=== harlowja is now known as harlowja_away [14:51] harmw_, around ? [15:00] hm.. harmw_ someone is asking for respecting mtu returned in dhcp. [15:00] i wish i'd have fixed it better with your staticroutes fix [15:01] so that we weren't hard coding that in a binary [15:01] but instead wrapping a udch call or something === harlowja_away is now known as harlowja [18:19] smoser: hr === harmw_ is now known as harmw [18:20] hey. [18:20] i wish i'd just wrapped udchpc [18:21] so that i didn't have re-build in order to change the options that it is called with on ifup [18:21] that just seems silly [18:21] ah [18:21] rebuilding a cirros image to chang a udhcpclint thingy, right [18:21] would be silly inded [18:23] is tht someone asking on lp? [18:23] * harmw likes some place to read the whole story :) [18:24] btw, Ive been trying to get me a new cirros image with updated kernel+mdev+hypervdrivers but after booting in hyperv it apparently forgot the initrd - while booting in qemu-kvm just goes without problems [18:30] yeah, ubuntu kernel will work fine in qemu-kvm with built in drivers. [18:31] not asking in launchpad, no. [18:31] they were just setting the mtu on a network and sending that in dhcp [18:31] and cirros wasn't oblighing [18:31] obliging [18:32] so cirros should just apply whatever DHCP options are inside the dhcp [18:33] yeah. [18:33] sounds like a cool feature [18:33] for 0.4 or somthing :p [18:33] i'm not really sure how udchp does it. [18:33] ie, udhcp is calling that ifupdown script [18:33] nah, its calling the default.script [18:33] and must be putting into environment varaibles names of the reponses. [18:33] from /usr/something [18:34] yeah. thats what i meant. [18:34] ifupdown -> udchp -> default.script [18:34] yep [18:36] http://git.busybox.net/busybox/tree/networking/udhcp/README.udhcpc?h=1_1_stable&id=179f41778880d0a233435f5c9ed04e17316f479a#n68 [18:37] we can hack default.script to just apply whatever is set, in a reasonable way [18:37] hm.. [18:37] so why the change we made for '-O' [18:37] well, not just everything ofc [18:37] -O made it get these extras in the first place [18:37] http://paste.ubuntu.com/7091665/ [18:38] so we have to one by one ask for things ? [18:38] [annoying] [18:38] -O,--request-option OPTRequest option OPT from server (cumulative) [18:39] thats annoying, yes [18:39] ill create a bug on LP first [18:39] for future reference [18:39] since I take it you're not diving in on this one right now :) [18:45] right. [18:45] did you try my last suggestion on that iamroot thing? [18:46] uhm [18:46] http://paste.ubuntu.com/7091714/ [18:46] hm, guess I missed that :| [18:46] sry, let me check [18:48] great, that works smoser [18:48] ok. i'm going to commit it then. [18:48] I've did a simple test with just those 2 functions [18:48] iamroot and asroot [18:49] its using sudo whn it should [18:49] so, ok [18:49] nice [18:52] smoser: please hold on the cirros-status merge [18:53] ? [18:53] grep DMI /var/log/messages gives no luck on HyperV [18:53] while dmesg|grep DMI does [18:57] really. [18:57] yup [18:57] the benefit of /var/log/messages would be that its permenant [18:57] dmesg is a circular buffer [18:57] indeed [18:58] though I wonder on how much is changed for dmesg, and how mch that would matter [18:58] i honestly dont even know how /var/log/messages is written :) [18:58] pretty much anything that "just worked" with buildroot i did not bother figuring out. [18:59] isn't syslog just doing that? [19:00] hm, now my qemu vm has no /etc/mdev.conf while the hyperv vm does [19:01] shouldnt I apply http://paste.ubuntu.com/7091714/ on my branch btw? then change this dmesg thingy and then you can just merge, right? [19:02] is that right? does your branch have nothing else? [19:02] if so, yeah, go for it. [19:03] your hyperv you re-built cirros [19:03] i thought [19:03] uhu, new kernel as well [19:03] right. well that'd be where mdev.conf came from silly [19:03] oh. odd. its built in. [19:04] i have the same disk.img loaded in qemu [19:15] odd [19:17] http://bazaar.launchpad.net/~harmw/cirros/cirros-sysinfo/changes [19:18] doesn't look like unwanted changes in there :) [19:19] https://code.launchpad.net/~harmw/cirros/cirros-sysinfo/+merge/209322 [19:19] k. [19:19] gracias [19:25] harmw, merged. and pushed. thankm you. [19:27] nice, thanks [19:34] hm, LP didn't mail me about any merge [19:46] smoser: any reason why console logs are disabled on tty0 from end of kernel boot until login prompt? [19:47] (its logged to file though) [19:48] yes. [19:48] :) [19:48] because there is no easy way to dot hat. [19:49] kernel messages go to each target of 'console=' on the cmdline. [19:49] indeed [19:49] but init's output (and any process it spawns) can only go to one place [19:49] hm yes [19:49] outside of using 'tee' or something. [19:49] so the code tries to figure out what the right place is. [19:49] and i prefer serial console to vga [19:50] +1 on that [19:50] i'm not opposed to some way to make it go to both places. [19:50] as long as it works. [19:51] ok [19:51] we have the same problem on ubntu too. [19:51] its a real PITA [19:51] the biggest pain, is that you can't actually tell if /dev/console is working [19:51] echo "foo" > /dev/console [19:52] works for a while [19:52] (kernel buffers) [19:52] and then it realizes that there is no /dev/ttyS0 (or wherever you had 'console=' on the cmdline) [19:52] and it then starts to fail [19:52] so, useless [19:53] https://bugs.launchpad.net/ubuntu/+source/cloud-initramfs-tools/+bug/1123220 [20:38] ah ys [20:39] smoser: any bugs/issues still open before 0.3.2 final? [20:47] no. [20:47] jsut time. [20:47] i jsut have to do it. [20:49] ok, thn take time :) [20:54] alright. so lets say i'm just going to go cowboy and do it. [20:54] and call it 0.3.2 with minimal-ish testing [20:54] but then we'll test it fairly well before we announce that anywhere. [20:55] you could go rc1.. :) [20:55] but 0.3.2 is better [20:55] yeah. the problem is then people like you just want to put more stuff in [20:55] (or people like me too, as i'd like the MTU thing fixed) [20:56] so 0.3.2 it is :) [21:09] you know, a young go getter like yourself could set up a nightly build service [21:09] :) [21:09] and automate all this. and get ifrasturcture infor gated c-i [21:09] doesn't LP supply thta? [21:09] it will build debs [21:10] i dont think i could shoe-horn this in there. [21:10] it might be interesting to try though :) [21:10] hmk, well hosting something Jenkins-like for cirros or c-i sounds like a nice thing :) [21:13] or just somthing manually, without Jenkins [21:18] i like the idea smoser [22:19] smoser: whre shhould I put nightly disk images