=== SDr|BOS is now known as SDr === Md_ is now known as Md === adrien_o1w is now known as adrien_oww [19:29] jodh, slangasek, xnox: there you go, cgroup proposal sent to the ML [19:30] stgraber: thanks!! [19:32] stgraber: \o/ [19:37] the only tricky bits for upstart should be the tear down when the job stops (as it has to check for sub-cgroups and existing tasks). Everything else should be as easy as parsing the cgroup stanza as a nih_list, iterating through it, replacing $auto by some relevant value and doing direct calls to the manager [19:37] no extra parsing should be required, the manager will do all the checks and return a failure code if something's off [21:50] xnox: hey, shouldn't libnih-dbus-dev depend on libdbus-1-dev? [21:50] I'm writing a small dbus daemon as an example for hallyn and noticed that I was missing some headers ;) [21:52] stgraber: according to pkg-config file it doesn't need it. but in practice it links against and uses headers from dbus. so yeah, it needs both fixing .pc file and dependencies. [21:52] stgraber: you want to fix it, or should I ? [21:53] feel free :) [21:53] stgraber: ok. [22:02] xnox: https://github.com/stgraber/cgmanager is the minimal dbus server+service I wrote. Nih really makes this easy once you've spent an hour going through its code to figure out how that's all supposed to work :) [22:02] lol =) yeah. [22:04] stgraber: the dbus stuff, and especially dbus-tests are the most twisted parts of libnih imho. [22:04] having had to debug the code generator a couple of times for upstart, I can certainly agree with that ;) [22:05] stgraber: =)))) i'm glad you did that ;-) [22:05] also the test suite is a major pain because a single char change in the code generator means hours of fun tweaking all the tests ;) [22:14] stgraber: i should finish up fixing the bugs in doc-comments and finally generate and publish libnih & upstart gtkdocs. [22:14] there are more and more libnih users now in the archive. [22:29] xnox: does "needs review" on https://code.launchpad.net/~xnox/upstart/kfreebsd/+merge/196028 mean you're not ready for review yet? [22:29] xnox: er, rather, "work in progress" [22:33] slangasek: it's not ready for review yet. some pieces are useful refactoring, but overall it doesn't pass tests nor compile on kfreebsd yet. [22:33] slangasek: thus at this point this is a standing merge-proposal for me to notice/check for bit-rot & merge conflicts. [22:34] ok [22:35] slangasek: i'm not sure what to do with libnih yet. I will stub out missing nih_watch api for now & it would be ready to be merged upstream...... although keybuk hasn't been receptive of a kfreebsd port yet. [22:35] * slangasek nods [22:36] speaking of which, 13h to my libnih NMU. http://ftp-master.debian.org/deferred.html [22:38] slangasek: second nmu. the first one is not merged upstream, although i believe i've addressed the review knit-picks (w.r.t. style and formatting) [22:39] https://github.com/keybuk/libnih/pull/2 [22:50] xnox: no luck with something like libinotify-kqueue? [22:51] oh I guess you considered it since you uploaded it to Debian ;) [22:53] stgraber: yeah, it fails libnih test-suite with fire =) [22:53] stgraber: cjwatson suggested to get inspiration from glib/gio code as it does alternate between inotify & kqueue/kevents [22:54] and that would be my next move.