[00:06] New bug: #181143 in bzr "inventory sha1 is not checked during routine operation" [Undecided,New] https://launchpad.net/bugs/181143 [00:13] In xml5, those sha1s are not reliable, because there are two xml5 serializations. [00:13] anybody have a problem with a little patch to allow the passing of a File() object in the output param of cmd_send._run() ? [00:15] abentley: we should _always_ have the sha1 in the revision and that of the inventory match within a given repo [00:16] abentley: if they don't, then check and reconcile should allow correcting that / reporting on discrepancies [00:17] abentley: so what I ened is a InterDiffingSerializer that also does Model1To2 stuff [00:20] abentley: is _install_revisions used by bundles? [00:21] Yes. [00:21] so I think we generate check errors from it [00:21] search for 'FIXME: TODO: The following loop *may* be overlapping/duplicate' [00:22] Old bundles only. [00:22] oh, new bundles don't use it? [00:22] They don't use it for file texts. [00:22] Not sure about the rest, but I think not. [00:23] hmm, I think CommitBuilder may fit in there now [00:23] For old bundles? [00:23] I think Jelmer had something that used it. [00:24] heres what I'd like to do. I'd like to allow InterDiffereringSerializer to match (non-rich-root, rich-root|subtree) [00:24] and use CommitBuilder in install_revisions which will take care of the differences automatically [00:25] does that sound ok to you? [00:26] Seems okay. [00:26] cool [00:26] Wait, [00:26] abentley: I don't use install_revisions() anymore [00:26] did you see the iter_inventories patch? [00:26] What about the tree_root knit? [00:26] abentley: commit builder abstracts that out [00:26] Okay. [00:27] you add the root, and a rich root repository will add an entry; non rich ones will assert that the revision is unmodified etc [00:27] What about Inventory.revision ? [00:28] thats the inventories creation id so set by the commit [00:29] IIUC [00:29] When converting from plain to rich-root, we assume the root revision changes every commit. [00:30] right; the old tree has inv.root.revision == inv.revision [00:30] so the commit builder will see those revisions when writing to a rich root repo [00:31] I'm going to start a new isolated branch for this [00:31] I'll make sure there are tests [00:31] Doesn't the commit builder have all sorts of clever logic determine the file's revision? [00:31] yes, thats what we're missing in install_revisions which causes it to generate bad file graphs [00:32] you're wondering if it will overwrite the information during conversion [00:32] If the commit builder commits to a rich-root repo, won't it assume the root revision should not change? [00:35] yeah, I'll need to extend the api to be able to force this [00:35] which is fine, it'll still be cleaner I think [00:35] jelmer: mtaylor@solace:~/src$ bzr svn-import --trees http://code.djangoproject.com/svn django [00:35] bzr: ERROR: Not a branch: "http://code.djangoproject.com/svn/django/branches/queryset-refactor/". [00:46] abentley, hi [00:55] poolie: hi [00:56] abentley, are you working at canonical now? [00:56] No, next week. [01:06] mtaylor: You may want to remove the entry for django in ~/.bazaar/subversion.conf [01:06] jelmer: really? what's that entry for? [01:07] remembers the branching scheme so it doesn't have to be recalculated each time you use bzr-svn [01:07] ah [01:09] how very, very weird [01:09] ? [01:10] jelmer: I tried to do an svn-import of http://reviewboard.googlecode.com/svn/trunk/reviewboard [01:10] and it "worked" [01:10] except it left stuff out [01:10] or [01:10] what sort of stuff did it leave out? [01:10] directories and stuff [01:10] the djblets dir for one [01:10] you would probably want to import the full repository [01:11] not just a part of it [01:11] yeah... sorry. I meant... http://reviewboard.googlecode.com/svn [01:11] (bad cut-n-paste) [01:11] where should djblets be? [01:12] the trunk has one dir, reviewboard [01:12] and djblets should be in that dir [01:12] at least, it is when I do an svn co [01:12] I don [01:12] 't see it in the Svn webview [01:13] are you sure you subversion checkout is up to date? [01:13] yeah - I didn't see it there either :) [01:13] I did this: [01:13] svn co http://reviewboard.googlecode.com/svn/trunk/ reviewboard-read-only [01:14] ls reviewboard-read-only/reviewboard [01:14] accounts contrib __init__.py README settings.py [01:14] AUTHORS COPYING iphone reports templates [01:14] autogen.sh devserver.sh m4 reviews test.py [01:14] ChangeLog diffviewer Makefile.am scmtools urls.py [01:14] conf djblets manage.py server.sh utils [01:14] configure.ac htdocs NEWS settings_local.py.tmpl webapi [01:14] which I find _exceptionally_ odd [01:15] externals perhaps ? [01:16] yup, externals [01:16] mtaylor: bzr-svn doesn't support externals yet [01:16] aha [01:16] stupid svn people and their externals :) [01:53] abentley: still around ? [01:53] I can answer a quick one. [01:53] when we load a non rich roots inventory, is inv.root.revision set correctly? [01:54] that is, is it set == inv.revision [01:54] I remember working on that, so I'd expect so. === bigdog_ is now known as bigdog [03:08] I know we don't yet have line-ending stuff in bzr, but is there any way to get CRLF line endings from cvsps-import, or is that strictly an issue of what cvsps itself puts out? I think it might be in Bazaar territory because --use-cvs doesn't change anything, and without it, [03:08] I think Bazaar uses direct RCS access. [03:08] It pretty much takes what comes out of the RCS files. [03:09] ...and RCS includes in its spec (or at least CVS does) that line endings should be LF only. [03:09] CVS converts per OS on checkout. [03:10] Interestingly, I got expected behavior (OS-proper endings) using Tailor, but cvsps-import gave me LF-only everywhere. [03:10] and Tailor missed all my CVS tags. :P [03:13] One guess is because tailor checks out the files into a tree to work on, while cvsps-import just captures the file versions from stdout. [03:17] Hmm... but I'd think stdout would be OS-translated too... so is this even sensibly addressable in Bzr or cvsps-import, or would it have to be in cvsps? Or somewhere else... maybe a cvsps wrapper that converts line endings--but that sounds dangerous without knowledge of file types etc... [03:17] cvsps just synthesizes the changesets. cvs or co is used to get the file versions. [03:19] My cvs checks out with CRLF though; this is why I'm surprised --use-cvs didn't help. [03:20] Yeah, but co'ing a dir tree doesn't necessarily give you the same thing as 'co -p'ing a file version to stdout. [03:20] I'm trying to capture bzr output test for use within a bash script. I've tried several things and googling proved fruitless. Anyone know what I'm missing? [03:20] text* [03:21] I can test that anyway; it's a good start. Thanks much. :) [03:22] i.e., echo " $( bzr update ) " > myfile [03:22] (I have some 40 projects I want to talk my company into moving from CVS to Bazaar, and they're all in CVS; so the plan is to come up with a nice clean conversion plan, then pull it off, and then start demonstrating pros :) [03:23] winzo_: just "bzr update > myfile", surely? [03:23] You'd think so. But it didn't work. [03:23] A newline char gets written, though. :? [03:24] What command are you using for your test? (bzr command I mean) [03:24] Possibly "bzr update > myfile 2>&1" then. [03:24] (to capture stderr as well as stdout) [03:24] that's where I was heading... [03:25] That was it! Looks like I was re-directing wrong. :D [03:25] Thanks guys! [03:28] how to install webserve [03:31] fullermd: cvs update -p produces CRLF here. Does --use-cvs literally make cvsps-import run the cvs command on my path? [03:34] AFAIK. [03:37] Filing a bug for lack of a better idea... [03:39] could some one please help in installing webserve [03:46] indu: If I can make a suggestion. Ask specific questions. [03:46] indu: e.g. 'When I XXX YYY happens' [03:56] New bug: #181161 in bzr-cvsps-import "Line endings go from CRLF to LF under Cygwin, but not with Tailor" [Undecided,New] https://launchpad.net/bugs/181161 [03:59] dlee: have you tried without cygwin? cyginw does some file mode seting magic [04:00] I don't happen to have a non-Cygwin cvsps, though I do have a non-cygwin cvs I think. [04:02] lifeless, i am not knowing the procedure of installing webserve [04:03] lifeless, i have seent he docs folder of the webserve source, where there is a README.txt file, which is giving information about how to configure webserve [04:03] which is to be done after the installation, of webserve, and no info about the installation procedure [04:04] its in python coding [04:05] indu: the README.txt has a section titled "Install" [04:06] spiv, yeah, its in the last line of the README, I am sorry, i missed seeing it. thanks [04:06] spiv, but without knowing this, i have done the copying of the webserve into the plugins fodler [04:07] and then I have done the proper configurations also in the apache.conf file [04:07] then I started webserve, using #bzr webserve --port=9099 "Official repository" http://localhost:9099/bazaar/ [04:08] now, when I am trying to access the http://localhost:9099/bazaar, its displaying nothing on the browser [04:08] Unbale to connect [04:09] Running "bzr webserve ..." has no relation to Apache AFAICT. [04:10] i.e. you run that command to run it as a standalone server. [04:29] spiv, now i have dont the all the explained configurations, and when i am accessing, through http://localhost/bazaar, my apache error.log is showing the mesage as [04:29] http://pastebin.com/m6e80c8b0 [04:36] indu: it sounds like you don't have bzrlib properly installed [04:37] woot, it pulls [04:37] indu: oh actually, that looks like a bug in the webserve plugin [04:38] indu: put "import bzrlib.commands" in the webserve plugin's __init__.py [04:39] indu: or just change the line that says "import bzrlib" to say "import bzrlib.commands" [04:42] any idea about this error message [04:47] If I do bzr send ... > outfile, then email outfile to a committer, (I) is this the best way to submit code changes where I can't commit/push, and (II) can the committer select among my (possibly multiple) commits in outfile easily and apply at will, in whole or in part? [04:48] dlee: (why not just run "bzr send", and let it start the email client for you?) [04:48] I'm afraid so far my emailings of this sort have just been typo fixes in comments, but I assume those are still of use. [04:49] dlee: yes, that's a good way to give your changes to someone else [04:49] Hehehe, good idea but this box isn't set up very well for that. This is Windows+Cygwin without Windows mail or well-configured Cygwin SMTP... mea culpa :) [04:49] dlee: they get all the revisions that you added, so they can do with them everything you can locally [04:49] dlee: including examine individual revisions, etc. [04:49] dlee: ah, I see :) [04:50] spiv, yeah, now i can see something in my browser, but an error message, saying section not found [04:50] Cool...because one thing I just saw involves a misspelled class name, and the author may wish not to mess with that [04:50] where as I have the repositories in the approprate lcoation [04:51] indu: I'd guess there's a problem with the configuration file for webserve, but I have no experience with that myself. [04:51] indu: so I can't offer much help on that [04:52] Of course, if they cherrypick revs, you lose the identity. But if upstream is in CVS anyway, it hardly matters... [04:52] There must be a way to grab --mail-to from bzr log or somesuch though... [04:53] fullermd: yeah, but "bzr send" doesn't make that situation any better or worse. [04:53] Oh this is for Bazaar plugins actually. I don't know Python (yet), so the best I can do so far is to send typo fixes while I read code and docs... [04:53] Yeah. `send` is uninvolved in that part of the process. [04:54] anyone else there, who can help me in this please [04:54] I read with a talking computer, and it seems to catch types of spellos that more easily escape the eye... if you type "Don't forget you hat," I'll catch it right quick; but if you type "Seems theirs a bug here," I won't... [04:57] Whereas that latter instantly grabs my eye and makes me want to beat the perpetrator with a 2x4 :p [04:57] hehehe [05:03] hmm, there seem to be a mistake in the my webserve conf file, could some one please correct it, http://pastebin.ca/846269 [05:05] New bug: #181165 in bzr-webserve ""import bzrlib" should be "import bzrlib.commands"" [Undecided,New] https://launchpad.net/bugs/181165 [05:05] indu: it would help to also tell people what the error you are getting is. [05:07] spiv, my error is on the browser only, and it has lot of details, i think i can take a screenshot and send, any site where I can upload the screenshots images [05:08] in simple, my Error messages on my browser says that ,, section not found: None [05:08] Section Not found [05:13] no one here, who can provide me help in webserve ? [05:13] Spiv: fwiw, bzr send direct to email now works here--nicer indeed. I may be the only Windows user using Mutt though.... [05:13] dlee: :) [05:21] Well, surely there's at least one other sensible Windows users in the wor... no, you're probably right. [05:22] fullermd :) [05:24] fullermd: btw, I tried checking through cvs-import code, and now I wonder if subprocess.popen() is capable of being told something that might help with the line ending conversion issue--but that's a taller stack of stuff to paw through than I want to start tonight... [05:26] Well, that's waay over my head :) [05:29] Does the cvs-import guy come in here? [05:34] Oh, yes, he's in here most days. But he's in my TZ; it's almost midnight now. [05:35] He's usually here during the day (say, 1500-2300Z). [05:35] Nick? [05:36] jam or jam-laptop or some such. [05:37] The list would probably also be a good place to check. [05:50] Since I submitted a bug on the cvsps-import line ending issue, if I do manage a code fix that works in my strange environment, would I be better to send it to Jam or somehow paste or attach it into the bug report? [05:52] dlee: you can attach patches to bugs [05:52] dlee: or even upload a branch to launchpad and link it to the bug. [05:53] Ah, the former doesn't surprise me but I didn't spot how to do it; the latter is beyond me to date... [05:53] but sounds cool [05:54] dlee: you can add attachments in followups to a bug report [05:54] we don't currently have a way to create an attachment when the bug is filed. [05:58] jamesh: gotcha; explains how I missed it. [06:10] jelmer: you around? is there any way to do an svn import do that svn tags show up as bzr tags and not bzr branches? [06:12] svn has tags? ;) [06:14] fullermd: as much as they have branches ... [06:39] no help in webserve i can get here? [06:40] indu: you've been getting help; its a shame it hasn't been enough to sort it all out yet. [06:40] indu: I suggest mailing the list; goffredo reads the list and he knows all about webserve [06:42] lifeless: Oh, hey, it just mentally penetrated that you're fiddling with inventories... [06:43] lifeless: How amenable is our current setup and/or your work to referencing unchanged files? [06:46] fullermd: EPARSE [06:46] is svn2bzr still in work? or has bzr-svn pretty much supplanted it? [06:46] lifeless: `bzr ci --unchanged -m'foo' asdf.c ; bzr log --limit=1 asdf.c (check for 'foo')` [06:47] fullermd: it won't record that [06:47] fullermd: a new version of asdf.c isn't recorded, nor does the metadata about it change. [06:48] Does it require major surgery to do so? [06:48] yes [06:48] whats the use case [06:48] Conversions, for one thing. Forced commits are very common in the CVS world for making notes on files (updating incorrect commit messages, etc) [06:49] Losing that information would make 'log $FILE' misleading at best. [06:49] And with --unchanged existing in bzr, one can imagine doing similar things; if we need --unchanged, we can just as well need to reference files in it, and having 'log' not show those revs would be Wrong. [06:50] so two things [06:52] conversions [06:52] hmmm [06:52] Of course, the very thought of "high fidelity CVS conversion" is giggle-inducing. [06:53] But still. [06:54] I think this is worth raising; the conversion argument is not strong - because the upshot of that argument is being a usperset of everything, good bad and ugly [06:54] Just a thought; ISTM that it's something we otter be able to represent, so as long as we're rototilling inventory handling anyway it's a capability to bear in mind. [06:54] the argument about 'what is --unchanged for' is however really good [06:54] if --unchanged is to allow logging extra information, there's no good reason you can't tickle some files in passing [06:55] we can represent a new change in the per file graph in the current disk format [06:55] the surgey is in the commit code is all [06:55] and in UI, to diferentiate between 'selecting' and 'annotating' [07:04] 'k. Should I file a bug, or open up a list discussion? [07:04] discussion i think [07:04] * fullermd nods. [07:05] Good 'nuff. Thanks. === asac_ is now known as asac === bitmonk_ is now known as bitmonk-sharp [11:59] mtaylor: No, bzr-svn doesn't support creating tags yet, that's an open bug. === mrevell is now known as mrevell-lunch === mrevell-lunch is now known as mrevell === cprov is now known as cprov-lunch === radix__ is now known as _radix === cprov-lunch is now known as cprov [17:45] hi [17:46] how do i release a lock on a launchpad repository ? [17:46] i have lost the connection during the initial stage of a push [17:46] ignas: tried bzr break-lock [17:47] thanks [18:12] hi all [18:13] quick question [18:14] I was trying to install bzr v1.0 onto a ubuntu feisty install and I am getting the following error "Failed to fetch http://bazaar-vcs.org/releases/debs/feisty/./Packages.gz MD5Sum mismatch" [18:14] does anyone have an idea if it is possible for me to fix this on my end or must I wait for changes at the bazaar site? [18:15] there is the obvious alternative that I can install from source, but I liked being able to apt-get the newest release... it will significantly beat version .15 that I have now [18:17] samiam: What did you add to your sources.list? [18:17] I don't see the actual deb files for feisty there [18:17] Odd_Bloke: deb http://bazaar-vcs.org/releases/debs/feisty ./ [18:17] but you can use http://launchpad.net/~bzr/+archive which seems to have them [18:18] I followed the instructions provided on http://bazaar-vcs.org/DistroDownloads [18:21] thanks luks [18:21] looks like getting it from launchpad works properly [18:23] BAH! The link from bazaar-vcs.org should be changed to the PPA [18:25] mtaylor: Well, PPA doesn't build stuff for feisty and dapper yet afaik... [18:25] jelmer: sure it does? [18:25] oh, rather, PPA itself can [18:25] the packaging would probably need altered to be able to build on all those [18:25] jelmer: I see bzr packages for dapper-gutsy [18:26] elmo: oh, ok. when was that added? [18:27] elmo: last time we discussed this (a bit before the holidays iirc) PPA supported only gutsy and hardy [18:27] jelmer: don't know actually, someone only pointed it out to me in the last week or so === mw is now known as mw|food [19:18] is there any way I can ask a branch what locks it has? === mw|food is now known as mw [19:26] thumper: I think 'bzr status -v' gives some information. [19:27] Peng: ta [19:28] Peng: it seems that status requires a working tree [19:28] Peng: which I don't have [19:28] hi! [19:28] if I want to convert a subversion repository to bazaar [19:28] thumper: Sorry, I meant info, not status. [19:29] and I have a lot of tags in my subversion repo (from every nightly build), what is the best way to get that tag info into bazaar without importing every single tag branch? [19:29] Peng: oh, ok, it isn't showing anything about locks [19:29] thumper: Even with -v? [19:29] yeah [19:30] morning statik [19:30] hey there thumper === cfbolz_ is now known as cfbolz === mwh___ is now known as mwh [20:13] Ok I'm *sure* I'm missing the obvious this time, but... if I have 320 revs in my local branch but I need to undo just revision 313, how do I do it? I'm used to applying what amounts to a reverse patch, then committing rev 321... but I've never done it in Bazaar. [20:16] You can merge it in reverse [20:16] bzr merge -r 313..312 [20:17] Might need to use: bzr merge -r 313..312 . [20:17] (with the . at the end) [20:21] Ah... I hadn't put the . for merge branch, so I was getting an error. Why does "bzr merge -c313 ." say "all changes applied successfully" but appear to do nothing at all? (Granted that was an experiment...) [20:23] dlee: Check 'bzr status'. [20:25] Odd_Bloke: I must be fine; "bzr status --versioned" is empty. [20:26] ...and "bzr diff" is also empty. [20:26] Peng: Worked, thanks. [21:17] dlee: '-c313' means '-r312..313', not '-r313..312'. And it's expected to do nothing, since it's an attempt to merge something you already have. [21:19] fullermd: That was my guess, but I experimented to see if it would figure out (from having nothing to do otherwise) that I wanted to go the other way. :) I was just surprised it said nothing about it, except to indicate success at doing nothing. [21:20] Is there a shortcut for reversing just one revision? I think subversion would do -C313. That was confusing to me when I saw it, but I'm slipping Bazaar methods into a project hand-over report and making them look as easy as pie so people might start liking Bazaar as much as I do. :) [21:20] -r313..312 is good enough otherwise. [21:25] hg has a backout command. === bitmonk_ is now known as bitmonk-sharp [21:39] peng: Could be done pretty quick with a plugin I bet. [21:45] Probably be done with an alias. [21:46] What the heck? [21:46] .bzr/repository.backup? [21:48] hmm, side effect of a failed upgrade most likely [21:48] I didn't know anything other than '.bzr.backup' was used. [21:48] I ran a 'find' and upgrade to pack-0.92 to upgrade branches a couple days ago. [21:48] Repos were already updated. [21:48] There was only a 'repository.backup' directory, no 'repository' directory. [21:49] hey guys - given a MergeDirective - what's the best way to get a list of revisions contained within it [21:49] And a .bzr.backup was created. [21:49] I'm guessing I want to get deserialize the MergeDirective.bundle, right? [21:49] Peng: strange; care to file a bug with any info you can think of that might be relevant? [21:50] Ok. [21:50] Very little info though. [21:51] ah... found it [21:52] Bug 145812, perhaps? [21:52] Launchpad bug 145812 in bzr "Upgrade can leave a broken repository (with backup)" [Low,Triaged] https://launchpad.net/bugs/145812 [21:53] Peng: so the repository is not working at all ? [21:53] I renamed it back to 'repository' and it's fine. [21:54] I have been using bzr under linux for a while now, I am now trying to get a National Instruments (NI) LabView (LV) project under revision control with bzr (1.0.0); my project directory has sub directories with whitespace (proj/a sub proj). This seems to cause problems when I try to branch. [21:55] I get the error "bzr: ERROR: Not a branch: "C:/Documents and .../Desktop/proj/a sub proj". what am I doing wrong? Does bzr on Microsoft Windows not handle whitespace as well? [21:56] it should [21:58] is "proj/a sub proj" the dir you are branching from, or to? [21:58] Just for a point of refrence, 'bzr branch "proj" "proj.a"' works fine ... [21:58] important ... note the ' " ' [21:59] GaryvdM: branching from [22:00] And if you go into that dir, is there a hidden directory called .bzr? [22:00] GaryvdM: So ... on a linux machine what I think I would type is "bzr branch proj/a\ sub\ proj proj/a\ sub\ proj.b" [22:00] Ah ah. [22:01] I just checked .bzr.log. [22:01] It is the bug I mentioned earlier. [22:01] GaryvdM: Under 'proj' there is a folder .bzr; under the folder 'proj' "bzr status" returns nothing. ie. every file in there (including 'a sub proj') are under bzr control. [22:02] on windows what you would type is: bzr branch "proj\a sub proj" "proj\ a sub proj.b" [22:03] lifeless: Comment? [22:03] I think your branch is "proj" - not "proj\a sub proj" [22:04] Peng: what bug earlier ? [22:04] Bug 145812. [22:04] Launchpad bug 145812 in bzr "Upgrade can leave a broken repository (with backup)" [Low,Triaged] https://launchpad.net/bugs/145812 [22:04] You triaged it a few months ago. :) [22:04] GaryvdM: Yep ... I was trying to branch an individual sub directory in the project ... not the whole project. My Bad. [22:04] Trivial to reproduce. [22:05] bzr init --pack-0.92-subtree; bzr upgrade [22:05] ok [22:05] I have no browser right now, will check it later [22:05] You don't have a browser open? [22:06] nope [22:06] performance testing [22:06] need the memory for cache [22:07] Ah. [22:08] With that 20-second commit I do with hg, when Firefox is really chewing RAM, it can take over a minute. :) [22:08] yah [22:26] New bug: #181367 in bzr "bzr update should work in a treeless bound branch" [Undecided,New] https://launchpad.net/bugs/181367 [22:47] good morning [23:05] hello! when I try to run bzr qcommit under gnome (ubuntu 7.10 gutsy), I get a window with no window decoration. When I run qtconfig, I get normal looking window decoration. Do I need to do anything special to make qbzr get window decoration? [23:16] statik: luks is the person you would want to talk to, when he's around [23:16] jelmer: ok, thanks [23:16] I did try just a very simple pyqt application, and I get window decoration [23:16] and I don't see any errors in ~/.bzr.log [23:23] lifeless: Now that PPA supports Dapper, Edgy and Feisty, do you have any plans to move the bazaar-vcs.org debian repo to PPA or would you rather keep it separate because of the signing? [23:26] statik: the best way to fix that is to use pygtk [23:26] mtaylor: did you know qbzr already has the commit message save and restore in it? [23:26] really? [23:26] yep [23:26] SWEET [23:27] you'll need to modify it to handle per-file commit messages, but that should be not so hard