[00:01] <maxb> .....that's a bit special.... "bzr multi-pull" is pulling the same branch over and over again!
[00:01] <maxb> Too many levels of symbolic links: u'/home/maxb/wc/bzr/bzr-svn/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/1.0/.bzr/branch-format'
[00:01] <maxb> heh
[00:02] <maxb> which would be fine, except there's not a symlink in sight
[00:03] <maxb> oh, yes there is
[01:27] <beuno> vila, ping
[01:27] <beuno> vila, what am I missing besides medusa to run all the tests?
[01:27] <beuno> Missing feature 'FTPServer' skipped 42 tests.
[01:28] <beuno> vila, http://paste.ubuntu.com/329884/
[01:31] <lifeless> beuno: python-medusa
[01:31] <lifeless> not medusa
[01:32] <beuno> hi lifeless
[01:32] <beuno> lemme try that..
[01:33] <beuno> (would make a truck-load of sense)
[01:33] <lifeless> beuno: (note that you can't run all the tests, ever).
[01:34] <lifeless> beuno: because windows isn't linux and linux isn't windows.
[01:34] <beuno> lifeless, I'm going to let other worry about windows for now
[01:34] <beuno> I get the same thing, sftp tests skipped
[01:35] <lifeless> beuno: ftp or sftp ?
[01:35] <beuno> lifeless, sftp
[01:35] <beuno> actually
[01:35] <beuno> FTPServer
[01:35] <beuno> Missing feature 'FTPServer' skipped 42 tests.
[01:35] <lifeless> python-medusa should have been it
[01:36] <beuno> it didn't  :/
[01:37] <beuno> I guess I can TDD anyway, and figure out the FTP part tomorrow
[01:37] <lifeless> beuno: what are you working on ?
[01:37] <beuno> lifeless, bzr-upload, going to implement a few new features
[01:38] <beuno> starting off with a .bzr-upload-ignores
[01:38] <lifeless> you can check the feature availability test
[01:39] <beuno> tell me more
[01:41] <lifeless> grep for FTPServerFeature, I think it is
[01:41] <lifeless> in bzrlib.transports.ftp
[01:42] <beuno> ftp/_gssapi.py:    if ftp_server.FTPServerFeature.available():
[01:43] <beuno> from bzrlib.tests import ftp_server
[01:44] <beuno> there's a comment: # medusa is bogus under python2.6
[01:44] <beuno> pyftpdlib...
[01:45] <beuno> can't find anythin in apt with it though
[01:48] <lifeless> keep pulling the string
[01:52] <beuno> hm
[01:55] <beuno> maybe pyftpd
[01:55] <beuno> nope
[03:21] <beuno> vila, https://code.edge.launchpad.net/~beuno/bzr-upload/ignores/+merge/15346
[03:21] <beuno> I managed to get some decent coding after *months*
[03:21] <beuno> TDD and everything
[03:23] <beuno> vila, I'm dissapointed you're not awake at 4:30am on a Saturday to review my branch
[03:24] <beuno> and I am also dissapointed in how long it takes Launchpad to generate a diff with no feedback at all
[10:40] <Glenjamin> hi guys, i'm trying to convince the company I work at to choose bzr over git for future projects, and i'm wondering what my options are for a central repository with secure access control
[10:41] <Glenjamin> is there some way to proxy the dedicated mode of bzr serve through apache or similar?
[10:44] <jszakmeister> You might want to check out: http://doc.bazaar-vcs.org/bzr.2.0/en/user-guide/http_smart_server.html
[10:45] <jszakmeister> I've not tried it yet (just don't have enough 'roundtuits)
[13:10] <thrope> hi - is it possible to pull/move a file from an unrelated repo preserving history?
[13:11] <thrope> sorry pull or copy I mean
[13:12] <thrope> I would like to have be able to import a single file with history from another repo
[13:13] <thrope> I'm guessing not but wondered if there was any trickery with rebase or something that could do it
[13:20] <jelmer> thrope: You can use merge
[13:20] <jelmer> thrope, but that'll pull of the contents of the other repo
[13:21] <thrope> I guess I could branch the source repo, delete everything but the file I want, then merge? (acutally I didn't think you could merge unrelated branches)
[13:23] <thrope> I get bzr: ERROR: Branches have no common ancestor, and no merge base revision was specified.  when I test merging
[13:24] <jelmer> you can merge unrelated branches by specifying -r0..-1
[13:24] <thrope> same with bzr merge -r 1 ../a
[13:24] <thrope> ah
[13:24] <thrope> great thanks
[13:29] <thrope> only thing is it pulls the whole history of other branch (i guess obvious) but I only really want to changes of that file
[13:49] <bing> hey folks.  i'm looking for a volunteer to take a cvs tree, and a svn tree and merge them into a single bzr tree.  i know it can't be much fun, but i figure *somebody* has to like doing this stuff.
[14:52] <thrope> mwhudson: another minor loggerhead thing... running another public instance at a nonroot url (but still https, apache proxypass)... if I go to https://www.host.net/bzr/ then everything works perfect but if I got to https://www.host.net/bzr (no trailing slash) then the repo links on the browsing page are broken (go to host.net/repo instead of host.net/bzr/repo)...
[14:52] <thrope> I tried playing with apache settings a bit, and have --prefix set in loggerheadd script; running the relative links patches from the other day
[15:26] <Glenjamin> we're looking at possibly setting up bazaar for our office, and i'm looking into how to get an access controlled central repository with access only over an encrypted transport - currently i think my options are bzr+https with auth via apache, and bzr+ssh with a wrapper like http://bazaar.launchpad.net/~bzr-pqm/bzr/bzr.dev/annotate/head%3A/contrib/bzr_ssh_path_limiter - can anyone offer some...
[15:26] <Glenjamin> ...advice?
[15:32] <vila> Glenjamin: Looks like you already found the right bits, do you have more precise questions ?
[15:32] <Glenjamin> well, what would be the performance implications of one or the other
[15:33] <vila> The performances should be roughly equivalent AFAIK, you're adding either SSL or ssh encryption anyway
[15:33] <Glenjamin> i think the second option would allow me to supply per-branch permissions as well - but i'm unsure if this would screw up with file ownership
[15:33] <vila> Ha
[15:34] <Glenjamin> my main issue is that i've got to convince our sysadmin that either of these is better than the git equivalent :)
[15:34] <vila> http[s] implies a single user on the remote host whereas ssh allows you to either use a single one or several, it depends on how you configure things
[15:35] <vila> What is the git equivalent on the admin side ?
[15:35] <Glenjamin> well if we were using the https approach, i think we'd have to wrap it with http auth
[15:35] <Glenjamin> there's either git-daemon, which is pretty much bzr serve, so no encryption
[15:35] <Glenjamin> or ssh, where there's a git-shell which is a bit like the python script i linked to above
[15:36] <vila> Right, so both use a single account on the remote host for ssh right ?
[15:37]  * vila thinks a single right should be enough...
[15:37] <Glenjamin> well i was mostly expecting that each developer should have their own credentials, in the https case this would be done via apache, and for ssh it'd just be different user accounts
[15:38] <beuno> vila, hi!
[15:38] <vila> beuno: hey !
[15:40] <vila> Glenjamin: so, if you go the several accounts route, you will encounter more problems to share, since the file system persmissions will come into play, what OS is used on the remote host ?
[15:40] <Glenjamin> ubuntu
[15:41] <Glenjamin> i suppose one user with multiple keys might be a decent compromise, so that each user has different credentials to log in
[15:41] <vila> that's a good point ;) What release (just out of curiosity but also to check which bzr version is available there)
[15:41] <beuno> vila, did you get my messages?
[15:41] <Glenjamin> i think its jaunty
[15:41] <Glenjamin> but i might be able to talk him into adding the bazaar ppa
[15:41] <vila> beuno: not yet I presume, I haven't check my mail yet, but I've seen your messages here, I'll will look at that closely and soon, but may be not right now :-)
[15:42] <vila> Glenjamin: gee, but you already know all the tricks ! Why are you asking questions :-)
[15:42] <Glenjamin> because its not me that needs convincing :)
[15:42] <vila> hehe
[15:42] <Glenjamin> i've been using bazaar for personal stuff, plus bzr-svn at work for a while
[15:43] <Glenjamin> we're looking at using a DVCS for new projects, but everyone else is leaning towards git, mainly because of the visibility of github
[15:43] <Glenjamin> ie. they want to use git because its probably more popular
[15:44] <vila> Glenjamin: So, IMHP, the safer (as in: simpler, less trouble, hence raising less questions from your admin) is the single ssh account and the additional script
[15:44] <Glenjamin> makes sense
[15:44] <Glenjamin> and then loggerhead proxied through apache for web browsing
[15:44] <vila> yup
[15:45] <vila> the biggest question may be which workflows you want to allow, is it a single trunk on the host or also the ability for each dev to mirror its own branches or do you want team branches where several dev can commit too, etc
[15:46] <Glenjamin> i think it'll be a nominated mainline per project, and then all development branches mirrored to the repo
[15:46] <Glenjamin> and there's no real reason i can see why some dev branches cant be shared
[15:46] <vila> beuno: regarding pyftpd, I'm pretty sure I documented the URL to download from somewhere, there is no debian/ubuntu package AFAIK, but it's a single .py file
[15:47] <vila> Glenjamin: then, would you restrict the sharing or should everybody be able to write everywhere ?
[15:47] <Glenjamin> hrm, ideally we'd want to be able to restrict some repos to certain users
[15:48] <vila> Did you look at the bzr_access script too ?
[15:48] <vila> disclaimer: I haven't use either
[15:50] <Glenjamin> i skimmed it earlier, loggerhead is having a fit any wont show it to me anymore :(
[15:50] <vila> Glenjamin: Hey ! Are you related to Neil Martinsen-Burell ???
[15:50] <Glenjamin> i'm not
[15:50] <Glenjamin> but if i recall he's been doing some stuff on a bug i reported
[15:51] <vila> gelnjamin.co.uk having trouble with http authenticated access after redirection ?
[15:51] <Glenjamin> yup
[15:51] <vila> bug #395714 ?
[15:51] <Glenjamin> thats the one
[15:52] <vila> That's weird, he seems to be able to reproduce the bug and I'm not :-/
[15:54] <Glenjamin> hrm, i'll update bzr on this machine and have a go
[15:54] <vila> Regarding your remark about loggerhead, you mean you don't have the bzr sources locally ?
[15:55] <Glenjamin> i'm aiming to replace the setup we have at the moment which is SVN over DAV with apache, where the devs dont have any other access to that box
[15:56] <vila> bzr_access should be far enough with a single ssh account then. At least that's how I understand its intent.
[15:59] <vila> beuno: still there ?
[16:00] <vila> beuno: good, enjoy your week-end :-D
[16:01] <Glenjamin> hrm, i think i'd only be able to get per-branch (ie. per-project) access control if i go back to the bzr+https approach
[16:02] <Glenjamin> the ssh command bazaar sends to start up the hpss using ssh doesnt know which branch its after
[16:04] <beuno> vila, kind of
[16:25] <Glenjamin> vila: updated to all the latest bzr releases, and i can still reproduce the bug
[16:25] <Glenjamin> which version are you using?
[16:26] <vila> Glenjamin: lp:~vila/bzr/395714-auth-redirect
[16:26] <vila> the bug is genuine, but Neil said even eith the fix in the branch above he still reproduces the bug
[16:32] <j85wilson> I have, in my code tree, several directories which contain code which is compiled into executables.  The executables have the same name as the directories.  I want to ignore them, but "Regular expression patterns may not include named or numbered groups."
[16:32] <j85wilson> Is there another way to do this?
[16:32] <j85wilson> RE:\(.*\)/\1 in .bzrignore does not accomplish my goal.
[16:33] <Glenjamin> just be explicit?
[16:33] <j85wilson> Directories get added frequently.  I don't want to have to, and don't want my users (physicists, we're writing analysis code, so my users are coders) to have to, edit .bzrignore all the time.
[16:35] <Glenjamin> could you change the build process to put all the executables in one /bin directory?
[16:35] <j85wilson> I don't think so.  Too much of a break with history-weighted process.
[16:36] <j85wilson> wait, maybe I've got an idea.
[16:36] <j85wilson> .*\(/.*\)*
[16:36] <j85wilson> not sure if I've got the grouping syntax and such correct for python regex
[16:36] <j85wilson> but RTFMing.
[16:37] <mathrick> hey
[16:37] <mathrick> is anyone working on bzr-git currently?
[16:37] <mathrick> putting up with git is beyond my patience, I'd really like to have a sane frontend to that idiocy
[16:38] <vila> mathrick: jelmer is
[16:38] <mathrick> jelmer: my hero!
[16:38] <mathrick> is it any usable? Last I checked (long time ago), it wasn't
[16:39] <vila> mathrick: AFAIK, it is, with less and less restrictions (I don't know the details though)
[16:39]  * mathrick goes to check it out
[16:39] <beuno> vila, I sumbitted a merge proposal for the ignore feature in bzr-upload. I still can't get the testsuite to tun FTPServer tests though
[16:40] <vila> beuno: if you're using python2.6 you need pyftpdlib
[16:40] <mathrick> seriously, git is just retarded
[16:40]  * jelmer waves
[16:40] <mathrick> jelmer: hola!
[16:40] <jelmer> mathrick, everything except for true push should be working
[16:40] <mathrick> cool, can I use it directly on a git-cloned repo, or do I need to go through bzr-git?
[16:41] <beuno> vila, where do I get it from?
[16:41]  * vila searching for the pyftpdlib URL
[16:41] <jelmer> mathrick: you should be able to use it directly on a git-cloned repo
[16:41] <mathrick> yay
[16:41]  * mathrick pants a little
[16:42] <jelmer> mathrick, although that'll still have some quirks since there is no equivalent of the index in bzr
[16:42] <mathrick> jelmer: right now just getting me bzr missing with git will be way awesomer than plain git :)
[16:42] <Glenjamin> i'm having trouble compiling bzr on windows - its saying cannoy find -lzdll, which i'm sure i've seen before and solved
[16:42] <Glenjamin> does it ring any bells for anyone?
[16:42] <jelmer> mathrick, that won't work on remote git repositories unfortunately though
[16:42] <vila> beuno: as mentioned in bzrlib/tests/ftp_server/pyftpdlib.py : http://code.google.com/p/pyftpdlib/
[16:43] <beuno> vila, where do I put it?
[16:43] <vila> beuno: no debian/ubuntu packages AFAIK, but it's a simple py file, put it somewhere in your PYTHONPATH
[16:43] <beuno> aiight
[16:43] <vila> $HOME/lib/python  :)
[16:43] <jelmer> mathrick: (it's impossible since the server doesn't provide enough data)
[16:44] <mathrick> bah
[16:44] <mathrick> why does it have to be so stupid and so popular?
[16:44] <vila> Glenjamin: rings a bell too... should be mentioned in setup.py or one of its included files
[16:45] <mathrick> jelmer: so how does it handle that situation? Or does it not? :)
[16:45] <vila> Glenjamin: or maybe you have to rename something.... isn't there a wiki page about building on windows ?
[16:47] <beuno> vila, that worked!  (and they all passed)
[16:47] <vila> Glenjamin: http://bazaar-vcs.org/WindowsInstall says:     *
[16:47] <vila>       Zlib - zlib123-dll.zip | Homepage Extract zlib123-dll.zip to temporary folder and copy zlib1.dll into C:\Windows\System32
[16:47] <vila> beuno: you rock !
[16:47] <Glenjamin> ah, thanks
[16:47] <beuno> vila, would be nice to make a package for that  *wink*
[16:47] <Glenjamin> there's about 6 different wiki pages for bazaar install on windows :D
[16:48] <vila> Glenjamin: feel free to add comments pointing to the one that works for you ! It's a wiki :D
[16:48] <Glenjamin> i'll test it first, i get the feeling that this step might be more for VS than GCC
[16:53] <jelmer> mathrick: it can't unless somebody fixes the git smart server
[16:54] <jelmer> mathrick, it'll simply tell you "not supported over git://"
[16:54] <j85wilson> well, I haven't been able to make that regex (or any variant thereof) work properly either.
[16:54] <Glenjamin> vila: right, finally managed to build the branch - i get a KeyError on 'protocol' now
[16:54] <Glenjamin> i wonder if it might be simpler to use a defaultdict with None for auth
[16:55] <j85wilson> does anybody know why "Regular expression patterns may not include named or numbered groups."?  This seems a silly restriction to someone who doesn't know anything.
[16:57] <vila> Glenjamin: grrr, I don't think so, the problem is I can't figure which codepath is involved in your case, can you provide me with a .bzr.log for a command run with -Dhttp (taking care of removing any auth header that may reveal your credentials (that includes Basic or Dihest related headers)
[16:58] <mathrick> woah, I haven't visited the bzr page in a long time it seems
[16:58]  * mathrick wonders at the shiny new look
[16:58] <Glenjamin> that url isnt a real branch, its just an empty repository with http auth on
[16:59] <vila> Glenjamin: -Dhttp will tell me :)
[17:00] <vila> Glenjamin: you can send me the .bzr.log by mail if you prefer, if any credentials are left, at least you'll have only one culprit to track :)
[17:00] <Glenjamin> http://pastebin.com/m32e6c375
[17:00] <Glenjamin> i'm fairly sure there arent any valid credentials for that auth config anyway
[17:01] <vila> bzr-svn !
[17:01] <vila> jelmer: are you using the _urllib2_wrappers ability to follow redirections ?
[17:02] <vila> Glenjamin: can you retry that with and additional --no-plugins ?
[17:03] <vila> s/and/an/
[17:03] <Glenjamin> yup
[17:03] <jelmer> vila: no
[17:03] <Glenjamin> works now
[17:03] <vila> :)
[17:03] <vila> :-/
[17:05] <vila> Glenjamin: ok, at least I know where to search (even if jelmer pretends to be innocent :-D)
[17:05] <jelmer> :)
[17:05] <Glenjamin> the issues all stem from trying to access non-existant keys in the auth dict
[17:06] <vila> Glenjamin: that's the apparent cause, a deeper one is that we shouldn't use that code path but one that create the auth dict correctly,
[17:08] <vila> http://pastebin.com/m241bae14 jelmer, you lied :-D
[17:09] <vila> that's really funny because *I* should provide OPTIONS to jelmer so he can use it safely :)
[17:09] <Glenjamin> _urlib2_wrappers.py line 360
[17:10] <vila> Glenjamin: look at the comment just above
[17:10] <jelmer> vila: that's not using the redirections following code though
[17:10] <Glenjamin> the comment just above says "containing at least..."
[17:10] <Glenjamin> and then creates dicts which clearly dont
[17:10] <vila> jelmer: yes it is, except we don;t have the same definition of redirections following code ;)
[17:11] <vila> Glenjamin: that's because callers should do that (look at http/__init__ create_auth)
[17:12] <vila> the redirection code in _urllib2_wrappers don't handle auth correctly (as that bug revealed) instead, callers are supposed to use the do_catching_redirections in transport/__init__
[17:14] <vila> jelmer: the bug is really that bzr http transport don't provide OPTIONS, don't you think ?
[17:14] <vila> rotfl
[17:15] <beuno> vila, thanks for the review,r eplied
[17:15] <jelmer> vila: yeah, HttpTransport should really have options
[17:16] <vila> for those not understanding the joke, we had a discussion with jelmer about whether OPTIONS should be supported by bzr even if it's isn't required internally :D
[17:16]  * vila feels dragged into working on bzr when he should be working on family backups :0)
[17:17] <Glenjamin> now i remember why i gave up trying to follow this call chain when i first reported this one :D
[17:18] <vila> beuno: about the README, I'm all for helping people find the relevant bits, but I'm against duplication since it leads to out-of-date info, how about putting the right pointers in the README ? Or separating the concerns and do that in another merge proposal ?
[17:19] <beuno> vila, sure, I'll drop it from that branch
[17:20] <vila> Glenjamin: yeah, the design is inherited from the python urllib2, in retrospect, it was a bad idea, but it helped building the actual implementation and all the related feaures: connection sharing, proxy support, various auth schemes, all of which weren't supported by urllib2 :-/
[17:21] <vila> eeerk low batteries for keyboard, priority interrupt
[17:50] <mathrick> jelmer: I ran into some crashes: http://pastebin.com/m5b10d31d
[17:54] <jelmer> mathrick, please file a bug
[17:58] <mathrick> ok
[19:58] <sjamaan> Is there a way to get bzr diff to output files which don't include the entire path?
[19:58] <sjamaan> ie, I do bzr diff under foo/bar/qux and it generates a diff for file foo/bar/qux/mooh.c, instead of just mooh.c
[23:19] <eelik> will colocated branches aka fasces be supported soon?
[23:46] <lifeless> eelik: they aren't the same thing, and no one knows :)
[23:46] <eelik> How they are different?
[23:47] <lifeless> fasces is a fullermd term to describe something he would like to see
[23:48] <lifeless> colocated branches is a jelmer created term to describe something he would like to see
[23:56] <eelik> but basically both describe the "git-style branches" where one working tree is shared by many branches in one directory?
[23:56] <eelik> That's a feature I'd like to see...
[23:57] <eelik> So there's actually no schedule for implementing either?