[11:57] can upstart reload service B for me whenever service A is restarted? [12:41] not yet === robbiew_ is now known as robbiew [15:40] ugh [15:40] I hate it when people find bugs that mean I have to spend hours writing 200+ test cases [15:41] Keybuk: which one this time? [15:41] sending a message to a disconnected d-bus connection results in an assertion error ;) [15:41] it doesn't actually affect upstart [15:41] just mountall [15:41] but it's in the most heavily tested bits of the code [15:41] so there's a _lot_ of coverage [15:42] and a test like "sending to a disconnected connection" deserves appearing in each of the coverage cases [15:42] *paste* modify *paste* modify *paste* modify *argh!* [15:42] Keybuk: join with me in the brave new world of software proving. [15:42] sadmac: "if it builds, ship it" ? [15:42] Keybuk: that's a start... [15:43] Keybuk: ultimately we want "if it builds under the most evil compiler ever spat out of hell, ship it." [15:45] compilers that find logic bugs for you. together we can make a difference. [15:49] llvm? :p [15:49] another good move [16:18] gnargh [16:18] arsing valgrind fuck hate bastard grr [16:20] New release name! [16:20] Upstart 1.1 "fuck hate bastard" [16:21] lol [17:02] * Keybuk tries to figure out what the hell he did with these tags and branches [17:04] how bzr [17:15] Meanwhile, i configured automatic failover to 3G if my ADSL is down. \o/ [17:52] oh, but of course [17:52] the valgrind errors are coming from the very last function in the file [17:55] Keybuk: the one you wrote when you are most tired no doubt [17:56] most rushed I think [18:16] argh [18:16] *one* missing free() [18:16] that's all it was [18:44] :-) [19:04] Keybuk: unless i'm reading it wrong, 0.6 contains compat code to talk to sysvinit after upgrade,but not to 0.3.x? [20:42] right [20:43] why? [20:43] because it's reasonably safe to just re-exec 0.3 to 0.6 ? [20:43] the runlevel implementation between sysvinit and upstart isn't sufficiently compatible to make that transition safe [20:45] Keybuk: but re-execing upstart just drops everything on the floor [20:45] where "everything" is not much in 0.3 ;) [20:45] at least in Ubuntu [20:45] Patches are, as always, Welcome :p [20:46] if you need some 0.3 compat code in there, please do add and I'll certainly accept :p [20:46] I'm just lazy [20:46] Keybuk: yeah, been meaning to write that state transfer thing. My point though was if 0.6 just ignores what 0.3 is doing and 0.3 is doing more or less what sysv is doing why can't it ignore what sysv is doing? [20:47] it's still there for Debian [20:47] because they asked to keep it for now [20:47] they like pain. [20:48] also, sysvinit is stupid and simple to code. [20:48] first requirements for fedora's new init system: 1) We must only ever ship 1 init system. 2) We must switch one and only one time. [20:49] 2 kinda fell apart, since Upstart is the switch that keeps on switching, but hey. :) [20:50] I guess 1 might fail too since I think somebody's still got initng packaged (at least I keep finding fedora users that use it.) [20:50] I thought you were NIHing your own? [20:50] * Keybuk read that on LWN somewhere :p [20:51] Keybuk: we were going to nih prcsys (which reeeely needed to be NIHd since its one of the silliest little pieces of code ever written.) [20:51] Keybuk: but prcsys turned out to be as bad of an idea as it was an implementation [20:52] ah [20:52] I heard it was going to be called PulseInit [20:53] Keybuk: ...that sounds familiar, but it wasn't what I was writing [20:53] maybe that was Lennart ;) [20:53] Keybuk: mine was rrn. AFAICT the only public announcement of it ever was at Fudcon, about 2 hours before the meeting where we decided to switch to upstart. [20:53] lol [20:55] Keybuk: would have dumped the code on the internet ages ago, but nobody wants it and... the last bit I was working on is a bit sketchy. It was some circular dependency detection logic, and despite consiting of about 40 lines of code and 2 additional elements in a struct, I still can't figure out how I'd meant it to work (which it doesn't) [20:55] so it was a closed-casket funeral === notting_ is now known as notting === robbiew is now known as robbiew_