[21:02] morning folks [21:05] thumper: you haven't turned into a zombie then? [21:06] no... survived with all my lives [21:09] thumper, fully encrypted email to public list? [21:09] hazmat: yeah, no idea what happened there [21:09] * thumper is trying to work it out [21:10] I can't even open the item in my sent mail :-( [21:10] I wrote heaps [21:10] damn [21:10] fwereade_, i'm not convinced that the something-changed hook is actually a step forward compared to the default-hook impl.. the former requires rewriting charms with an entirely different frame of mind to utilize, while the later is a simple optimization that removes tedium. [21:11] although both need a feature flag [21:29] thumper: defeated by the gpg :P [21:34] rick_h__: yeah... [22:37] * thumper goes to make another coffee [22:52] lol - Panic: Invalid user tag "not touched" (PC=0x414676) [22:52] panic: Invalid user tag "" [22:55] gnn,,, [22:55] hmm... even === Guest79828 is now known as wallyworld [23:24] thumper: waigani menn0 https://github.com/juju/juju/pull/531 [23:25] davecheney, thumper, waigani: and here's mine https://github.com/juju/juju/pull/508 [23:25] davecheney: looking at yours now [23:26] hmm... [23:26] thumper: interesting lxc issue in bug 1357552 - some but not all containers can't see the http server on port 8040 which is used to provide tools into the containers via curl. sigh [23:26] Bug #1357552: lxc containers created, juju can't seen or communicate with them

[23:27] I'm wanting the git equivalent of 'bzr diff -r -2..-1' [23:27] I want the diff of the last merge [23:27] git diff HEAD^ HEAD doesn't quite do it [23:27] as it shows the last commit, not the last merge diff [23:27] any git gurus? [23:27] wallyworld: that's weird [23:27] wallyworld: sounds like an lxc networking problem [23:28] we have a few lxc issues at the moment [23:28] * thumper nods [23:28] :-( [23:28] i could understand if all containers failed [23:28] but not some [23:29] thumper: i'm thinking we should just mount a host dir into the container and grab the tools from there rather than a http server [23:29] wallyworld: why not use wget -c [23:29] i wonder if that will work for nested [23:29] thumper: does "git show -m " give you what you want? [23:29] then it'll loop automagically [23:30] wallyworld: I'd really prefer if we didn't treat lxc too specially [23:30] but I can see the desire [23:30] davecheney: i have no idea, someone else used curl [23:30] menn0: shows the last commit [23:30] not the last merge [23:30] thumper: lxc is already special - there's a big hack in machine agent to start the http server :-( [23:31] wallyworld: yeah, but we are changing that aren't we? [23:31] we want machine 0 to be a container [23:31] any change like this just makes that move harder [23:31] thumper: sure, but we do need a way to get stuff passed in [23:31] agreed [23:32] but lets not make it hard for us to move where we want to go [23:33] how is mounting a host dir making it hard? don't we want to do that anyway for charm develppment? [23:33] thumper: does this help? http://stackoverflow.com/a/7335580 [23:35] wallyworld: yes we do (ish) [23:37] hah... found out why we are now panicing making user tags [23:37] in the past, when we created a user tag from a string, there was no validation [23:37] now there is [23:37] thumper: BOMM! [23:37] names.NewXXXTag is only for testing [23:38] or when you are very very sure that what you are passing it is a valid tag [23:38] in which case, you should probably just use names.ParseXXXTag [23:39] should still be valid though [23:39] what is the line ? [23:40] environs/cloudinit_test.go line 49 ish [23:41] although I have another [23:41] Tag: names.NewUserTag(info.APICredentials().User), [23:41] from juju/api.go [23:41] seems that the user is blank [23:41] blank isn't a valid tag [23:41] s/tag/user/ [23:41] grr [23:41] bingo [23:42] thumper: why do the apiserver tests embed another test suite ? [23:42] type rsyslogSuite struct { testing.JujuConnSuite *commontesting.EnvironWatcherTest [23:42] NFI [23:42] the embedding with a pointer means that the EnvironWEatcherTets are run when the rsyslogSuite tests are rung [23:42] run [23:42] ah... [23:42] yes, that is by design [23:42] continue [23:43] * thumper thinks back [23:43] s.EnvironWatcherTest = commontesting.NewEnvironWatcherTest( [23:43] api, s.State, s.resources, commontesting.NoSecrets) [23:43] oh [23:43] there were situations where each other test had a copy of the common test functions [23:43] this is somet facade thing [23:43] I consolidated them [23:43] yeah [23:43] hmm [23:43] i still don't see what it does [23:43] it isn't great [23:43] it repeats a lot of stuff [23:44] but it is better than it was [23:44] well [23:44] i sort of do [23:44] I agree it isn't correct yet [23:44] just 'less shit' [23:44] 'cos it is this logic that does the canAuth("") logic [23:46] lucky(~/src/github.com/juju/juju/state/apiserver) % grep -e 'can.*(\"' -r * [23:46] common/environwatcher.go: if !canWatch("") { [23:46] common/environwatcher.go: if !canReadSecrets("") { [23:46] common/environmachineswatcher.go: if !canWatch("") { [23:46] provisioner/provisioner.go: if !canWatch("") { [23:46] the best i can make from this logic is [23:46] if !anyone is allowed to watch ... [23:52] davecheney: review done [23:55] ta [23:57] thumper: you are assigned to bug 1320543, are you working on it? [23:57] Bug #1320543: debug-log uses internal names for filtering

[23:57] wallyworld: no [23:57] ok [23:57] * wallyworld unassigns