[07:21] <jelmer> moin
[13:14] <jazon> hello everybody I have a problem with my bzr server when I do a push on command line it say me 'not a branch', when I try with bzr explorer it say me 'This transport does not update the working tree'
[13:15] <jelmer> jazon: is the location you're pushing from aqctually a branch?
[13:15] <jazon> Yes
[13:15] <jelmer> jazon: what does "bzr info" say about it?
[13:17] <jazon> jelmer http://pastebin.com/4RuAA94X
[13:23] <jazon> can you help me?
[13:23] <jelmer> jazon: I mean, what does "bzr info" say about your local branch?
[13:24] <jazon> I wasn't in good directory...
[13:26] <jazon> http://pastebin.com/fVr1VAyK
[13:28] <jelmer> jazon: so 'bzr push' works now?
[13:29] <jazon> no, it's say me look at bzr help working-trees
[13:30] <jazon> I don't understand what is the matter
[13:31] <jelmer> jazon: if you push to a remote location, bzr will only update the remote branch
[13:31] <jelmer> not the remote working tree
[13:32] <jelmer> for that, you need to run 'bzr up' on the remote server
[13:32] <jazon> I have to be connected permanently on my server with ssh?
[13:34] <jelmer> jazon: no, you just have to run 'bzr up' after you push to the branch
[13:34] <jelmer> jazon: (if you care about the working tree on the server side)
[13:36] <jazon> I have to run bzr up on the server no?
[13:42] <abentley> jelmer: thanks for your comment on my merge proposal.  Did you mean to vote "approve"?
[13:45] <jelmer> jazon: if you want to have the working tree up to date, yes
[13:45] <jelmer> abentley: Euhm, yes.. one sec
[13:49] <jelmer> abentley: done
[13:49] <abentley> jelmer: thanks.
[18:50] <Qaghan> Guys, for some reason I cannot commit this project to Launchpad, even though I got my ssh set up and everything. I keep getting the bzr: ERROR: Cannot lock LockDir() error
[22:23] <mgrandi> hello
[22:29] <jelmer> hi mgrandi
[22:30] <mgrandi> do you happen to have gpg installed without an agent?
[22:34] <jelmer> mgrandi: I didn't hit this particular issue but some other people did
[22:34] <mgrandi> yeah
[22:34] <mgrandi> well i meant
[22:34] <mgrandi> i'm writing up a response now, but from reading the gpg man page, it seems that $GPG_AGENT_INFO will be defined if you have an agent
[22:34] <jelmer> mgrandi: mgz was able to reproduce it, apparently withtout a gpg agent
[22:34] <mgrandi> and i'm trying to make sure thats the case
[22:34] <mgrandi> cause then we can just check to see if that variable is define, and do the right thing from there
[22:35] <mgrandi> i have no idea what happens on windows
[22:35] <jelmer> mgrandi: I'm wondering if we should be specifying this option at all - shouldn't gpg already try to do this by itself?
[22:35] <mgrandi> thats what i think
[22:36] <mgrandi> but the whole issue was that while bzr explorer, that gpg would be printing stuff out to a tty
[22:36] <mgrandi> gpg: cannot open `/dev/tty': No such device or address
[22:36] <mgrandi>  bzr: ERROR: Failed to GPG sign data with command "['gpg', '--clearsign']"
[22:36] <mgrandi> caused that ^
[22:37] <jelmer> mgrandi: without a gpg agent?
[22:37] <mgrandi> that was with an agent
[22:37] <jelmer> mgrandi: why would it need a tty if you have an agent?
[22:38] <mgrandi> good question. gpg says on the --no-tty option: "Make sure that the TTY (terminal) is never used for any output. This option is needed in some cases because GnuPG sometimes prints warnings to the TTY even if --batch is used."
[22:38] <mgrandi> and --quiet means "try and be as quiet as possible"
[22:39] <mgrandi> i dont know what happens if you tried to use bzr explorer to sign without an agent
[22:39] <jelmer> mgrandi: it's not wrong to use the TTY
[22:39] <mgrandi> yes, but on linux it was saying that it can't find /dev/tty
[22:40] <mgrandi> which makes me think that it doesn't "have" a tty to print out to
[22:40] <mgrandi> cause its a pyqt4 app
[22:40] <jelmer> mgrandi: was that bit fatal though?
[22:40] <mgrandi> it caused signing to fail
[22:40] <mgrandi> the code just says "failed to gpg sign" if gpg doesn't return 0
[22:40] <jelmer> mgrandi: but it tries to use the gpg agent first
[22:43] <mgrandi> i know
[22:43] <mgrandi> i just tried on my mac, if you enter the wrong password, it prints "gpg: Invalid passphrase; please try again ..."
[22:43] <mgrandi> to the terminal
[22:43] <jelmer> rigfht, it falls back to the tty if using the gpg agent doesn't work
[22:43] <mgrandi> im not exactly sure whats going on with the original bug report, since it worked on windows and mac fine
[22:43] <mgrandi> with just 'gpg --clearsign'
[22:44] <mgrandi> but then on linux, somehow running bzr explorer meant it didn't have a tty?
[22:44] <mgrandi> should it have one?
[22:44] <jelmer> mgrandi: I think we should revert the use of --no-tty for now, as it causes a serious regression
[22:44] <jelmer> mgrandi: right, bzr explorer usually wouldn't have a tty
[22:45] <mgrandi> hmm.
[22:45] <mgrandi> thats why i was thinking of just checking to see if we have an agent
[22:45] <mgrandi> cause --no-tty works fine if you have an agent, but if you dont, you need the terminal to enter your passphrase
[22:46] <jelmer> mgrandi: that means there won't be any fallback to the terminal if the gpg agent doesn't work
[22:47] <jelmer> mgrandi: I'm worrying we're trying to do gpg's job for it
[22:47] <jelmer> this sort of logic really belongs in gpg, not bzr
[22:47] <mgrandi> yeah. so if we don't want to use --no-tty, then how do we solve the original problem , with signing failing cause of no /dev/tty
[22:48] <mgrandi> you say you want to use terminal as a fallback, but then if we are using bzr explorer, then there is no terminal (in most cases) to fall back to
[22:48] <mgrandi> err, 1/3 cases
[22:50] <jelmer> mgrandi: I think that might warrant some more investigation, but this is a serious regression.. let's at least fix that bit
[22:50] <mgrandi> yeah
[22:50] <mgrandi> go for it
[22:50] <mgrandi> or do i have to do something
[22:50] <jelmer> mgrandi: no, that's okay
[22:51] <mgrandi> it seems that all of this is just sprouting from the fact that we are using gpg as a separate program rather then a nice api we can bind to, but don't think there is any way around that
[22:51] <jelmer> mgrandi: it would be interesting to use pygmge to do the actual signing
[22:51] <jelmer> but that's a significantly larger change
[22:51] <mgrandi> yeah
[22:51] <mgrandi> but we might have to cause we are juggling where bzr is being run as
[22:52] <mgrandi> command line, bzr explorer started from terminal, bzr explorer without a terminal,
[22:52] <mgrandi> and isn't that the thing that is required to check signatures?
[22:53] <jelmer> mgrandi: how do you mean?
[22:54] <mgrandi> if you try and run bzr verify signatures without it installed you get this: bzr: ERROR: python-gpgme is not installed, it is needed to verify signatures
[22:55] <jelmer> mgrandi: ah, yes, it is
[23:00] <mgrandi> looking at 'gpgme', the gpg folks say that applications should use that, so that seems like that would be the better solution
[23:02] <jelmer> mgrandi: perhaps the best thing to do is to revert this fix, and just consider moving gpgme for signing a proper fix for bug 847388
[23:02] <mgrandi> yeah
[23:02] <mgrandi> i say revert it, the bzr explorer bug can be worked around by just using the terminal to commit