[08:39] Anyone up? [11:01] hi! [11:02] i have a problem, i was commiting a (new) file to my repository and lost the internet connection. now this happens: bzr: ERROR: Cannot lock LockDir(sftp://bzr@aaa.dyndns.org/srv/bzr/aaa/.bzr/branch/lock): File exists: '/srv/bzr/aaa/.bzr/branch/lock': Failure: unable to mkdir [11:02] (when i try to commit) [11:02] what can i do now? [11:04] halp! [11:05] i've tried doing bzr break-lock to no avail [13:23] cheater__: hmm [13:23] cheater__: what url do you give to break-lock, and what does it say? [14:53] LarstiQ, that's all it says! [14:54] cheater__: this is what break-lock says? [14:59] break-lock says nothing [15:01] cheater__: what argument do you give to break-lock? [15:03] none [15:03] what should i be giving it? [15:03] oh btw, the lock file that commit mentions changes every time [15:43] cheater__: (it helps if you address me so I get a highlight and check this channel) [15:44] cheater__: try `bzr break-lock sftp://bzr@aaa.dyndns.org/srv/bzr/aaa/`, possibly with 'aaa' replaced by the relevant information [15:48] ok let me try [15:54] LarstiQ, no improvement [15:55] cheater__: same error? [15:55] or, same nothing? [15:55] same error [15:55] cannot LockDir [15:56] bzr: ERROR: Cannot lock LockDir(sftp://bzr@aaa.dyndns.org/srv/bzr/aaa/.bzr/branch/lock): File exists: '/srv/bzr/aaa/.bzr/branch/lock': Failure: unable to mkdir [15:56] cheater__: is /srv/bzr/aaa/.bzr/branch/lock a directory or a file? [15:57] just a sec [15:58] it's nothing. [15:58] * LarstiQ blinks [15:58] it used to be a dir, but i was getting that error, so i moved it to lock-old [15:58] and the error persists [15:58] euh [15:59] that is at least somewhat weird [15:59] cheater__: does ~/.bzr.log have more info? [15:59] on the account i'm trying to commit from? [15:59] cheater__: yeah [16:00] just a sec [16:02] yeah [16:02] in fact i think it's a lock that's on client side [16:02] how does bzr do locking on the client side? [16:04] * LarstiQ would need to read code for that [16:04] lockdir.py [16:04] among things :) [16:04] that's where the exception happens [16:04] cheater__: I believe there are some traces of os locking left [16:04] cheater__: do you have a full traceback? [16:10] http://pastebin.com/E3BREgAv [16:11] LarstiQ, ^ [16:11] ah hm [16:12] * LarstiQ looks at the code [16:15] cheater__: the code notices .bzr/branch/lock doesn't exist and tries to make it, and that then fails [16:15] not sure why [16:16] cheater__: if you are in a position to make a backup of /srv/bzr/aaa that might aid future debugging as to what is going on here [16:16] yeah i'm root on that server [16:16] cheater__: as to a workaround now, what happens if you create .bzr/branch/lock/ yourself? [16:17] cheater__: another idea is to use bzr+ssh:// instead of sftp:// [16:17] cheater__: you said earlier you had a problem so you moved it out of the way [16:17] cheater__: hm [16:17] yeah i moved it on the server [16:17] but this error seems to be happening on the client [16:17] * LarstiQ wonders if something funky is going on with the filesystem [16:18] cheater__: the client is reporting the error it gets from the transport [16:18] cheater__: but you could check if there is a /srv/bzr/aaa/.bzr/branch/lock on the client [16:20] no, /srv is empty. [16:21] right, so I remain convinced then that it is reporting about the serverside situation [16:21] * LarstiQ goes through the release-notes in the meantime [16:21] cheater__: not the same bug, but https://bugs.launchpad.net/bzr/+bug/733350 got fixed in 2.3.2 [16:21] Ubuntu bug 733350 in bzr (Ubuntu Natty) "LockContention error when pushing (with new tag) to a bound branch" [High,Fix committed] [16:22] cheater__: I need to run to the store before it closes [16:24] let me have a look [16:24] LarstiQ, thanks for your help. if you have time afterwards please give me a shout. [16:25] cheater__: yeah, I'll be back in 45 minutes or so [16:28] LarstiQ, cool! [17:52] cheater__: messed up timing with dinner a bit, before I go back, did you try making the lock directory and seeing how that changed things? [18:05] LarstiQ, actually no, let me try [18:07] actually the lock dir has been created automatically. [18:07] i tried committing again, and now it argued that a subdir of lock exists [18:07] bzr: ERROR: Cannot lock LockDir(sftp://bzr@aaa.dyndns.org/srv/bzr/aaa/.bzr/branch/lock): File exists: '/srv/bzr/aaa/.bzr/branch/lock/gacus7hzmf.tmp': Failure: unable to mkdir [18:22] cheater__: it sounds like something weird with either the filesystem or ssh is going on [18:25] i restarted ssh and it didn't help [18:26] oh wait! [18:26] it did! [18:26] but not really. [18:26] now, bzr forgot my name ?? [18:26] i need to do whoami again on the client. wtf? [18:28] hah it's uploading [18:29] yay, restarting ssh did it [18:29] but why did bzr forget who i am? [18:29] that's just bizarre. [18:32] cheater__: *blink* [18:33] cheater__: did something happen to your bazaar config files? [18:33] strange indeed [18:33] nope! [18:33] the thing is i'm on a very faulty internet connection [18:34] like 30% packet loss [18:34] i've had to kill bzr a few times when trying to commit [18:34] maybe it was in a half-open state or something [18:37] ah hmm [18:38] cheater__: amything weird in ~/.bazaar ? [18:38] leftover locks maybe [18:55] let's see [18:56] nope ~/.bazaar/lock is empty [18:56] but of course it could have been non-empty before [18:56] in FACT [18:56] what i think might have happened is that i killed bzr by ctrl-c but somehow an ssh connection kept lingering [18:57] i've noticed runaway processes when i do ctrl-c on bzr: i'll do ctrl-c while it's logging in, and it will later keep spamming my tty with ssh authentication challenge dialogs [18:57] so i'm thinking maybe restarting openssh on the SERVER has taken that out [18:59] i think that's fairly clunky. i was actually fairly disappointed. [20:08] cheater__: that ssh repeating is an openssh thing [20:09] don't know about serverside lingering processes === yofel_ is now known as yofel [23:52] LarstiQ, i think the process was lingering CLIENT-side [23:52] except when the server quit, so has the client process