libera/#devuan-dev/ Saturday, 2019-10-05

* rrq finally got my d-i iso build branches to work00:02
rrqnew netinstall candidates at https://borta.rrq.id.au -- with firmware00:03
rrq(I think) please test00:03
Jjp137rrq, the standard install (both graphical and console) asks the user to choose an init system; shouldn't that be for the expert install only?01:37
golinuxIt used to be but hard to find.  Trying to make it more noticeable.   Maybe that should only be during an expert install.01:40
golinuxOr maybe it needs an option for "defaut init" . . . assuming the user knows what an init is.  On thing leads to another . . .01:42
golinuxOn > One01:43
golinux(afk01:43
golinux)01:43
Jjp137yeah I recall it being hard to find even within an expert install; in 2.0, one had to load the "choose-init" component and it's really easy to miss01:45
rrqafaiu, the choice of "where" it comes in is programmed in the udeb (choose-init), while the ISO composition choice is whether to include it at all or not.01:46
fsmithredIn the 2.0 isos there's a checkbox in the installer components list that causes choose-init to run later01:46
fsmithredbut it's easy to miss it01:47
golinuxIf we want to support alternate inits they need to be more obviously available but to the audience that won;t be confused by that question.01:49
fsmithredI agree01:49
golinuxOff to a big bowl of kabocha01:51
rrqchoose-init=0.2 is the current ascii version. beowulf has 0.2.101:53
rrqthe ISO packing choice is whether or not to include it.01:54
rrqthe udeb itself "decides" where and how it appears in menues01:54
rrqafaiui01:55
fsmithredI just went through network setup twice - once in the beginning and once when asked to choose a mirror, right before tasksel02:03
fsmithredit asks if I want non-free but doesn't ask if I want contrib02:04
rrqyes, I'm not sure why. I guess the first time is for installing the OS and the second time is for the installed OS.02:04
fsmithredmaybe. I didn't get that when i installed buster a few days ago02:05
fsmithredand I don't recall doing it before in devuan.02:05
rrqI'm not sure which udeb(s) are doing that. (again: packing choice is not wrt to function :))02:06
rrqto tune the process, we need to upgrade ascii udeb(s) to suit.02:07
fsmithredupgrade to change their behavior?02:07
fsmithredwhy do we need udebs? (philosophical question)02:08
rrqyes, that's a new version of them/some. and ascii has particular versions.02:08
rrqonly some udebs are forked.02:08
rrqso far02:09
fsmithred(shudder)02:09
rrqthe installer "is" a collection of udebs.02:09
fsmithredI understand that the udeb is a trimmed-down package so it fits on the CD02:09
fsmithredbut why do we need packages?02:09
fsmithredthat's just to run the installer?02:10
fsmithredthe installed system gets built from real deb packages, right?02:10
rrqbasically a udeb is a puzzle piece for the installer, and the insteller "is" exactly the collection of choosen udebs.02:10
fsmithredbrb, need to talk to the gas man02:10
masonAh, it's really a µdeb then, eh? :)02:11
rrq:)02:11
fsmithredhow do you do that?02:13
rrqwhe? what?02:14
fsmithredoh, hot shit. Impossible to miss 'Select an init system'02:14
fsmithredhow to type mu02:14
fsmithred?02:14
fsmithred(in one char)02:14
masonfsmithred: I use xim, so it's meta-m-u02:15
masonSorry, compose-m-u02:16
masonCan type µ, and ☭, and ², and π, and a host of other goodies.02:16
masonfsmithred: In my .xsession I say export GTK_IM_MODULE=xim and export QT_IM_MODULE=xim and unset XMODIFIERS, that last being a bit cargo-cultish.02:17
fsmithredthanks02:18
masonfsmithred: Happy to share. xim has gone out of favour, but I hugely prefer it to the newer stuff.02:19
rrqιτ'σ αλσο ποσσιβλε το ψηανγε κευβοαρδ02:20
fsmithredinstall is done. Looks like I got the whole desktop ok02:20
rrqit's also possible to change keyboard02:20
masonAh, I couldn't read the ψηανγε02:20
masonI got the rest!02:20
rrqthe old CD-2&3 only extend "main" .. no contrib or non-free02:21
fsmithredconrib gets added with non-free02:22
masonbbiab - ISP is falling apart and the dropped packets are making IRC almost impossible02:22
rrqold = ascii 2.0.002:22
fsmithredyeah, that makes sense02:22
fsmithredall the firmware goes on CD102:23
rrqI might mimic that (like I did for install & cd-1)02:24
fsmithredwhat do you mean?02:25
rrqI took the same pool packages02:26
fsmithredyeah, that should work02:26
fsmithredfor beowulf, there will be some changes in package names02:26
fsmithredit's always that way02:27
rrqi've got the poolonly "build flavour" working .. about to upload i386_cd-[123]02:33
rrqfor testing02:33
rrqwe'll need to decide whether udeb upgrades are needed for ascii 2.1, and do them, if so.02:34
rrqi'll round it off (later) today with the i386_DVD (which might become too large)02:38
fsmithredthat's the other thing that always happens02:40
rrqmaybe have to drop the "gnome" option or something ;)02:42
golinux"gnome" has never been an option in the installer02:52
rrqtexlive-extra is 1.1G (on ascii 2.0.0 i386 dvd-1)02:56
golinuxI have no idea what that is02:57
rrqwell if your where publishing research papers around the millenium shift, you'd know :)03:00
rrqgnome-* is only shy of .25G03:01
rrqin parity (+10%) with the k* packages (on dvd-1)03:04
golinuxThe official lineup:  Default: Xfce03:04
golinuxAlternate: Cinnamon, KDE, LXQt, MATE03:04
golinuxOthers: available post-install03:04
fsmithredhave to go out. bbl.03:07
rrqI would suggest pushing texlive-extra onto a dvd-2 (if any). would resolve the size issue and even give room for more games or something03:07
Unit193So I may have asked this before, but it was lost in scrollback.  Any chance Devuan will set up a copy of tracker.d.o?  Seems it's portable enough for Kali to do so, https://pkg.kali.org/03:08
golinuxrrq: Note that the install docs might also be a few kb larger if we add  a screenshot + text for the openrc option03:28
golinuxbbiab03:29
masonfsmithred: Ah, I forgot to mention include "/usr/share/X11/locale/en_US.UTF-8/Compose" in ~/.XCompose. And my ISP is still dropping packets like a champ.03:52
rrqUnit193: many Devuanites are beutifying atm, but though it's good with means to increase communication, the service cost of a web site is generally 1 man day per month per page, plus any set up cost of, say 10 man days; all in all a tad expensive for the present resource pool. Maybe you should chaeck at #devuan if someone has anything like that set up that you could attach to. And then there's the option04:06
rrqthat you commit yourself to the task.04:06
golinux+1!04:46
Unit193rrq: OK, so it seems like it hasn't really, due to lack of time and manpower.  That really makes sense, yeah.05:09
gnu_srs1LeePen: See the mail to -dev about mate in Debian.11:49
rrqfsmithred: production line all in hand, I believe; now it's a matter of getting the doc/ dir right13:35
rrq( or docs/ rather )13:36
LeePengnu_srs1: Thanks. Just replying.14:17
LeePenrrq: Hi. I am still being held up by the failed arm64 builds from a few weeks ago.14:31
LeePenWhen I diffed the successful<->failed logs, some of the arm64 build host versions we behind, notable devuan-jenkins-glue, and it was14:32
LeePennot using eatmydata. Is it worth/possible to update the build environment and see if that fixes the build?14:32
rrqdid you check the most recent debian-installer builds .. if Dan's done something?14:33
rrqI don't think I have access to the build host14:33
LeePenOK, thanks. I haven't heard that he has, but I will chack the debian-installer build logs.14:34
rrqjenkins-debian-glue version 0.17.0+devuan1  .. is old?14:34
rrq(yes)14:34
LeePenI will have to check the exact versions again.14:35
rrqyes, that's same as the last failing policykit14:35
rrqhmm there's some build configuration on Jenkins .. still, safest to get Dan's help14:43
LeePenOK. I'll wait until he is around. Thanks.14:44

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!