[10:38] so localechooser/supported-locales [10:39] is something supposed to update this? [10:40] I'm noticing that (for example) xxx-en-gb packages are removed by ubiquity, because that contains "en_US.UTF-8" and then that's passed on to check-language-support -l. [10:48] ah debian-installer/locale too, but this is apparently not set in the en_GB case? [11:07] hmm, it is, but something's overwriting it [11:08] I guess [11:09] xnox: want to help me? :-) [11:10] https://paste.debian.net/1046987 [11:10] between line 13 and 24 it got changed back [12:20] localechooser/supported-locales horum... i can't recall how it was done. either like at localechooser built-time or like at ubiquity built-time or like at langpack/language updates time [12:21] i think i never learned how to do translations (and supported locales) updates in ubiquity. [12:21] cause somebody else picked it up on the team, or like i rage quit working for the company, or some such. [12:21] better luck with infinity maybe?! [12:28] something is making it get reverted back to en_US [12:29] I don't yet understand all this debconf business [12:31] Laney, all i know that it is `different` given the monkey-patch at ./d-i/patches/localechooser-post-base-installer.patch [12:31] cause langpacks are done differently in ubiquity, but we still wanted to support the same d-i preseed questions or some such [12:32] yeah I don't think it is localechooser itself [12:32] it's fine after that finishes (line 13) [12:32] I'm looking for something in ubiquity I think [12:54] ooooh. there mayhave been cdimage stuff as well, cause i think it writes out to disk as to what langauge packs are included on the cd. [12:54] cause i think it we did (at one point) supported "re-rolling" the iso with more or less language packs without like rebuilding ubiquity. [12:55] literarly cannot remember any of those details. [12:56] it's to do with the debconf database in install.py not seeing the updates that happen from the timezone page [12:59] juliank was working on those sort of things lately (frontend database locking stuff, but maybe that was dpkg/apt rather than debconf) [13:05] ok, I found a change that fixes it I think, not sure if it's right though [13:18] * Laney fake news :( [15:33] https://bugs.launchpad.net/ubuntu/+source/ubiquity/+bug/1797579 [15:34] cyphermox: xnox: maybe one of you knows the debconf stuff in ubiquity better than me and can tell me why that patch is wrong and/or how to do it better [15:34] would be nice to fix in cosmic if we can [15:34] i think we need juliank for this. [15:36] 😱 [15:36] the cavalry [15:36] yeah, there's only one instance of debconf somewhere at the beginning [15:36] there's this parallel db thing [15:37] I don't know what on earth that is [15:37] no clue. [15:37] wasn't this working before though? [15:38] presumably it was at some time [15:38] I didn't try previous releases tho [15:38] I'm suspicious of console-setup [15:38] maybe I have a old iso or something [15:38] that could have gone broken, as it tends to every release in some obscure way [15:38] * Laney has wily [15:38] actually, trusty [15:41] Laney: in your bug, which part of the log says the installer forgot about the language? or is it not shown? [15:41] oh, right, only install.py [15:41] cyphermox: where it says Oct 12 15:48:42 ubuntu /install.py[9568]: keeping language packs for: en_US.UTF-8 [15:41] it's not actually completely forgotten [15:41] e.g. I get en_GB locale [15:41] yeah [15:42] well, I can see install.py having a separate, wrong debconf [15:42] it does have that [15:43] I ended up booting a xenial iso [15:43] it has this bug [15:44] bit different though, in that one localechooser is run too late [15:46] hold on [15:47] is that when you pick language in the selector in ubiquity? [15:47] or when you pick language from gfxboot? [15:47] the two might behave differently. [15:47] (it's wrong, but I think I see why they would behave differently) [15:48] I did pick english at gfxboot, and then english and english(uk) and clicked on the UK in ubiquity [15:49] well, timezone won't change anything but timezone [15:50] english in gfxboot I don't remember what it sets [15:50] me neither [15:50] but it does at least know that I chose en_GB after running localechooser [15:51] yeah, but I wonder if the issue is that one of the code paths doesn't set debian-installer/locale somewhere it should be set [15:54] db_get can see it right up until it can't any more (in install.py) [15:55] ok [15:55] can you check what pkgsel/language-packs is set to? [15:56] sure, at which point? [16:00] uh, anywhere after picking language, and then around the end of the install (when install.py runs) [16:00] I don't think it'd be much different between the two [16:00] wait, no [16:00] don't bother [16:01] I remember this is a preseeded thing on CD [16:08] it's not? [16:08] hrm [16:08] what's this? [16:09] I expected pkgsel/language-packs to be preseeded to something [16:09] but I went to check all the way back to precise [16:12] that codepath is not used in a regular iso install [16:12] it's the localechooser/supported-locales one [16:12] supplemented with debian-installer/locale which is what ubiquity decided you wanted [16:14] (that last one is the one that is buggily not set properly) [16:18] well, either one should be set to /something/ [16:18] supported-locales is en_US.UTF-8 for me [16:20] ok [16:20] so maybe we screwed up localechooser? [17:03] ohhhh [17:03] is this a race? [17:03] we start installing in the background right? [18:15] yeah [18:15] but like I said, the timezone doesn't really matter [18:15] you don't start install until after partitioning, which is after picking language [18:15] (but before picking keyboard) [18:15] that said, you did point out what matters is localchooser, which is run as you pick language, not when you pick keybaord [18:16] (and it's preseeded even earlier if you pick language at gfxboot) [18:17] I think we should debug this further, add enought prints in the code to output what the debconf values are at the time install_misc.py runs [18:42] Laney: could be racey [18:42] looks like I got fr_CA working first try [22:31] cyphermox: it takes the country you pick in the timezone step into account [22:31] I think the fix would be basically to not start installing until after ubi-tiemzone [22:32] but you have to do that weird debconf stop / start thing to make sure it is committed at that time afaics [22:32] will test it a bit more and probably propose something over the weekend