zyga-mbp | good morning | 05:56 |
---|---|---|
pstolowski | morning | 07:01 |
mvo | good morning pstolowski | 07:14 |
pstolowski | mvo: hi! could you please merge https://github.com/snapcore/snapd/pull/10447 ? it only failed on the lxd test that cachio fixed with the other PR | 07:15 |
mvo | pstolowski: sure | 07:15 |
mborzecki | meh, damn matrix bridge | 07:16 |
mborzecki | mvo: pstolowski: hey | 07:16 |
mborzecki | i've been sitting there since 630, wondering why nobody has showed up | 07:17 |
zyga-mbp | mborzecki hahaha | 07:19 |
zyga-mbp | I've been here :) | 07:19 |
zyga-mbp | well, since around 8 | 07:19 |
zyga-mbp | I've filed https://git.ostc-eu.org/OSTC/tools/oh-spread/-/issues/8 | 07:19 |
mvo | mborzecki: good morning to you as well :) | 07:20 |
mvo | mborzecki: hahaha | 07:20 |
mvo | zyga-mbp: ! good morning | 07:21 |
mborzecki | mvo: yeah, heh | 07:21 |
zyga-mbp | and also https://git.ostc-eu.org/OSTC/tools/oh-spread/-/issues/9 | 07:21 |
zyga-mbp | if anyone from the snapd team is interested I would love to include you in code review | 07:21 |
zyga-mbp | if not that's fine, I'll just push on | 07:21 |
mborzecki | mvo: mardy is around 730 usually, then you join ~830 and pawel joins around 9, so something didn't add up ;) | 07:21 |
mborzecki | heh, so some people in #irc:matrix.org have noticed equally confusing behavior | 07:23 |
mborzecki | mvo: can you take a look at https://github.com/snapcore/snapd/pull/10432 ? | 07:27 |
mborzecki | heh, so i'm confused, libera.chat are running their own bridge to matrix, but it seems like it's out of sync :/ | 07:31 |
zyga-mbp | mborzecki matrix is out of sync! | 07:32 |
zyga-mbp | mborzecki the agents are coming ;) | 07:32 |
mborzecki | haha, took the wrong pill this morning ;) | 07:32 |
mborzecki | waiting for agent smith to knock on my door | 07:32 |
mborzecki | ok, going to school, back in 1.5h or os | 07:38 |
mborzecki | so | 07:38 |
pedronis | mvo: hi, we are getting more? random core 20 failures on https://github.com/snapcore/snapd/pull/10446, I don't think it's the new code (because it deterministic) but it's annoying | 07:40 |
mvo | mborzecki I will look at 10432 sure! | 07:41 |
mvo | pedronis: let me check | 07:41 |
pstolowski | mvo: can you land that one too? https://github.com/snapcore/snapd/pull/10428 (also, only lxd test failure fixed by cachio) | 07:47 |
mvo | pstolowski: sure | 07:49 |
pstolowski | mvo: thank you! | 07:50 |
mvo | 10448 needs a second review (should be simple) | 08:14 |
pstolowski | pedronis: i've updated error handling with ignore flag in https://github.com/snapcore/snapd/pull/10408, you should probably look at the last commit | 08:36 |
pedronis | pstolowski: I'll look in a bit | 08:47 |
pstolowski | ty | 08:49 |
pedronis | mvo: question in https://github.com/snapcore/snapd/pull/10449/ ? | 09:14 |
mvo | pedronis_: looking | 09:26 |
mvo | pedronis_: I can remove the random things in the end, I misunderstood the initial suggestion. will fix it, sorry | 09:27 |
mborzecki | re | 09:41 |
mborzecki | hm still doesn't work | 09:44 |
=== pedronis_ is now known as pedronis | ||
dot-tobias | hi everyone | 10:34 |
rbasak | Is this a core snap regression? https://bugs.launchpad.net/snapd/+bug/1933392 | 11:05 |
rbasak | I've not tested, but I'm not sure if that report is in the right place | 11:05 |
rbasak | It's apparently regressing the certbot snap on Debian. | 11:05 |
dot-tobias | Building a fresh Core image with ubuntu-image, I currently get this fatal error: invalid layout of volume <myvolume>: cannot resolve content for structure #0 at index 0: cannot find "dtbs" in kernel info from "/tmp/tmplm3ypa5a/unpack/kernel" | 11:09 |
ogra | rbasak, you could never "run snapctl on the host" ... that only works internally | 11:09 |
dot-tobias | Building an image worked yesterday around 3pm UTC+1, but I saw that the pi-kernel snaps from the 18-pi channel have been updated yesterday. Found https://forum.snapcraft.io/t/failed-to-build-pi-uc20-image/24918 but using edge snaps isn't really an option for prod images. | 11:11 |
ogra | not sure about the REST API stuff, but the call to snapctl is definitely nt supposed to work unles you run it from "snap run --shell ..." from inbside some snap | 11:11 |
ogra | geez ... my typing ... must be friday ... | 11:11 |
pstolowski | jamesh: hi, there is a todo: check g_icon_serialize in your notification prototype; is it optional? is gtk notification api able to deal both with icon names & serialized icon? | 11:14 |
ogra | dot-tobias, https://forum.snapcraft.io/t/changes-in-pi-pi-kernel-dtb-handling/24369 update your gadget | 11:15 |
rbasak | ogra: can you glance at the github issue linked please? In that case, it _is_ running from inside the snap. The reporter tried to generalise it in the LP report. | 11:15 |
ogra | aha | 11:15 |
dot-tobias | ogra: I've cherrypicked https://github.com/snapcore/pi-gadget/commit/e452ccb3c6608e6f1de85e51e161706e780f01b1 and https://github.com/snapcore/pi-gadget/commit/fd40afabf0f5bcfcebb673a16a4046a0aa4234e6 to my gadget, that resolved the original error (can't remember the wording) and produced a working image. The same gadget, which includes said changes and worked yesterday, now suddenly fails … I'll triple-check my gadget to be sure | 11:20 |
mborzecki | dot-tobias: it's a new thing, pi-kernel must have finally caught up https://bugs.launchpad.net/ubuntu/+source/snapd/+bug/1907056 | 11:35 |
mborzecki | tbh the feature was made specifically for pi | 11:35 |
pstolowski | cachio_: do you have a system where snap debug state hangs? | 11:53 |
cachio_ | pstolowski, let me reserve that | 11:54 |
pstolowski | cachio_: i actually just need state.json from such system, if you can download one that would be enough | 11:55 |
cachio_ | pstolowski, I already reserved 1 device | 11:55 |
pstolowski | ok thanks | 11:55 |
cachio_ | it is being provisioned | 11:55 |
dot-tobias | mborzecki: Thanks, I reverted the changes to my gadget, image is building again. Looking forward to have DTBS from the kernel snap, though 😊 | 12:14 |
ogra | yeah, it is definitely the best thing since sliced bread ! | 12:17 |
ogra | rbasak, (sorry, got dragged into a meeting) if you suspect a core regression the first thing i'd do is a snap revert core and compare ... ad given the curl command in the bug i'd also check if snap login makes a difference | 12:19 |
ogra | i doubt the core changes are that big nowadays given 16.04 is ESM ... so such a regression is not super likely (possible still indeed) | 12:19 |
rbasak | ogra: I don't have a reproducer here readily available unfortunately - I was just asked because I wrote the snap originally. | 12:22 |
rbasak | So I don't have anything to revert. | 12:22 |
rbasak | According to https://github.com/certbot/certbot/issues/8922#issuecomment-868451477 microk8s is affected too | 12:22 |
rbasak | I suspect maybe a regression in snapd rather than core? Anyway I'm just passing on the message. | 12:23 |
ogra | yeah, perhaps ... | 12:23 |
ogra | in either case soeone should ask in the issue to revert core to the former version to see if it is really the new core | 12:23 |
ogra | its the easiest bit to verify this | 12:24 |
rbasak | If it reproduces as reported then that's a bug, no? | 12:24 |
rbasak | I'd like for upstream to do the debugging :-/ | 12:24 |
ogra | it is | 12:24 |
ogra | ... but it helps a lot if the reporter can nail down the two revision of core so you know where to lokk | 12:25 |
ogra | *look | 12:25 |
rbasak | It's not possible to do a bisection in the general case, because previous snap binaries are not made available except to the snap developer. | 12:26 |
rbasak | That decision has a cost: you can't reasonably expect affected users to do that kind of debugging. | 12:27 |
ogra | rbasak, the last local revision is still there ... indeed that doesnt help when you install snapd/core from scratch | 12:58 |
ogra | so affected users shuld be always able to revert ... just *new* users wont be (as the only have the current core version they just freshly installed) | 12:59 |
rbasak | ogra: well, OK, but affected users can't delegate - eg. I can't help. | 13:01 |
rbasak | Or, at least, it's artificially made difficult for me to help | 13:02 |
ogra | yes | 13:02 |
pedronis | rbasak: ogra: fwiw, we have asked somebody on the team to investigate this | 13:48 |
ogra | 👍 | 13:48 |
rbasak | Thanks! | 13:49 |
mborzecki | pstolowski: can you take a look at https://github.com/snapcore/snapd/pull/10460 ? sergiusens reported this in the forum https://forum.snapcraft.io/t/snap-download-slips-in-extra-logging-with-download-size/25151 | 14:11 |
pstolowski | looking | 14:12 |
mborzecki | cachio_: can you try to update the tumbleweed image? | 14:20 |
cachio_ | mborzecki, sure | 14:20 |
mborzecki | cachio_: thanks, hoepfully the image will be better this time ;) the builds in obs keep on working | 14:21 |
cachio_ | mborzecki, nice | 14:21 |
cachio_ | I'll create a new one right now | 14:22 |
pstolowski | pedronis: the debian issue seems to be caused by /run/snapd-snap.socket vs /run/snapd.socket rename; debian 9 has a very old snapd 2.21 deb, the old snapctl from the deb on the host can't talk to new snapd | 15:35 |
pedronis | pstolowski: that's not a rename, or I'm cofunsed we have two sockets | 15:36 |
pedronis | pstolowski: one is for snapctl and one is for the general api | 15:37 |
pedronis | pstolowski: /run/snapd.socket is the general API socket (use by the snap command for example | 15:37 |
pstolowski | pedronis: yeah, sorry, no rename, we have two indeed. what i see with strace is the old snapctl is trying to talk to /run/snapd.socket | 15:37 |
pstolowski | did we have just one socket in the past? | 15:38 |
pedronis | not sure, we need to go dig into the release branches I suppose | 15:38 |
pstolowski | good idea | 15:40 |
pedronis | pstolowski: it was already like that: https://github.com/snapcore/snapd/blob/release/2.29/cmd/snapctl/main.go#L38 | 15:41 |
pedronis | so maybe there's no socket or something is wrong with apparmor? | 15:41 |
pedronis | deos the socket exist? | 15:41 |
pedronis | I mean which sockets are in /run | 15:41 |
pstolowski | pedronis: that's 2.29 though, we should be looking at 2.21 | 15:42 |
pedronis | ah, sorry I misread | 15:42 |
pedronis | seems we didn't have two sockets then | 15:43 |
pedronis | but I'm still confused why is the snapctl new though? | 15:43 |
pstolowski | pedronis: we have two sockets because snapd 2.51 comes from core. but afaiu the code of snapctl from 2.21 (which comes from the deb) expects to talk to /run/snapd.socket | 15:46 |
pedronis | pstolowski: sorry, you mean snapctl comes from core, but snapd comes from the 2.21 deb? | 15:47 |
pstolowski | pedronis: no. snapd comes from core, snap --versin shows 2.51. but for snapctl we don't reexec, so the one from deb 2.21 is used | 15:49 |
pedronis | because this is a classic snap? | 15:49 |
pstolowski | pedronis: i only tried snapctl manually so far, let me try a classic snap | 15:50 |
pedronis | I suppose you want to install a confined snap and a classic one, and try snapctl from their snap run --shell | 15:51 |
pstolowski | pedronis: i tried confined snap it was fine, it used snapctl from core | 15:52 |
pstolowski | pedronis: yep, classic snap uses snapctl from the host, so fails | 15:54 |
pedronis | this is not a regression though, I expect things behaves like this since a long time | 15:54 |
pedronis | I wonder what changes, and why a classic snap is using snapctl at all? | 15:54 |
pedronis | what changed | 15:55 |
pedronis | pstolowski: I suppose we need to try to snap mentioned in the bug? I don't know why it is using snapctl | 15:55 |
pedronis | rbasak: did certbot always use snapctl? this doesn't seem a regression | 15:56 |
rbasak | pedronis: it has done for a long time | 15:58 |
rbasak | Let me find the thread | 15:58 |
pedronis | anyway I'm not quite sure how it ever worked on debian 9 | 15:58 |
pstolowski | pedronis: yes this seems to be very old | 15:59 |
rbasak | pedronis: see https://forum.snapcraft.io/t/certbot-request-for-classic-snap-approval/6240/19?u=rbasak and the following post | 16:01 |
pstolowski | pedronis: it uses snapctl unset from prepare-plug-plugin, i think that's all | 16:01 |
rbasak | That's why it's using snapctl - it was a condition of classic snap approval to check for that | 16:01 |
pedronis | anyway as I said unsure it ever worked on debian 9 given the version of snapd it shipped with. We could think whether there's a way to fix this, but is not new afaict | 16:02 |
rbasak | pedronis: I might have jumped the gun in assuming it was a regression. Now that I look at the report, I don't see the reporter actually saying that. | 16:02 |
rbasak | Apologies for that | 16:03 |
rbasak | I suppose it's possible that nobody has tried using the certbot snap with a plugin on Debian 9 before? | 16:04 |
pstolowski | one sec | 16:04 |
rbasak | Given how well used it is, that does seem odd to me still though. | 16:04 |
rbasak | But also, installing core seems to regress it? The certbot snap doesn't need core - IIRC it's based on core20. | 16:05 |
pedronis | it might also depends what other snaps are installed | 16:05 |
pedronis | though given that old snapd it might always install core as well | 16:06 |
pedronis | not sure | 16:06 |
pstolowski | the bug report isn't complete afaict | 16:07 |
pstolowski | the github issue it links to is complete | 16:07 |
pstolowski | snap set certbot trust-plugin-with-root=ok followed by snap install certbot-dns-cloudflare | 16:07 |
pstolowski | triggers it | 16:07 |
pstolowski | because of prepare-plug hook uses snapctl | 16:08 |
pstolowski | anyway.. it's pretty clear the old snapctl from the deb can't talk to new snapd and must be like that for a few years already. i can also reproduce it with snap run --shell certbot | 16:09 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!