libera/#devuan-dev/ Tuesday, 2020-05-12

LeePeng4570n: Nice to know somebody uses it. Our live-build packages are very out of date wrt Debian. Would he like to update them?10:54
LeePenAnyway, after a hiccup last night, the new debian-installer builds are in the repository.10:55
bgstack15I'm going to take a brief poke at bug #458 about accountsservice. If it's as easy as compiling with a -DNOSYSTEMD then maybe it can happen by this user.13:07
masonbgstack15: I haven't looked at the source yet. Do they ship init scripts? Nothing I could see suggested it was heavily tied into systemd.15:07
LeePenmason: does it not work?15:07
LeePenAs far as I know it is started by dbus activation?15:08
masonLeePen: Oh, I hadn't dug in wildly far. I assumed it started as a normal service and talked to things via dbus, but I could be confused.15:09
masonI'd read something talking about looking for the running service to determine if it was active, which suggested that it'd always be running if in use.15:09
LeePenCertainly on my beowulf vm it is started after you login15:10
masonOh, hrm. I was mucking with my kid's system, I didn't notice it running, and then I saw that there were unit files and no init script. Am I more confused than I thought?15:11
bgstack15the systemd unit is type=dbus15:11
bgstack15I've already drafted an init script.15:11
masonHow's anything launch it with dbus if it isn't there, running, registered with dbus?15:11
bgstack15It has a dependency on dbus.15:12
bgstack15And start-stop-daemon runs it just fine.15:12
bgstack15Unfortunately though I've never had any luck getting lightdm to show me user pictures. It probably has to do with my freeipa domain users, and I don't enumerate users in lightdm anyway.15:13
bgstack15i have no idea what else accountsservice is supposed to do.15:13
masonbgstack15: I found a random bug talking about its use in this role. Half a sec.15:13
mason I think15:14
masonLeePen: I'm curious how it starts on your Beowulf VM without something explicitly starting it. I may not be understanding how dbus-activated services work, but it's just random bits on disk unless something knows to start it... Or does dbus have some sort of directory with pointers to services that can be run or something? I wouldn't have looked for that.15:16
LeePenYes. Let me find it.15:17
masonAh, so there's no need for anything to run on boot? I'm curious why none of the LightDM greeters were doing per-user background screens. Maybe the bug is somewhere else. My daughter's computer uses a fairly vanilla XFCE4.15:24
masonPer the slick-greeter bug I linked, I wasn't seeing the user file populated with a background.15:24
mason(That said, dbus is as much an abomination as systemd.)15:25
bgstack15Hm, does all this mean that the init script is not required?15:25
LeePenAbsolutely. I dislike dbus activation. In fact I have disabled it in the elogind packaging in favour of an init script.15:25
LeePenbgstack15: I think so at the moment.15:26
masonbgstack15: Yeah, I think that's the gist of it, although LeePen will be the authoritative voice.15:26
bgstack15Unfortunately I cannot tell that lightdm is doing anything at all with accountsservice, in the logs.15:26
bgstack15But my accountsservice is running (from my init script) with some of the basic user data in /var/lib/AccountsService15:26
masonbgstack15: Does anything populate a background image spec there?15:27
bgstack15but then I have no reason to think my user-specified background would be defined at all15:27
bgstack15is that set in mugshot?15:27
bgstack15I use fluxbox primarily...15:27
masonNot sure. Openbox here, but I use XFCE4 for the kids since it gives them a bar out of the box that they can click on - they're not up to dmenu yet, since only two of them are old enough to read.15:28
bgstack15A while back I did learn about the supposed ~/.face file that does appear to get propagated into /var/lib/AccountsService15:29
bgstack15but seems to just die there without ever being displayed.15:29
masonPart of my desire for slick-greeter was to have the user list visible with background images as a visual cue, so my daughter sees her name and get familiar with it. We have a short password for her as her login task.15:29
masons/get fam/gets fam/15:29
bgstack15I'm not personally attached to affecting the login screen based on what username you type, and at present I require the user to type his name in. No Win-XP style user selector.15:30
masonYeah, agreed. For me, I just use XDM.15:30
bgstack15but then I'm also using freeipa users so... not sure if user enumeration would be feasible/possible/whatever.15:30
bgstack15I've noticed that fprintd does get triggered nicely upon typing my username and tabbing out of the username box.15:31
bgstack15So it would make sense that it should be able to populate the little generic user icon with my preferred display picture.15:31
bgstack15But it's not a high priority for me.15:31
masonI need to get an account on our git. I haven't needed one for the stuff I've been doing so far, but I'll want one eventually.15:37
LeePenmason: Based on a quick search, per-user backgrounds isn't supported by lightdm.15:47
bgstack15all of lightdm, or dependent on greeter?15:48
LeePenYou can have a user list though:
masonLeePen: That's unusual. Ubuntu was using LightDM until 16.04 and that's how it got per-user backgrounds, via slick-greeter. I could have sworn I saw the right config hooks as well, independent of the show/hide user list knob.15:50
masonLeePen: But yeah, the user list knob works, both for the lightdm-gtk-greeter and slick-greeter.15:50
bgstack15from lightdm-gtk-greeter.conf:15:51
bgstack15#  user-background = false|true ("true" by default)  Display user background (if available)15:51
masonbgstack15: My assumption was that "if available" meant "if AccountsService is running" hence my bug, but clearly I don't have an accurate picture in my head.15:52
bgstack15It would make sense that lightdm (-gtk-greeter) would want to use gsettings... oh boy.
bgstack15But this is a very fascinating read:
masonbgstack15: Back, reading.16:05
masonbgstack15: Interesting script in that second link.16:09
bgstack15But, from the accountsservice d/changelog: accountsservice (0.6.15-3) unstable; urgency=low16:09
bgstack15  * Move GIR package into the new section "introspection".  * debian/patches/0011-add-background-file-support.patch     - Add support for a BackgroundFile property, similar to IconFile.       Apply on Ubuntu only. (LP: #844081)16:09
bgstack15I cannot find anything in the code from the debian/master branch about backgroundfile property.16:10
masonbgstack15: Ah, that sort of dovetails with something saying ...something... was no longer Ubuntu-specific in that slick-greeter bug. Re-reading.16:10
masonI wonder if this exists in Debian's LightDM:
bgstack15well, not anything to do with 'backgroundfile' but if there's some sort of extensibility features in AccountsService that might suffice.16:20
bgstack15Oof: another thing trying to take over (albeit just another fd.o thing anyways):
bgstack15Changes in 0.6.35: * add extension interface16:25
HumanGeekhello one off you know if there is a standard/good practice to run content of an init.d script as a specify user ?16:38
bgstack15yes, there would be. I usually poke around in the extant examples in /etc/init.d until I find a service script that runs its daemon as a particular user.16:39
bgstack15Are you running your daemon with start-stop-daemon? That takes a --user $USERNAME parameter.16:40
HumanGeekbgstack15, first thx, i use sysd2v from to migrate the gitlab-sidekyq systemd unit to init script and it s my first time so i don t really know i use start stop16:43
bgstack15I was actually just using that two hours ago as discussed before you logged in! I found its output lacking but then I learned it wasn't really required and due to the systemd unit being type=dbus, it probably wasn't going to work anyway.16:45
bgstack15What is the type of the unit you're converting?16:45
HumanGeek is my current work16:46
HumanGeekand the original unit is16:46
masonSo, I'm curious - is there a Rosetta Stone for package building? 'apt source' seems to snag sources and do the equivalent of 'rpm -bp' to them, applying packaged patches and so forth.16:47
bgstack15I don't know of any, mason. I had to learn debuild skills the hard way, which is not terribly related to rpmbuild.16:47
bgstack15if you want a personal guide to the Wrong Way to build Dpkgs (TM), pm me16:48
masonI'm wondering if there's a complete equivalence table anyone knows of, as while I muck with RPM packages and in the distant past Solaris/sysv packages, I'm far less familiar with Debian packaging gizzards.16:48
HumanGeek the unit16:48
bgstack15If you want the crazy long reproducible builds process, the Devuan doc isn't bad.
masonFrom what I've heard from folks there are potentially many wrong ways and there's argument about if there is a right way. :P16:49
bgstack15Well, amusingly yes, in the Debian+ world, there's about 5 ways to build a dpkg, and more ways than that to build an apt repo.16:49
HumanGeekthe right way is alway better16:49
masonAh, good idea. Since Devuan is the target, whatever Devuan uses as stylistic choices would be correct for me. Thanks.16:49
bgstack15Rpmbuild was so straightforward, and so was createrepo.16:49
masonMm, I'm very fond of rpmbuild, and mock isn't bad.16:49
bgstack15Never figured out mock.16:49
bgstack15I think rpm recently added "triggers" concept16:50
masonIt's mostly just scaffolding that sets up clean chroots in which to build. Kind of like Poudriere, if you're familiar with that.16:50
bgstack15I wasn't smart enough to figure out Poudriere. I tried.16:50
masonThat's more a question of Poudriere being hairy than you as a user not being smart enough.16:51
bgstack15yeah, the git-buildpackage (gbp) or pbuilder mechanisms do the whole chroot ordeal, which is why they take forever but they make reproducible builds.16:51
bgstack15debuild is my goto for personal packages and also for alpha testing.16:51
bgstack15but, HumanGeek, haven't forgotten you.16:51
bgstack15Oh, good. Type=simple. I would really hope the sysd2v worked for you.16:52
bgstack15But just compare your new script to various extant scripts in /etc/init.d. You're going to want to have a case "$1", and either have status, start, stop, etc. call $1 and then you will want to define bash functions in the script for start() { echo foo; } or just have those steps right in the case logic.16:53
bgstack15make sure your script, near the beginning, checks for file /etc/gitlab/gitlab-debian.conf and then dot-sources it. (Based on EnvironmentFile from the unit file). Although in a unit file, EnvironmentFile value that starts with a dash is the way to indicate that the environment file is optional and should be skipped if absent.16:55
bgstack15So you might choose to require the /etc/gitlab/gitlab-debian.conf file.16:55
bgstack15[ -r /etc/gitlab/gitlab-debian.conf ] && . /etc/gitlab/gitlab-debian.conf16:56
bgstack15is how the bash guys like to write it16:56
HumanGeekbgstack15, i will try something like you says17:35
golinuxEvilham: I made some changes to the beta site over the weekend but they never made it to the live site.  Could you please have a look?21:29
golinuxI saw the changes in git but didn't check the live site. Thankfully LeePen did:

Generated by 2.17.0 by Marius Gedminas - find it at!