=== kentbout is now known as kentb [16:13] cjwatson: mornin' [16:14] cjwatson: I have a log, it turns out that 50mirror isn't even being called, and 'choose-mirror' is being run from main-menu [16:14] cjwatson: the problem for us is that choose-mirror (the C binary?) is trying to 'validate' our mirror [16:14] cjwatson: but our mirror doesn't follow the standard format [16:14] cjwatson: so 'validation' fails [16:14] cjwatson: even though our mirror is totally able to serve packages [16:15] cjwatson: so nominally I want to skip choose-mirror: we have a apt-setup generator that will fill in the right lines for us :/ [16:34] antarus: hmm.... well use early-command or partman-early-command to replace choose-mirror with whatever you want it to do =) [16:35] xnox: it was unclear to me if choose-mirror was installed during early-command [16:35] but yes that was raised :/ [16:35] Eh, early-command shouldn't be necessary. One moment [16:36] antarus: I'd like to look at the log myself, if possible [16:42] cjwatson: yeah, I'm just redacted bits [16:42] redacting* [16:45] cjwatson: hmmm it looks like in this log I tried setting d-i mirror/http and d-i mirror/directory and d-i mirror/proxy [16:46] and the only bug was in d-i mirror proxy, because my proxy ACL was bogus [16:46] (I am not sure what you are looking for in the log, per se, and I've changed a few parameters since Friday ;p) [16:49] (The logs are super helpful, in terms of the DEBCONF_DEBUG stuff, but they have not uncovered any bugs or odd behavior, the code is behaving as expected.) [16:50] antarus: I always find it much easier to read through it myself than to operate at one remove [16:50] And they give me exact values for everything [16:50] true enough [16:50] For instance "mirror/http" doesn't exist so either you've mistranscribed or your preseed file is wrong [16:50] This is the sort of thing I prefer to look at directly in the log [16:50] Tells me about order of events too [16:51] pft! ;p [16:51] And I'm not sure how your mirror might not follow the standard format in a way that can't be handled [16:51] blame thomas [16:51] :) [16:51] Well, no, I'm not sure this is *possible* [16:51] You may just be unaware of the facilities [16:51] ahh [16:53] let me look at my error log again, perhaps we can just fiddle with mirror/http/directory or soemthing [16:53] I admit when this was set up it was a rush job done by someone else, luckily that fellow is back from holiday [16:53] There's also a thing to tweak handling of security [16:53] apt-setup/security_{host,path} [16:54] choose-mirror is always run for netboot installs because the netboot installer retrieves bits of itself from the mirror too [16:55] So fixing up your mirror preseeding is mandatory for netboot installs [16:57] yeah it is possible that we are actualyl being naughty and fetching from us.archive.ubuntu.com [16:57] (but just the installer bits) [16:58] If you fix that up then you *can* skip the choose-mirror call in apt-setup, but it may not buy you very much [16:59] The only plausible alternative is to PXE-boot something like a monolithic d-i mini.iso which you've built specially; that's possible but it's starting to get a bit off the reservation [17:00] (The monolithic flavour has everything built into the initrd rather than relying on the network. But it's not built by default and mainly intended for use by d-i developers.) [17:02] ahh [17:02] so I talked to the guy who set this upe [17:02] his understanding is that our repo names are not the repo names the installer is expecting [17:03] which is a bug we need to fix in the future [17:03] ;p [17:03] I need concrete details if at all possible [17:03] Preferably that log :) [17:03] well here [17:03] let me set the mirror lines to point at rapture [17:03] and do an install [17:03] and then I'll have that log [17:04] There are definitely meant to be ways to cope with this kind of thing, and if it can't then I suspect I will consider it a d-i bug [17:06] on a positive note, this did lead to a bug in our pxe infrastructure that downcased extra arguments ;p [17:08] heh [17:09] ahh [17:09] asprintf(&command, "wget -q %s://%s%s/dists/%s/Release -O - | grep -E '^(Suite|Codename):'", [17:09] protocol, hostname, directory, name); [17:09] I see [17:10] where is 'name' set.. [17:11] it is possible that if we preseed 'suite' to be the right name, it would work [17:11] its not clear if preseeding the suite has other negative naming effects [17:12] That's one possibility, yes [17:12] For the most part I think it should be OK [17:12] * antarus nods [17:12] I'll try taht next [17:17] cjwatson: ok, how shall I get this log to you? [17:18] cjwatson@ubuntu.com is fine [17:18] feel free to encrypt if necessary [17:21] nah I've removed all of the juicy bits ;p [17:21] ugh [17:21] chromeos-- [17:29] ok log sent [17:48] hrm, suite sets codename... [17:48] so I need to change that as well [17:48] Should be picked up from the archive's Release file [17:48] (I'll have a look properly later, getting towards dinnertime here) [17:49] oh [17:49] Codename: ubuntu-precise-proposed-main-20130614-2 [17:49] yeah [17:49] thats a bug on our side [17:49] ;p [20:21] cjwatson: the more I look at this, the more the hate piles on ;) [20:26] * antarus thinks he has something that will work though [22:29] cjwatson: and another bug in our mirroring system ;p