380871 in Bazaar "support for colocated branches" [High,In progress] https://launchpad.net/bugs/380871 [18:23] jelmer, thanks [18:25] jelmer, it says fix released for bzr-git? [18:28] hmm oh well [18:32] Noldorin__: the required fixes for bzr-git have happened, but that doesn't really matter until the bzr part is there [18:33] jelmer, got it. [18:33] cheers [18:33] brb [19:00] what bug tracking software integrates the best with bzr? :) [19:15] Kraln: Probably Launchpad [19:17] probably; what about self-hosted stuff [19:22] Kraln: I'm not sure; I know there is e.g. sloecode, I don't know if that does bug tracking though [19:22] Kraln: things like trac also have plugins that add bzr support [19:26] * jelmer__ realizes he doesn't actually have a lot of useful info [19:26] sorry === yofel_ is now known as yofel [21:18] What does 'lost connection during test' mean? [21:18] https://launchpadlibrarian.net/76377611/buildlog_ubuntu-natty-i386.bzr-svn_1.1.0~bzr3767-1~bazaar1~natty1_FAILEDTOBUILD.txt.gz [21:28] maxb: it's a subunit error [21:29] basically, it can mean various different things - in this case I suspect it means that the test runner crashed [21:29] worth retrying the build to see if it goes away? [21:29] Is this using the latest subvertpy, 0.8.0 ? [21:30] 0.8.3 [21:30] There were some ref counting issues in subvertpy fixed a while ago that were causing segfaults during the bzr-svn testsuite [21:31] in that case, it's probably a new one :( [22:17] hi all [22:18] g'morning poolie [22:18] how are you today? [22:18] hi, welcome back [22:18] good thanks [22:18] going to do some reviews and piloting today [22:18] my own queue is almost empty which is good [22:18] thanks, it's good to be back.. though I think I got quite a bit of stuff done last week too :) [22:19] poolie: thanks, I'll make sure to have a look at your MPs [22:20] thanks, welcome back [23:51] how can I split my ./debian folder into a separate branch and move it out of the repository into a completely differnet folder? [23:57] systemclient: bzr split in your debian directory will create a separate branch [23:58] systemclient: after you commit that, you can safely "bzr branch /path/to/thefolder/debian /tmp/newlocation"