=== pizzaiolo is now known as pizza | ||
bryceh | https://autopkgtest.ubuntu.com/packages/n/node-caniuse-db/jammy/i386 --- what does "code 14" mean? | 01:56 |
---|---|---|
sarnold | darktable in #ubuntu had a hirsute -> impish upgrade blow up, and usrmerge seems to have been involved somehow, https://pastebin.com/AnVB5zJv -- while trying to walk him through piecing his system back together, some of us found we've got /bin -> /usr/bin symlinks, and some of us don't have such symlinks.. what's the long-term plan for harmonizing everyone's setups? | 02:19 |
sarnold | (sigh, darktrick, not darktable) | 02:19 |
DarkTrick | I added a persistent question here: https://askubuntu.com/questions/1379811/usrmerge-of-bin-sbin-lib-how-is-the-system-supposed-to-be | 02:45 |
DarkTrick | I hope I got the details right. I hope for a quick answer to straighten out my system :) | 02:46 |
mwhudson | sarnold: using do-release-upgrade vs editing /apt/sources.list directly? | 03:08 |
sarnold | mwhudson: do-release-upgrade, heh, that was my first question :) | 03:09 |
sarnold | oh hah DarkTrick's in here, too :) I guess I'm not super-observant | 03:09 |
mwhudson | although usrmerge is recommended by ubuntu-minimal now so you'd have to try moderately hard not to have it? | 03:10 |
sarnold | mwhudson: hmnm, maybe; I certainly went through an awkard enough install process myself that I'm not surprised I don't have the usrmerge package installed, but I *do* have the /bin -> /usr/bin etc symlinks. (I don't understand it.) | 03:12 |
mwhudson | sarnold: well either your system came that way or you had it and uninstalled it i guess? | 03:12 |
sarnold | mwhudson: bashing-om probably didn't go to much effort to avoid usrmerge, but he doesn't have the symlinks; he's probably got an old install upgraded a few times though | 03:12 |
sarnold | mwhudson: my system is the ubuntu root on zfs with luks guide, who knows what exactly that is :) that's why I'm so surprised I've got the symlinks.. | 03:13 |
mwhudson | sarnold: which version did you start with? | 03:13 |
sarnold | mwhudson: oh that's an excellent question. do you know off-hand where apport gets that from? :) | 03:14 |
mwhudson | sarnold: /var/lib/something/media-info | 03:15 |
mwhudson | or something | 03:15 |
mwhudson | ah v | 03:15 |
mwhudson | /var/log/installer/media-info | 03:15 |
mwhudson | of course i can't remember when we started making merged images so | 03:16 |
sarnold | mwhudson: thanks, not on this system :( bummer. | 03:17 |
sarnold | mwhudson: my memory is something like 19.10 but .. it's a very fuzzy memory. | 03:17 |
mwhudson | sarnold: debian enabled merged-usr by default in jun 2018 it seems, not sure about ubuntu | 03:19 |
DarkTrick | Just for reference, I started with 18, I think | 03:19 |
mwhudson | (default in debootstrap) | 03:19 |
=== genii is now known as genii-core | ||
DarkTrick | mwhudson, that means I should've had a merged dir from the beginning, but I don't .... | 03:20 |
mwhudson | DarkTrick: 18.what? | 03:20 |
sarnold | mwhudson: oh! that'd do it.. my install was debootstrap :) | 03:20 |
mwhudson | 18.04 wouldn't | 03:20 |
mwhudson | 18.10 might have? | 03:20 |
DarkTrick | mwhudson, the LTS version | 03:20 |
mwhudson | yeah that's 18.04 | 03:20 |
* DarkTrick was sooo close... | 03:21 | |
mwhudson | ah i think disco / 19.04 + were merged by default | 03:21 |
mwhudson | maybe? | 03:22 |
mwhudson | somewhere around then | 03:22 |
DarkTrick | mwhudson, so I understand it's rather recommended to have usrmerge (?) and if not I should get it | 03:25 |
DarkTrick | what about symlinks for individual files VS symlinks of the whole directory, like sarnold has? | 03:26 |
mwhudson | DarkTrick: yes i think for impish+ the idea is that unmerged is no longer supported, not sure though | 03:26 |
mwhudson | DarkTrick: uh no, symlinks from /bin to /usr/bin etc is the expectation | 03:26 |
mwhudson | there are long and tedious arguments on debian-devel about this | 03:26 |
DarkTrick | mwhudson, so, the whole directory, right? | 03:27 |
mwhudson | yea | 03:27 |
DarkTrick | is this what would be done if I'd call `usrmerge` ? | 03:27 |
DarkTrick | (the program) | 03:27 |
mwhudson | well it | 03:27 |
mwhudson | well it's all done by installing the package | 03:28 |
mwhudson | it does all the stuff in it's maintainer scripts i think | 03:28 |
DarkTrick | mwhudson, I have the package, but it's not merged | 03:28 |
mwhudson | if your system is already wedged i don't know if it's going to work though | 03:28 |
mwhudson | DarkTrick: uhh | 03:28 |
* mwhudson checks some confidently made assertions | 03:28 | |
DarkTrick | worst case would be to manually move all non-links to /usr/* and create a symlink to the folder? | 03:29 |
mwhudson | DarkTrick: congratulations, i guess? | 03:29 |
mwhudson | DarkTrick: /usr/lib/usrmerge/convert-usrmerge is the script that does the deed | 03:30 |
DarkTrick | so either I break my system now or I fix it :D | 03:30 |
mwhudson | i'm not sure i'd be brave enough to do this by hand | 03:30 |
DarkTrick | who needs a casino, I have Ubuntu xD | 03:31 |
mwhudson | there are going to be moments during the process where the system is in a very strange state i assume | 03:31 |
mwhudson | like /lib64/ld-linux-x86-64.so.2 being a dangling symlink | 03:31 |
sarnold | maybe start a root sash or busybox first? enable all tools? something like that.. | 03:32 |
mwhudson | reboot with break=bottom, run chroot /root /usr/lib/usrmerge/convert-usrmerge ? | 03:33 |
mwhudson | anyway good luck! | 03:33 |
sarnold | mwhudson: wat | 03:35 |
mwhudson | sarnold: doing it all from the initrd | 03:35 |
sarnold | mwhudson: google's giving me nothing but swimwear photos | 03:35 |
sarnold | https://sources.debian.org/src/debirf/0.38/src/debirf/?hl=219#L219 | 03:36 |
sarnold | what the heck is debirf? :) | 03:36 |
sarnold | here we go, this is more like it https://sources.debian.org/src/initramfs-tools/0.140/init/?hl=254#L159 | 03:39 |
mwhudson | sarnold: "build a kernel and initrd to run Debian from RAM" | 03:40 |
mwhudson | not heard of that one | 03:40 |
sarnold | mwhudson: ohhh hmm that might be useful, I've wondered before if my power9 system would be more reliable if I ran it from tmpfs or ramdisk.. (*sniff*) | 03:41 |
juliank | File "/snap/git-ubuntu/660/usr/lib/python3/dist-packages/gitubuntu/prepare_upload.py", line 112, in push | 10:11 |
juliank | assert gitubuntu.importer.VCS_GIT_URL_VALIDATION.fullmatch( | 10:11 |
juliank | hooray? | 10:11 |
juliank | The reason for this is I had insteadOf configured for launchpad so it always uses ssh to talk to it | 10:36 |
=== cpaelzer_ is now known as cpaelzer | ||
=== sem2peie- is now known as sem2peie | ||
=== alan_g_ is now known as alan_g | ||
rbasak | juliank: ah - so git-ubuntu is unable to determine what URL to put in the changes file for you that doesn't use insteadOf. | 12:07 |
juliank | rbasak: basically same cause as https://bugs.launchpad.net/usd-importer/+bug/1942985 | 12:08 |
ubottu | Launchpad bug 1942985 in usd-importer "push-for-upload verification breaks on non-default but valid remotes" [Medium, Triaged] | 12:08 |
rbasak | I'm not sure if git-ubuntu should expand insteadOf or not. It'd be nice if it did, but I'm not keen on chasing that functionality. It'd be nice if git could expand it for us. | 12:09 |
juliank | rbasak: well, a fix for that will fix mine too; though maybe insteadOf is easier to fix itself by an option somewhere to not follow insteadOf | 12:09 |
rbasak | No, I think it's a different bug. | 12:09 |
juliank | rbasak: No, the issue is that insteadOf expands the normal https:// url that git-ubuntu configured to ssh:// | 12:09 |
juliank | s/expands/replaces/ | 12:09 |
rbasak | Oh, OK. | 12:09 |
rbasak | So there's another failure case here too. | 12:09 |
juliank | And pygit2 follows insteadOf apparently when you calculate the uri | 12:10 |
rbasak | If someone uses insteadOf to create a virtual alias such as lp:, then that won't work either. | 12:10 |
juliank | right, I have | 12:10 |
juliank | [url "git+ssh://juliank@git.launchpad.net/"] | 12:10 |
juliank | insteadOf = lp: | 12:10 |
juliank | insteadOf = https://git.launchpad.net/ | 12:10 |
juliank | so pygit2 returned git+ssh:// instead of whatever git-ubuntu actually configured in .git/config | 12:11 |
rbasak | "git-ubuntu clone" configures https://git.launchpad.net/... | 12:12 |
rbasak | (for fetch; git+ssh for push) | 12:12 |
juliank | yes | 12:12 |
juliank | didn't want to write it :) | 12:12 |
rbasak | I agree then that Christian's bug fixed will also fix your issue. | 12:12 |
rbasak | Sort of by accident though. I think git-ubuntu should indeed be using the actual URL used. But this one can be special cased. | 12:13 |
=== sem2peie- is now known as sem2peie | ||
DarkTrick | mwhudson, sarnold: Just back from work. I thought I'd follow up on my problem: usrmerge did only help half way: if two files exist it will report that one, single file, you have to solve it by hand, run again and wait for the next error to occur and handle. I ended up changing the usrmerge script to rename files instead of throw errors. So far everything seems ok. | 12:56 |
mapreri | ddstreet, teward: I wrote up a core-dev application. I'd be happy if you could write something about me ^^ https://wiki.ubuntu.com/MattiaRizzolo/CoreDevApplication | 14:14 |
=== genii-core is now known as genii | ||
=== sem2peie- is now known as sem2peie | ||
sarnold | DarkTrick: oh! that's a clever fix to make it run reasonable speed :) | 19:42 |
Unit193 | mapreri: G'luck! | 19:43 |
=== sarnold_ is now known as sarnold | ||
Unit193 | ahasenack: Looks like it's all merged, so I guess that's good then. | 20:58 |
ahasenack | I proposed a PR in salsa too, but I have to update it with what I have actually uploaded | 20:59 |
ahasenack | there was some review feedback on the ubuntu side that made me change it a bit | 20:59 |
Unit193 | Yeah, saw that bit about tests failing not actually failing the build and all. | 21:18 |
=== sem2peie- is now known as sem2peie | ||
=== nibbon_0 is now known as nibbon_ |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!