[00:40] PR snapcraft#2749 closed: storeapi: add StoreErrorList to handle store errors [05:23] morning === jamespage_ is now known as jamespage === abeato_ is now known as abeato [06:40] zyga: i'm playing a bit with repck in spread, instead of a 5.5MB archive i usually got, i'm getting 2.8MB now [06:50] mvo: morning [06:56] hey mborzecki ! good morning [06:57] PR snapd#7582 closed: spread: include mounts list in task debug output === pstolowski|afk is now known as pstolowski [07:04] morning [07:05] Good morning [07:05] I’ll start later today, just doing morning dog walk [07:05] See you in 30 [07:06] mborzecki: try that with spread doing gitignore and the matching snapd branch please [07:31] back [08:05] mvo: #7558 can land, or do want to wait for Samuele? [08:05] PR #7558: boot,dirs,image: various refinements in the prepare-image code switched to seedwriter [08:09] mborzecki: I updated https://github.com/snapcore/snapd/pull/7547/files a little [08:09] PR #7547: many: use a dedicated named cgroup hierarchy for tracking [08:21] Chipaca: o/ - FYI, I worked around the this issue with services failing on refresh, and filed this bug for your consideration: [08:21] https://bugs.launchpad.net/snapd/+bug/1847723 [08:21] Bug #1847723: Config hook is run after services start during refresh [08:22] bloodearnest: what about the different -refresh hooks? [08:22] bloodearnest: e.g. post-refresh? [08:22] bloodearnest: ISTR post-refresh comes right before start-services [08:23] Chipaca: so, my fix was to regenerate the config in post-refresh in my snap. But I think that that argubly shouldn't be nessesary? [08:23] ah i see you mention that i the bug [08:23] yeah [08:23] I'm sure there's things I'm missing, but it was a suprise to me that the config hook ran after services had been started [08:24] bloodearnest: I'll let pstolowski and/or pedronis figure out whether the current behaviour is correct or not :-) [08:24] sure, thanks for your time [08:30] bloodearnest, Chipaca yes, this (config hook running after services start) has been known for a long time (and we know it's annoying), we touched that briefly with pedronis even recently with the conclusion that we might reconsider the order if there are strong arguments [08:32] pstolowski: great. Well, consider that bug this user's vote on the isssue :) It's a dual of the install/config hook interplay really, where services can start before config hook as run there too. [08:48] bloodearnest: OTOH I can imagine snaps with the opposite problem, needing the services running for 'configure' to make sense? [08:53] brb, installing eoaiun [08:53] Chipaca: oh my :-) [08:53] I wonder what your experience will be like [08:53] zyga: slow [08:54] the only machines i have that i'm willing to sacrifice to the goddess eoan are slow [08:55] PR snapd#7586 opened: interfaces/optical-drive: additional permission needed by mir-kiosk-kodi === mborzeck1 is now known as mborzecki [09:00] PR snapd#7587 opened: spread: generate delta when using google backend [09:01] zyga: mvo: ^^ this should fix the source package size uploaded by spread for a while [09:01] nice [09:01] thank you [09:13] Chipaca: very possiblly :) [09:13] mborzecki: nice one! [09:14] does anyone know How to statistics the number of software in gnome-software [09:15] xiaoji: gnome-software is a tarball, what it shows when you build and run depends heavily on configuration [09:20] xiaoji: by "software" do you mean packages? apps? binaries? [09:20] xiaoji: and do you mean the number installed, available for install, ...? [09:21] xiaoji: and do you mean how many does gnome-software show, or do you mean how do you get gnome-software to show you the number? [09:27] thanks [09:27] yes [09:28] I want to know how many software can be installed and have been installed in gnome-software [09:28] PR snapd#7588 opened: cmd/snap: add a "snap internal portal-info" command [09:31] jamesh: that's not the first internal command we have, any reason to add the new toplevel 'internal', etc? [09:32] * zyga installs gobs of updates [09:33] xiaoji: why are you asking in this channel? [09:35] I have asked on many channel and many IM.. [09:35] sorry [09:36] xiaoji: where? [09:37] xiaoji: as zyga pointed at, gnome-software is the frontend to a number of backends, and not all those backends offer a count. Also you haven't told me what "software" is, nor on what distro you are [09:38] for example, if "software" is a package, and you are on a debian-derived distro, you can answer "how many are installed" via the backend tools [09:38] i guess the same is true for other distros [09:38] so it's not clear you know what you are asking [09:38] so, how to answer you? your question is unclear, to begin with [09:38] I am on ubuntu 18.04 [09:39] are you using flatpak? [09:39] no,only install software from software-center [09:40] i thought you could get flatpak via the software-center, but ok [09:40] xiaoji: and, what is "software"? [09:41] xiaoji: is a library-only debian package, "software"? [09:41] xiaoji: is bash "software"? [09:46] .. I dont know [09:47] xiaoji: ok. Going back a step, why do you want to know statistics about this? [09:48] I saw many software in gnome-software,they have icon. I want know how many [09:51] PR snapd#7558 closed: boot,dirs,image: various refinements in the prepare-image code switched to seedwriter [09:56] Chipaca: I seem to remember zyga suggesting it during previous discussions, and it seemed like a reasonable idea. I can remove that part of the PR quite easily though [09:59] jamesh: I still like it [10:09] jamesh: i don't think it's a bad idea, and i like the refactor, but i feel it should be separate from the feature [10:09] and maybe we should discuss moving other internal commands to this [10:10] Chipaca: I can split it out easily enough: it's the first commit. [10:10] jamesh: neat [10:10] jamesh: I'd expect pedronis to have an Opinion about that, and he's off today [10:10] just fyi [10:17] Wimpress: is there anything like an eoan mate installer already? [10:19] facubatista: "eoan mate", https://flic.kr/p/6wqZBC [10:20] xiaoji: so, for an approximation [10:20] xiaoji: open a terminal, and see if you have the "appstreamcli" command [10:21] xiaoji: if you do, then the number of things gnome software offers you is the number in Summary of "appstreamcli status", *plus*, the output of "wc -l /var/cache/snapd/names" [10:22] approximately; some things might be counted twice in the above [10:22] if you _don't_ have appstreamcli, you need to 'sudo apt install appstream' and then 'sudo apt update' for the number to be right [10:23] both numbers should be ~2k or more [10:23] although it varies by arch and region [10:41] PR snapd#7589 opened: cmd/snap: add ability to register "snap internal" commands [10:50] Chipaca: Do you mean an iso image for Ubuntu MATE 19.10? [10:51] Wimpress: i think i found it at http://cdimage.ubuntu.com/ubuntu-mate/daily-live/current/HEADER.html [10:51] heh, dunno why google took me to the header, but ok [10:51] Yep, that's what you want 👍 [11:03] pstolowski: replied in #7443 [11:03] PR #7443: timeutil: fix schedules with ambiguous nth weekday spans [11:04] mborzecki: ok [11:21] Wimpress: ooh, installer crashed :-) [11:23] Chipaca, a mate, indeed, why eoan? [11:23] facubatista: eoan → mañanero [11:23] roughly [11:23] Chipaca, ah, didn't know that [11:24] facubatista: https://en.wikipedia.org/wiki/Eos [11:24] facubatista: "diosa de la aurora" [11:25] facubatista: I would've written that as "diosa del aurora" but maybe la->el to avoid vowel clash is frowned on these days? [11:25] Chipaca, much funnier name the roman equivalent [11:26] "mater matuta" [11:26] facubatista: :) [11:27] xnox: if the ubuntu installer crashes, is there anything useful or interesting for you? [11:27] or should i just do over [11:27] (maybe not pick zfs next iteration) [11:28] Chipaca: if it installed /var/log/installer otheriwse /var/log/syslog /var/log/partman and like /var/log/ubiquity [11:29] Chipaca: possibly check with didrocks & jibel as they have been working on zfs bit in desktop iso. [11:29] xnox: it did not install, I don't think, it was fairly early in the process [11:29] ack [11:29] I now (after a few minutes) got the "system program problem detected" dialogue [11:29] maybe the "report problem" button gets all those bits of info? [11:29] * Chipaca might be overly optimistic [11:29] it will [11:30] but you can also do $ ubuntu-bug ubiquity [11:30] (will ask for SSO login.....) [11:30] * Chipaca clicks the button and waits patiently [11:30] xnox: ack [11:30] zyga: hmm lxd snap https://paste.ubuntu.com/p/HzGx7PG4pc/ [11:30] Chipaca, give me the bug # once it's reported [11:31] jibel: will do, thanks [11:31] still waiting on apport [11:32] xnox: the "report bug" button didn't seem to do anything fwiw, at least, no apport activity from there [11:32] Chipaca: ubuntu-bug ubiquity [11:32] yep [11:32] Chipaca: at least that should make a thing in /var/crash/* [11:33] Chipaca: any files from /var/crash/* will be useful [11:33] oh god my lp password with no password manager [11:33] * Chipaca reaches for scp [11:37] jibel: #1847748 [11:37] Bug #1847748: installer crashed

[11:38] mvo, zyga, mborzecki: https://bugzilla.redhat.com/show_bug.cgi?id=1760568 [11:38] any of you have any idea of a way to fix this? [11:38] jibel: let me know if you need any more info from the system before i nuke it and start over [11:39] Eighth_Doctor: saw it in the monring, but wanted to try it myself [11:39] 👍️ [11:39] Chipaca, thanks. I'll comment on the report [11:40] Eighth_Doctor: but yes, /var/home is a problem [11:41] Eighth_Doctor: also, not sure why silverblue is not using a bind mount for /var/home -> /home [11:41] Eighth_Doctor: looking [11:41] aha, this one [11:41] mborzecki: that's what langdon and I said too... [11:41] but unfortunately it's an ostree thing :( [11:41] there is a way to fix it but not anytime soon IMO [11:42] enough stuff to fix :/ [11:42] we could remap $HOME to /home/$LOGNAME [11:42] but that's far from close [11:42] * pstolowski walk & lunch [11:46] output from gtest is an unreadable mess [11:48] zyga: duh, they changed it like 2 releases back [11:48] yeah, but it only matters when you have to look [11:55] zyga: can we add `/var/home` as an allowed redirection? [11:55] or maybe just follow the symlink and bind-mount the target? [11:56] (I assume that's what we're doing here with the homedir perm?) [11:56] Is there an interface that would allow a command to stop and/or start a service? https://forum.snapcraft.io/t/stop-and-or-start-a-service-from-a-command/13649 [11:56] Eighth_Doctor: not easily [11:56] there's more to it than that [12:01] off to pick up the kids [12:02] jibel: from your comment I gather you don't need more info and I can re-install [12:02] Chipaca mvo https://github.com/snapcore/snapd/pull/7590 [12:02] PR #7590: seed: seed16: run adding snaps in parallel [12:02] PR snapd#7590 opened: seed: seed16: run adding snaps in parallel [12:02] Chipaca if you help me to land this, I will owe you one extra beer :P [12:05] ondra: hmmmmmmmmmmmm [12:05] ondra: I suspect it needs locks in addEssential, at least [12:05] ondra: but that's just a quick pass [12:05] ondra: I'll take a deeper look later [12:05] Chipaca not really [12:05] Chipaca I did keep order there [12:06] ondra: you are writing to a shared map from multiple goroutines [12:06] Chipaca have a look, only non assync is gadget [12:06] ondra: that's what i mean [12:06] Chipaca errors are written to shared map [12:06] Chipaca but there does not seem to be anything which would be related othetrwise [12:07] have a look [12:07] will do [12:07] ondra: in any case just the fact that you found out this speedup opportunity is awesome [12:08] so I'm happy [12:08] I might be even happier after reviewing this, who knows [12:08] i just don't have the mental bandwidth right now to do it properly :-) [12:09] Chipaca of course, this is not critical :) [12:09] Chipaca and it's not remedy to all, make no different on single core, or if you have 3 snaps [12:10] ondra: yeah, for that we need go to take ian's sha3 patches [12:10] ondra: there we get ~60% speedups across all armhf [12:10] Chipaca imagine combined those two :) [12:10] ondra: ikr [12:11] Chipaca unless that assembly code also utilises multicore [12:12] i'm sure there exist people that can write multicore assembly [12:12] i think it's all black magic anyway === ricab is now known as ricab|lunch [12:26] hi, every time I open firefox it tries to read /var/lib/snapd/desktop/icons which is blocked by apparmor. Is it something fixable? [12:27] vidal72[m]: what snapd version are you on? [12:28] Chipaca: one from ubuntu 19.04 [12:28] vidal72[m]: what does 'snap version' say? [12:29] I'm not on that machine atm, give me a sec [12:31] Chipaca: 2.42 series 16 [12:32] vidal72[m]: are you sure you're seeing these messages with that version? (it's from yesterday) [12:32] old stuff then :P [12:32] yes, just reproduced minute ago [12:32] jamesh: have you seen this? [12:33] vidal72[m], is that on app startup or if you do some specific action (opening a file or whatnot) [12:33] Chipaca, yes, I've enough info, thanks [12:34] jibel: good because i'm already finishing the re-install :-) this time around i had no issues [12:35] Chipaca: it's not surprising that apps would try to read various paths under /var/lib/snapd/desktop: that directory is included in $XDG_DATA_DIRS, and various specs say to check all locations contained there [12:35] ogra: app startup [12:35] jamesh: hrm, i thought we'd added desktop/icons as ok as part of your icons work [12:35] maybe i was confused then [12:36] Chipaca: not for the benefit of confined apps: that was for the host [12:36] ah [12:36] should we add a rule just to not have it spam syslog? [12:37] (i don't mean a rule to allow the read) [12:37] +1 [12:38] perhaps [12:38] is it sensitive dir? I think desktop/applicationds are allowed to read [12:38] browser-support currently allows access to /var/lib/snapd/desktop/applications, yes [12:39] but it probably shouldn't [12:39] it's not really useful and is a data leak [12:40] I think those should be both allowed or both denied [12:41] my icons folder is empty anyway [12:42] presumably firefox was trying to access /var/lib/snapd/desktop/icons previously, but it didn't warn due to the directory not existing [12:43] as long as we have that directory in XDG_DATA_DIRS, I agree that warnings for access denials are just noise [12:48] jamesh: desktop-legacy also allows desktop/applications https://github.com/snapcore/snapd/blob/release/2.42/interfaces/builtin/desktop_legacy.go#L229 [12:49] vidal72[m]: yep. I don't think it is useful there either, since neither interface provides a way to launch those applications or access e.g. referenced icons [12:51] heh, why it was added then? isn't it needed for xdg-open? [12:51] which doesn't work in ff anyway but that's another story... [12:52] mborzecki: https://github.com/snapcore/snapd/pull/7591 [12:52] PR #7591: cmd/snap-confine: remove loads of dead code [12:53] PR snapd#7591 opened: cmd/snap-confine: remove loads of dead code [12:53] zyga: can you lead the standup? I don't think I'm going to make it [12:54] syre [12:54] *sure [12:54] zyga: thynks [12:54] ny proolbm [12:54] :) [12:54] vidal72[m]: I wouldn't be surprised if the rules grew out of watching the audit logs for various apps without fully thinking through whether it made sense [12:55] see you at the standup [12:55] zyga: probably not, that's the point :) [12:55] jamesh: should I make PR which deny desktop/{applications,icons} then? [12:55] vidal72[m]: similar to the icons case, /var/lib/snapd/desktop/applications existed, so it would generate audit messages without the rule allowing read access [12:55] vidal72[m]: jamesh: suspect jdstrand might need to peep into this [12:56] vidal72[m]: perhaps start a thread on the forum and tag @jdstrand. [12:56] ok [12:57] another thing: did anyone considered allowing writing to @{PROC}/@{pid}/clear_refs https://forum.snapcraft.io/t/call-for-testing-chromium-browser-deb-to-snap-transition/11772/17 [13:01] zyga: wait i thought mvo was off today [13:04] Chipaca: he's here - he changed his mind apparently. [13:08] Chipaca: I was considering it [13:08] jamesh: done https://forum.snapcraft.io/t/reading-var-lib-snapd-desktop-applications-icons/13650 [13:36] mvo: thank you for the review :) [13:36] did you read the commit messages? [13:36] interestingly snap-store snap wants to read my /home/user/documents dir on each startup which is blocked === ricab|lunch is now known as ricab [13:45] hmm, how does one use (install, run) snaps inside a docker container? is that even plausible? [13:46] 🤯 [13:46] roadmr: you have tips? :) [13:46] only to run away as fast as you can! [13:46] * roadmr does, woohooo [13:47] well I was hoping zyga would just tell me that it all works :D [13:47] sparkiegeek: no can do :( [13:47] sorry, docker is not able to run docker either [13:48] k, am I basically SOL? [13:48] I'm afraid so [13:48] sparkiegeek: why must it be docker? [13:49] zyga: np, knowing is better than flailing around [13:49] roadmr: Reasons™ [13:49] aha, got it :0 [13:49] heeh :) [13:57] man this system feels slow even over ssh [13:57] ondra, Chipaca: IIRC the SHA3 NEON assembly is not parallelized, so each core has it's own NEON processor and the parallel speedup that ondra has should multiple together with the assembly patches [13:59] we should also test on an imx6ull to see if there is any impact on single-core SoCs (positive or negative) [14:03] the imx6ull would definitely benefit from the SHA3 patches (I actually tested on one of those :-D ), probably not so much from ondra's parallelization patch though [14:03] well, the first is a given, the latter is what caused my question :) [14:04] it might improve, leaving the ordering to the kernel effectively ... or it might degrade because it tries to parallelize where it can not [14:12] ogra go seems to be somehow smart about loading, so if one tasks pegs the core, it will probably not try to run too many in parallel [14:14] * zyga -> lunch [14:20] cmatsuoka: after much chuntering, I have a booted core! [14:22] jibel: not being too scientific, it seems on this system I need to let it finish disk activity for partitioning before choosing country and creating the user, or it fails [14:22] jibel: "not too scientific" == tried it both ways twice [14:36] popey: Wimpress: I think mate would be better if it waited for seeded before giving you login [14:36] better first boot, you'd get the welcome app for one [14:41] ondra, well, i'd still perfer if you could hand it to cwayne to see actual results on real imx6ull hw [14:46] cmatsuoka: - Update assets from gadget "pc" (44) (cannot read current gadget snap details: invalid volume "pc": invalid structure #2 ("Recovery"): invalid role "system-recovery": unsupported role) [14:46] cmatsuoka: expected? [14:50] ogra feel free to get him involved [14:50] well, you got the patched binary :P [14:50] i'm just sayig we should make sure to have them in the loop before this goes to stable [14:51] i guess they'll run QA anyway [14:51] (on that HW) [14:59] cmatsuoka: mvo: core20 needs to drop the ConditionKernelCommandLine=snap_core from snapd.system-shutdown.service [14:59] :) [14:59] i need to dig into what creates that now that it's automagically put in etc by somethingorother [14:59] not now tho; now, tea [15:00] sil2100, hey, not sure if you saw my update to https://github.com/CanonicalLtd/ubuntu-image/pull/175 ? [15:00] PR CanonicalLtd/ubuntu-image#175: Little kernel bootloader support [15:03] chrisccoulson: indeed, good point [15:06] mvo, was that meant for me? [15:08] chrisccoulson: sorry, for Chipaca [15:09] * Chipaca hands chrisccoulson a "got called Chipaca" beginners badge [15:11] * roadmr is now nown as Choadmr [15:13] * Chipaca wonders if roadmr knows what a choad is [15:13] * roadmr googles [15:13] * roadmr is now known as google-before-changing-nick-mr [15:13] :) [15:23] Chipaca: thoughts on #1847788 ? is this a bug or "AsDesigned"? [15:23] Bug #1847788: reverting and installing a snap deletes a disabled revision that should be kept [15:24] Is there a problem if I maintain a channel branch (eg edge/strict) with a strictly confined snap revision while the rest of the snap revisions in the channel are classic? https://forum.snapcraft.io/t/mixing-strictly-confined-and-classic-snap-revisions-in-a-channel/13652 [15:25] kjackal: migrations from one and the other are not automatic [15:25] you can go classic -> strict [15:25] ijohnson: the one that's removed is _blacklisted_ (because revert) [15:25] but not the other way around [15:25] ijohnson: so that changes the behaviour a bit [15:25] ijohnson: you could instead snap refresh --revision x1, that should dwyw? [15:25] Chipaca: I want to keep revision x2 though [15:26] (though this is for a test, so bit of an arbitrary use case) [15:26] thanks zyga, it is going to be only me using this channel branch for demo purposes (i hope) [15:26] ijohnson: yeah, it sounds like you want a refresh, not a revert [15:26] ijohnson: it's also possible it's a bug, but i think the refresh should leave x2 there [15:26] basically I want to keep like 4 revisions of a snap around, going back and forth between them with reverts and refreshes without any of them being garbage colledged [15:26] if that makes sense [15:27] kjackal: you can use a branch [15:27] kjackal: not a channel [15:27] ijohnson: why not with refreshes? [15:27] that's good for demos [15:27] I do need a revert though because I'm testing that revert does the right thing too :-) [15:28] ijohnson: ¯\_(ツ)_/¯ revert will discard things after 'current' [15:28] zyga: I am probably messing up the terminology edge/strict is a channel with strict being a branch, yes/no ? [15:28] no, that's a risk level [15:28] Chipaca: yes that seems to be what it does now, but do you think that's a bug or AsDesigned? [15:28] you can create an actual branch [15:28] like demo [15:28] edge/demo [15:28] or demo/edge, I don't recall [15:29] people need to refresh to it explicitly [15:29] ijohnson: i mean, https://github.com/snapcore/snapd/blob/master/overlord/snapstate/snapstate.go#L291 [15:29] zyga: demo/edge has demo as a track, while edge/demo has demo as a branch IIRC [15:29] ijohnson: // discard everything after current [15:29] ijohnson: it's very intentional :-) [15:29] I thought the channel had a // format [15:30] you are probably right, as I said I don't recall [15:30] my whole point is that it's a good way for actual demos [15:30] kjackal: it should be fine, subject to the limitations of branches [15:30] yeap, thank you people [15:30] i mean, they expire after a while unless you keep on pushing to them [15:30] that sort of thing [15:31] Chipaca: git blame says this was from mvo with commit message "review feedback" :-/ [15:31] alright if it is AsDesigned that's fine I guess [15:31] just makes my tests more complicated [15:31] "review feedback" is the worst [15:31] ijohnson: https://github.com/snapcore/snapd/pull/1457 [15:31] PR #1457: snapstate: drop revisions after "current" on refresh [15:32] i mean, it's not good, but you grab the commit hash and look it up and you see the review feedback [15:32] so it's not _the worst_ :) [15:32] oh god, "retest this please" [15:33] * Chipaca hugs spread [15:35] ijohnson: what do you need the revision to still be there after revert for? [15:35] ijohnson: i mean, which is the path that you're testing for? [15:36] because I want to reuse the revision later in a different permutation, i.e. one test case is reverting backwards in time to that revision to make sure things work, another use case is to refresh to that revision [15:36]