LeePen | rq | 11:20 |
---|---|---|
rrq | ye | 11:26 |
LeePen | Well spotted! | 11:27 |
fsmithred | rrq, I just sent you email | 11:45 |
rrq | ta | 11:46 |
fsmithred | ok, grub is happy to say Devuan in the boot menu via /etc/default/grub as long as the bootloader is in /boot/efi/EFI/debian | 11:58 |
fsmithred | and that requires ID=debian in /etc/os-release | 12:00 |
bgstack15 | Does parazyd spend any time in these irc channels for Devuan? | 16:09 |
bgstack15 | Or should I send him an email, about the devuan-sanity systemctl inclusion? | 16:09 |
LeePen | bgstack15: he is often here | 16:12 |
LeePen | fsmithred: I am very unhappy to go back to ID=debian | 16:13 |
LeePen | Can we make grub use ID_LIKE instead? | 16:13 |
bgstack15 | would that involve forking grub? | 16:24 |
bgstack15 | (i.e., recompiling) | 16:24 |
LeePen | bgstack15: I don't know. I don't have any uefi hardware and use lilo! | 16:25 |
bgstack15 | If ID_LIKE is a mandatory part of (the apparently systemd-related) /etc/os-release, then I think we could ask upstream grub to use it. | 16:25 |
bgstack15 | oh crap, did I just think about asking somebody to go use a systemd-related thing? | 16:25 |
fsmithred | yeah, we would need to fork grub | 16:25 |
bgstack15 | I submit myself for punishment. | 16:25 |
fsmithred | other way to deal with it is to mess with /etc/default/grub | 16:26 |
fsmithred | but I think we're not allowed to do that unless we fork grub | 16:26 |
bgstack15 | Lol, I'm even using grub-legacy in a few places. | 16:26 |
fsmithred | I can do it in the live isos | 16:26 |
fsmithred | hey, refractainstaller still supports grub-legacy | 16:27 |
bgstack15 | I forget why some of my [virtual] machines use grub-legacy. | 16:27 |
bgstack15 | It was probably easier to automate with a preseed file. | 16:27 |
fsmithred | already do it in the live isos for the grub theme | 16:27 |
fsmithred | but that doesn't work in the installer isos - /etc/default/grub is not present early enough. | 16:28 |
fsmithred | (i.e. when desktop-base gets installed) | 16:28 |
bgstack15 | grepped through the source at https://salsa.debian.org/grub-team/grub-legacy.git (which I guess is grub 1) and didn't see anything about os-release at all. | 16:31 |
bgstack15 | Maybe it's one of the debian scripts that helps assemble the grub-mkconfig scriptlets that uses /etc/os-release? | 16:31 |
fsmithred | maybe os-prober | 16:32 |
bgstack15 | That already sounds easier to fork if necessary. | 16:33 |
fsmithred | os-prober: /usr/lib/os-probes/mounted/90linux-distro | 16:36 |
bgstack15 | Oh, and this discussion is about EFI. I don't boot using UEFI at all. | 16:36 |
fsmithred | looks at /etc/os-release, /etc/issue | 16:36 |
bgstack15 | I turn that **** off. | 16:36 |
fsmithred | legacy won't do efi | 16:36 |
fsmithred | it also looks at /etc/lsb-release | 16:37 |
fsmithred | which usually is not present | 16:37 |
fsmithred | but will override what lsb_release returns | 16:38 |
bgstack15 | Well, it appears to read NAME= from /etc/os-release. | 16:38 |
fsmithred | overrides GRUB_DISTRIBUTOR, too | 16:38 |
bgstack15 | But if os-release is absent, it tries a million /etc/*release alternatives, and Devuan is in there, near the end! | 16:38 |
bgstack15 | But my systems have a /etc/debian-release so it'll short-circuit early on with just "Debian" | 16:39 |
fsmithred | it reads os-release first? | 16:39 |
bgstack15 | affirmative, in that 90linux-distro | 16:40 |
bgstack15 | Not sure what this script does, or when it is called. | 16:40 |
bgstack15 | Perhaps this is what populates grub's menu entries? | 16:40 |
fsmithred | update-grub uses it to make the boot menu | 16:40 |
bgstack15 | It only pulls NAME and PRETTY_NAME though | 16:40 |
bgstack15 | nothing about ID or ID_LIKE | 16:41 |
fsmithred | if I remember correctly there is some comment about black magic in there | 16:41 |
bgstack15 | I see nothing about black magic in the 90linux-distros file. | 16:42 |
fsmithred | is somewhere in os-prober files | 16:42 |
bgstack15 | Now, that file does comment that it has a higher number for lower priority because of the perceived lower reliability of this sort of check. | 16:42 |
bgstack15 | So os-probes/mounted/x86/05efi probably comes first | 16:43 |
fsmithred | yeah, that makes sense | 16:43 |
bgstack15 | and maybe the EFI bootloader entry somehow is labeled as Debian? | 16:43 |
fsmithred | yeah, that's a requirement for grub-efi-amd64-signed | 16:43 |
bgstack15 | I'm one of those old-timey lameos who likes the way things were so I don't really want to use EFI. | 16:44 |
fsmithred | if the bootloader is not in /boot/efi/EFI/debian then it can't find grub.cfg | 16:44 |
fsmithred | it has few advantages | 16:44 |
fsmithred | probably not worth the trouble it causes | 16:45 |
mason | I recently realized that it hinders mobility - unless you use the default naming, you can't just move disks between systems. | 16:45 |
mason | So I'll probably start using the default naming in my installs. | 16:45 |
fsmithred | which names? | 16:45 |
fsmithred | oh, if you move a disk to another box, you still need to register the bootloader with the system | 16:46 |
fsmithred | nvram | 16:46 |
bgstack15 | Is that a EFI thing? | 16:47 |
fsmithred | yeah | 16:47 |
fsmithred | there's a command to do that | 16:47 |
bgstack15 | Hm. | 16:47 |
fsmithred | efibootmger -d ... -p... <path> | 16:47 |
fsmithred | oh! | 16:49 |
fsmithred | hot shit! it works. | 16:49 |
fsmithred | ID=devuan, ID_LIKE=debian, update-grub, reboot, menu says Devuan, and it boots. | 16:50 |
fsmithred | boot/efi/EFI/debian/ was the last dir to be touched | 16:50 |
fsmithred | I'm gonna delete the devuan dir and see what happens | 16:51 |
fsmithred | confirmed | 16:52 |
fsmithred | it uses the debian bootloader but says devuan in the boot menu | 16:52 |
fsmithred | I just noticed that /etc/os-release is a symlink to /usr/lib/os-release | 16:54 |
fsmithred | so that can be changed in base-files and we should be all set. | 16:54 |
fsmithred | I'm assuming it will work with secure boot. It's working with the signed grub now. | 16:54 |
bgstack15 | And what else will break?! This feels like something to stick just to unstable for now. | 16:54 |
fsmithred | not sure if it will break third-party software that expects debian | 16:55 |
fsmithred | brb | 16:55 |
fsmithred | we will need to change it for the beowulf point-release. Can't have uefi people landing on a grub prompt. | 17:02 |
LeePen | fsmithred: I am really against changing ID back to debian. | 17:08 |
fsmithred | why? | 17:08 |
LeePen | It breaks other things: postgresql-common for 1. | 17:08 |
fsmithred | nm | 17:08 |
fsmithred | don't | 17:08 |
fsmithred | ID_LIKE=debian | 17:08 |
LeePen | Yes, absolutely | 17:09 |
fsmithred | it works. I just tested it in beowulf. | 17:10 |
fsmithred | also, while we're in base-files, in chimaera, 3 needs to be 4 | 17:11 |
fsmithred | in whatever | 17:11 |
fsmithred | issue | 17:11 |
mason | fsmithred: Ah, sorry, didn't see that there was traffic for me for lack of a "mason" hilight. The default path is EFI/BOOT/BOOTX64.EFI. In general, if you don't have a valid boot entry, that'll work, and it'll also make a disk bootable from the UEFI menus. | 17:11 |
fsmithred | mason, you don't get that by default, do you? | 17:12 |
fsmithred | I think you have to do expert install and choose "install grub to removable media path" | 17:13 |
mason | fsmithred: Some systems will install there by default, potentially in addition to the expected OS-named directory. I don't install anything using an installer any more, so I can't remember who does that still. | 17:13 |
mason | fsmithred: Yeah, that sounds right. | 17:13 |
fsmithred | base-files already has the right settings in os-release in beowulf-proposed-updates branch in git and in repo | 17:34 |
fsmithred | I guess I can close the bug report I opened this morning | 17:34 |
fsmithred | and beowulf-proposed-updates are included in the iso | 17:38 |
fsmithred | so why didn't it work? | 17:39 |
fsmithred | I did another install. Having ID=devuan and ID_LIKE=debian does not work. The bootloader is in a dir named devuan, not debian. | 18:22 |
fsmithred | Those are the default settings provided by base-files in beowulf-proposed-updates (and in the installer isos build#171) | 18:24 |
bgstack15 | How does Ubuntu do it? | 18:36 |
bgstack15 | they're a Debian derivative too. | 18:36 |
fsmithred | they have their own signing key from MS | 18:41 |
bgstack15 | Oh, M$ Ubuntu | 18:41 |
bgstack15 | lol | 18:41 |
fsmithred | thye had theirs years before debian did | 18:41 |
mason | Oh, you guys are talking SecureBoot, not just UEFI boot...? | 18:50 |
LeePen | bgstack15: fsmithred: I have just been playing with teh lightdm/orca/xserver-share issue. | 18:54 |
fsmithred | what did you find? | 18:54 |
bgstack15 | Well, I did mean how does Ubuntu handle the ID= and ID_LIKE=. | 18:55 |
bgstack15 | But if they have some special stuff in SecureBoot then it's possible that gets found first by os-prober; no dang /etc/os-release munging required! | 18:55 |
fsmithred | mason, I haven't tested secure boot, but if you boot in uefi mode, you get the signed grub. | 18:55 |
LeePen | With the new beowulf-proposed-updates debian-pulseaudio-config-override package installed, orca talks for me in both lightdm and xfce session with xserver-share=false. | 18:56 |
LeePen | Can somebody else verify that? | 18:56 |
LeePen | If so we can revert xserver-share (which will fix #529) and still have orca talk. | 18:57 |
fsmithred | did you just add that package? | 18:57 |
LeePen | Yes and revert the xserver-share change in /etc/lightdm/lightdm.conf | 18:57 |
fsmithred | that wasn't changed | 18:58 |
fsmithred | it was changed in /usr/share/lightdm/lightdm.conf.d/01-debian-something | 18:58 |
LeePen | I can only see it in /etc/lightdm/lightdm.conf | 19:00 |
fsmithred | that should be a commented line | 19:00 |
fsmithred | unless that's where the new package does it | 19:00 |
fsmithred | I can test it soon | 19:01 |
LeePen | Thanks | 19:01 |
fsmithred | still waiting for ubuntu to boot | 19:01 |
fsmithred | ID=ubuntu, ID_LIKE=debian | 19:03 |
LeePen | fsmithred: maybe that was my editing or /etc/lightdm/lightdm.conf. | 19:09 |
LeePen | Does lightdm even read /usr/share/lightdm/lightdm.conf.d? | 19:09 |
LeePen | bgstack15: ^^^? | 19:09 |
bgstack15 | let me check for you. | 19:10 |
fsmithred | bgstack15, your change was from true to false? | 19:11 |
bgstack15 | in b-w-u for lightdm, the change hard-codes "xserver-share=false" to underneath /usr/share/lightdm/lightdm.conf.d/01_debian.conf | 19:12 |
bgstack15 | Before that, there is no explicit configured xserver-share setting, so lightdm would use the default. | 19:12 |
fsmithred | ok, I changed it to true and rebooted, but still get voice | 19:12 |
fsmithred | I'll try in /etc | 19:13 |
fsmithred | ok, I'm confused. What do I need to do to return it to the default that was before your changes to lightdm? | 19:14 |
bgstack15 | remove the xserver-share line from /usr/share/lightdm/lightdm.conf.d/01_debian.conf | 19:14 |
bgstack15 | and reboot or restart lightdm (beware of the implications of that, of course) | 19:15 |
fsmithred | ok | 19:15 |
bgstack15 | I am assuming the default would be whatever is described in the comments of the main /etc/lightdm/lightdm.conf | 19:16 |
fsmithred | I had to downgrade lightdm to break the screen reader | 19:21 |
fsmithred | nothing at lightdm login screen (after installing the override package) | 19:24 |
fsmithred | I need lunch before I mess with this more | 19:25 |
LeePen | Hmmm. It is working for me both in lightdm and the subsequent xfce session. | 19:25 |
LeePen | Tea time here. | 19:25 |
fsmithred | difference is that I downgraded lightdm | 19:26 |
fsmithred | could not get the screen reader to stop by changing the config | 19:26 |
fsmithred | ok, nothing is working the way it should | 20:45 |
fsmithred | purging a package does not remove its files | 20:48 |
fsmithred | runnning with autospawn=no uncommented does not break sound | 20:49 |
fsmithred | removing xserver-share=false from /usr/share/lightdm/lightdm.conf.d/01-debian-whatever does not break the screen reader | 20:50 |
fsmithred | but downgrading to the version in beowulf main does break the screen reader. | 20:50 |
fsmithred | what else is different between the two packages? | 20:51 |
bgstack15 | let me run that diff | 20:55 |
bgstack15 | which exact version is in beowulf? | 20:56 |
bgstack15 | Assuming beowulf is using the Debian release from git tag debian/1.26.0-4, then the 1.26.0-4+devuan1 changelog includes | 20:57 |
bgstack15 | add session optional pam_elogind.so to lightdm-greeter (Closes: #438) | 20:57 |
fsmithred | ah | 20:58 |
fsmithred | yeah, that does it | 21:05 |
fsmithred | need pam-elogind.so and debian-pulse-ovveride. Don't need xserver-share=false | 21:05 |
fsmithred | bbl | 21:06 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!