Snap on Arch/Liri OS

snapd
,
21

@zyga-snapd sure, I published it here. Thanks :slight_smile:

22

Having cleaned up all the build files and rerun makepkg, I got some different error messages: http://paste.ubuntu.com/25615596/.
Most interestingly, after rerunning makepkg again, they donā€™t show up again.

23

Ha, I thought that was only affecting my machine somehow. I have no idea what is going on, it feels like golang bug or kernel bug. I spent a few hours debugging that and I ended up with a dump of the data that failed to parse as YAML. Instead of the expected content I saw various corrupted data that I could not explain. No idea what the problem was.

Can you tell me which CPU you used? Did you use any virtualization systems? Can you tell me which filesystem you used?

In my experiments the problem reliably went away (as in, subsequent attempts had a chance to succeed) if I removed the source checkout and tried again. When I kept data on disk I got no success. Maybe the page cache was broken somehow?

In general I saw this only on Arch so Iā€™d say it is a kernel-related change as the same hardware was successfully running Fedora and openSUSE.

1 Like
24

Iā€™d add something to build/install snap-exec at around https://github.com/tim-sueberkrueb/snapd-2.27.6-1-lirios-experiment/blob/master/PKGBUILD#L67

I will provide patches but itā€™s way too late for that tonight.

25

@zyga-snapd wow, that sounds like a nasty bug indeed! My system:

cpu: IntelĀ® Coreā„¢ i7-7500U CPU @ 2.70GHz Ɨ 4 
kernel: 4.12.13-1-ARCH
os: Liri OS/Arch
filesystem: ext4

I didnā€™t use any virtualization system(s) nor a container.

I will provide patches but itā€™s way too late for that tonight.

Yeah, no hurry, thanks a lot!

26

My system used older 35xx Intel Core i5 (now I envy your shiny machine :slight_smile: ) so I suspect it is more of a kernel bug than a VM or hardware bug.

27

Looks like it :slight_smile: Let me know if I can somehow help to debug it.

28

Can you please try the new snapd-git package from AUR?

29

Itā€™s not AUR yet, thereā€™s just one remaining test that fails on Arch currently. Related PR fixing this is right here: cmd: pretend we're running on Ubuntu in TestExecInCoreSnapUnsetsDidReā€¦ by bboozzoo Ā· Pull Request #4235 Ā· canonical/snapd Ā· GitHub

Other than that, can you clone snapd repository and follow instructions from this post Updates to snapd package on Arch - #35 by mborzecki ?

30

Hey @zyga-snapd, @mborzecki thanks for the news/info, I will try (hopefully) this weekend :slight_smile:

31

That PR is now in. Is there anything else left here, or is it all sorted and we can drop the tracking tags?

32

Hey, I just got to try building and installing the latest snapd from git on Liri OS and it just works for strictly confined snaps now! This is awesome, thanks a lot!
I still have issues running some classic snaps, however. e.g. vscode and atom wonā€™t launch with the message libselinux.so.1: cannot open shared object file: No such file or directory.

33

Thanks for feedback. Glad to hear that things work for you now.

As for vscode snap, another user reported this problem right here Installing vscode snap on Arch Linux. I did some digging and IMO the snap does not ship with all dependencies included. I think itā€™s best if you reach out to the creator of that snap and make them aware of these problems.

1 Like