[05:14] morning [05:43] o/ [05:44] mborzecki: can you look at https://github.com/snapcore/snapd/pull/8726 [05:44] it's a one liner and I have a follow-up it opens [05:44] PR #8726: tests: silence stderr from dbus-monitor [05:46] zyga: ack [05:47] for some context, privately I've been using a version of https://github.com/snapcore/snapd/pull/8728 [05:47] PR #8728: tests: detect stray dbus-daemon [05:48] and it essentially prints a TODO list of tests to improve [05:49] mborzecki: I replied but let's chat here [05:49] right now there's nothing more on stderr, if it goes to a file it'd be rather short-lived as that directory goes away when session-tool quits [06:13] zyga: if something fails does is the directory removed too? [06:16] yes [06:16] I looked at the rest of the script and we log other things there [06:16] but it's a remnant of development when I just would comment out the cleanup logic [06:21] fedora-32 is listed in the workflows but doesn't exist in spread.yaml, this is making it fail constantly [06:24] pedronis: oh? [06:24] perhaps mvo added that with the caching PR [06:25] seems so [06:25] zyga: see here: https://github.com/snapcore/snapd/pull/8729 [06:25] PR #8729: snap,many: mv Open to snapfile pkg to support add'l options to Container methods [06:27] yeah, indeed, it's there in the patch [06:28] it's not marked as requires so it was masked by opensuse failing as well, I suspect [06:29] if we cannot get a working image today I'll drop it from the matrix with a comment [06:29] it was marked as required [06:29] or maybe I'm confused [06:29] anyway [06:30] #8729 needs a 2nd review [06:30] PR #8729: snap,many: mv Open to snapfile pkg to support add'l options to Container methods [06:37] i'll fix the sanpd selinux policy today so we'll be able to merge cachio's fedora 32 pr [06:39] pedronis: https://github.com/snapcore/snapd/pull/8729#pullrequestreview-418005303 [06:39] PR #8729: snap,many: mv Open to snapfile pkg to support add'l options to Container methods [06:39] mvo: mborzecki: hi, Pawel's #8697 needs reviews [06:39] mborzecki: cool [06:39] PR #8697: packaging: build cmd/snap and cmd/snap-bootstrap with nomanagers tag [06:41] pedronis: sure, looking [06:42] PR snapd#8723 closed: github: remove workaround for bug 133 in actions/cache [06:43] PR snapd#8721 closed: devicestate, sysconfig: revert support for cloud.cfg.d/ in the gadget [06:55] PR snapd#8729 closed: snap,many: mv Open to snapfile pkg to support add'l options to Container methods [07:03] morning [07:14] PR snapd#8730 opened: data/selinux: allow snapd to remove/create the its socket [07:16] quick drive by fix when testing package updates pushed by Eighth_Doctor ^^ [07:16] mborzecki: hi, can you take another look at https://github.com/snapcore/snapd/pull/8697 ? [07:16] PR #8697: packaging: build cmd/snap and cmd/snap-bootstrap with nomanagers tag [07:16] sure [07:22] pstolowski: small comment https://github.com/snapcore/snapd/pull/8697/files#r430205753 [07:22] PR #8697: packaging: build cmd/snap and cmd/snap-bootstrap with nomanagers tag [07:23] not incorrect, just spreading the awesomeness of make [07:30] zyga: that's nice, thank you [07:31] zyga: oh, nice! [07:31] mvo: the prereq has landed squash merged, now #8711 needs a 2nd review and hopefully to be (almost) green [07:31] PR #8711: snap/squashfs: also symlink snap Install with uc20 seed snap dir layout

[07:31] zyga: this should make some of the other stuff we do much more readable if we use this trick [07:32] pedronis: \o/ I have a look [07:32] mvo: cool :) [07:33] mvo: it's annoying we need to repeat this across packaging [07:33] I wish we'd use snaps.mk more [07:34] brb [07:34] zyga: unfortunately I don't think that's an option for debian/ because if we ever want someone to maintain this it has to be as clsoe as possible to a "standard" [07:35] mvo: I don't think so, it's not unlike dh_auto* and using a makefile in the upstream project [07:35] mvo: it's all about integrating dpkg-buildflags inside our native makefile [07:43] PR snapd#8731 opened: snap,many: mv Open to snapfile pkg to support add'l options to Container methods (2.45) [07:43] pedronis: I backported 8711 to 2.45, straight cherry-pick did unfortunately not work [07:44] mvo: it's because of the prereq? #8729 ? [07:44] PR #8729: snap,many: mv Open to snapfile pkg to support add'l options to Container methods [07:44] or something in it itself? [07:46] mvo: ah, you backported the prereq actually https://github.com/snapcore/snapd/pull/8731 ? [07:46] PR #8731: snap,many: mv Open to snapfile pkg to support add'l options to Container methods (2.45) [07:47] mvo: there's a bit of extra tests in boot, do they pass? [07:51] mvo: we should backport #8721 [07:51] PR #8721: devicestate, sysconfig: revert support for cloud.cfg.d/ in the gadget [07:59] pstolowski: with this trick we could actually reduce quite a lot of clutter in the code building commands [07:59] pstolowski: with this trick we could actually reduce quite a lot of clutter in the code building commands [07:59] zyga: ok, but perhaps as followups [07:59] yes, totally [07:59] zyga: i'd really like to get this landed to unblock stuff [07:59] let me do it after your bits land [08:02] pedronis: 8721> yes, looking at this too [08:09] pstolowski: can you take a look at https://github.com/snapcore/snapd/pull/8730 ? [08:09] PR #8730: data/selinux: allow snapd to remove/create the its socket [08:13] mborzecki: https://github.com/snapcore/snapd/pull/8656#discussion_r430233548 [08:13] PR #8656: snap-mgmt: perform cleanup of user services [08:16] siure [08:20] i'll force push to cachio's f32 pr and drop the last commit he made [08:20] PR snapd#8726 closed: tests: silence stderr from dbus-monitor [08:21] ta mvo [08:21] and mborzecki :) [08:21] PR snapd#8725 closed: testutil/exec.go: set PATH after running shellcheck [08:24] PR snapd#8730 closed: data/selinux: allow snapd to remove/create the its socket [08:25] PR snapd#8604 closed: interfaces/builtin/desktop: do not mount fonts cache on distros with quirks [08:41] 8351 needs a second review [08:44] pedronis: do you want to do a review of 8661? it has two +1 at this point (but none from you) [08:47] PR snapd#8732 opened: data/selinux: update policy to allow forked processes to call getpw*() [08:47] pstolowski: another one for you ^^ [08:47] mborzecki: siure [08:47] grr [08:47] *sure [08:48] zyga: can you remind me where in the spread suite do we check for leftovers from namespaces etc? [08:48] pstolowski: what do you mean by leftover from namespaces? [08:49] zyga: sorry, probably mount changes; anything that lxd can leave behind [08:50] I see [08:50] I don't think we do that regularly [08:50] we kind of do it in one spot in mount-ns test [08:50] a branch that I have long open tried to but it picked up too many failures so it was unable to land [08:50] we may get that, a bit, with invariant-tool [08:50] zyga: aah, mount-ns test, i think that was it [08:50] but it needs to be done still [08:52] thanks [08:57] brb [09:00] mvo: ? I don't see the 2nd review [09:01] mvo: as I remarked I think we should apply mborzecki idea of trying not to restart if nothing has changed [09:01] mvo: I think it needs a review from pstolowski [09:03] pedronis, mvo i'll review it in a bit (and also the vitality PR) [09:04] re [09:06] PR snapd#8711 closed: snap/squashfs: also symlink snap Install with uc20 seed snap dir layout

[09:06] mvo: ^ I squash-merged it [09:08] pedronis: \o/ [09:09] pedronis: re 8661> sorry, ian did two +1 I misread that [09:10] ah, was confused for a sec [09:10] pstolowski: actualFsTypeAndMountOptions doesn't need to return an error anymore as well [09:10] pedronis: missed that, thanks [09:11] pstolowski: np, but that's where I started thinking about this, because I had move the call to it and then wondered if it would change the failure modes but actually it can't error in itself [09:12] *I had you [09:13] pstolowski: mvo: do we need/want the preseed fixes in 2.45.1 ? [09:14] quick errand, back in 30 [09:16] PR core20#65 closed: Do not attempt signing changes during snap build [09:16] mvo: this is the comment I was referring to in 8661: https://github.com/snapcore/snapd/pull/8661#discussion_r427920348 [09:16] PR #8661: configcore: add "service.console-conf.disable" config option [09:16] pedronis: that would be nice to have [09:17] Issue core20#63 closed: fix travis builds with new subiquity from git [09:17] Issue core20#66 opened: Use focal base in travis [09:17] otherwise it will take long time before it's easily usable with lxd [09:19] pedronis: preseed> it would be a good idea, yes [09:20] pedronis: 8661> aha, sorry, I will look at this [09:24] PR core20#49 closed: fix broken symlinks in /etc/writable [09:26] mvo: I will most likely miss standup today [09:26] mvo: I'll send notes to the standup doc [09:27] I may join from a phone if I have the conditions to [09:30] zyga: ok [09:30] pstolowski: mvo: #8697 failed on centos-8, seems unrelated but not sure [09:30] PR #8697: packaging: build cmd/snap and cmd/snap-bootstrap with nomanagers tag [09:31] pedronis: yes, i've restarted it [09:35] * zyga goes to look at 8697 restart not wasting money anymore [09:47] re [10:07] mborzecki: Ian made a suggestion in #8718 then it should be good to go [10:07] PR #8718: boot, many: require mode in modeenv [10:18] pedronis ijohnson thanks for the reviews [10:18] * zyga hugs mvo [10:18] mvo: the caching behavior is BRILLIANT [10:18] 2 seconds [10:18] need to address some commentrs in 8716 too [10:18] 0.00 cost [10:18] zyga: hah nice :P [10:18] I wonder if this will show up in load charts on the spread runner [10:19] as we're usually bound by memory [10:19] morning folks [10:20] ijohnson: heya [10:20] zyga: re: your comment on 8729 about the header being one byte larger than magic, the magic bytes is just 4 bytes, so a header of 20 bytes already is 16 bytes larger ? [10:20] hey mborzecki [10:20] ijohnson: yeah but it's a constannt [10:20] PR snapd#8733 opened: tests: port document-portal-activation to session-tool [10:20] it's okay but not for the right reasons :) [10:20] zyga: so you think it should be only 5 bytes ? [10:21] I think it should be exactly len(magic) + 1 [10:21] :D [10:21] got it [10:23] PR snapd#8734 opened: tests: port interfaces-wayland to session-tool [10:23] PR snapd#8735 opened: snap/snapfile,squashfs: followups from 8729 [10:23] zyga: applied in a followup ^ [10:23] super [10:23] thanks! [10:24] I was going to do it myself but I was working on a streak of patches [10:33] mborzecki: https://github.com/snapcore/snapd/pull/8736 [10:33] PR #8736: tests: log stderr from dbus-monitor [10:33] ijohnson: not a fan of len(magic)+1, it's not particularly better or worse than the existing especially now that its in squashfs itself [10:33] PR snapd#8736 opened: tests: log stderr from dbus-monitor [10:34] * zyga returns to https://github.com/snapcore/snapd/pull/7825 [10:34] PR #7825: many: use transient scope for tracking apps and hooks [10:34] ijohnson: I'll push something else there [10:34] pedronis: ok [10:34] ijohnson: I mean a 5 bytes file is never a valid squashfs anyway [10:34] well that's not really for us to decide ? [10:35] unsquashfs is the ultimate decider of if it's a valid squashfs ? [10:35] ijohnson: not but as a predictor this is worse than before [10:35] ijohnson: if we didn't care we could use len(magic) and Equal [10:35] I mean for something that's called magic I wouldn't expect it to be perfect [10:36] why the +1 [10:36] anyway I will do yet something else [10:36] and I think we are splitting hairs over a tiny thing [10:36] sure feel free [10:36] I just took zyga's suggestion [10:36] I have no strong feelings about this really [10:36] just feels a bit pointless [10:36] ijohnson: I know, I disagree with the suggestion, doing nothing would have been fine [10:37] ijohnson: also fn to f is not an improvement either [10:37] ijohnson: anyway, it's not your problem anymore, I will push something there [10:37] PR snapd#8737 opened: snap/squashfs: also symlink snap Install with uc20 seed snap dir layout (2.45) [10:41] Cannot allocate google:opensuse-tumbleweed-64: cannot allocate new Google server for opensuse-tumbleweed-64: invalid value for field 'resource.shieldedInstanceConfig': ''. Shielded VM Config can only be set when using a UEFI-compatible disk [10:49] pstolowski: known [10:49] sergio needs to fix the images [10:49] ok [11:07] PR snapd#8731 closed: snap,many: mv Open to snapfile pkg to support add'l options to Container methods (2.45) [11:12] PR snapd#8737 closed: snap/squashfs: also symlink snap Install with uc20 seed snap dir layout (2.45) [11:13] PR snapd#8738 opened: snap,many: mv Open to snapfile pkg to support add'l options to Container methods (2.45) [11:25] ijohnson: once I was looking there, I added some tests to 8735 as well [11:25] pedronis: ok, seems the scope of the PR has greatly increased though [11:27] anyways approved it [11:27] * ijohnson afk for bit [11:32] pstolowski: zyga: I tweaked/expanded a little bit #8729, it needs a re-review [11:32] PR #8729: snap,many: mv Open to snapfile pkg to support add'l options to Container methods [11:32] yeah, reviewing now [11:32] okay [11:32] pedronis: i've commented on the tooling proposal [11:35] pedronis: reviewed [11:37] pedronis: hmm but 8729 was merged? [11:37] heh sorry [11:38] pstolowski: I meant #8735, but another review is not stricly needed anymore [11:38] ah 8735 i guess [11:38] PR #8735: snap/snapfile,squashfs: followups from 8729 [11:38] yep [11:40] pstolowski: seems 8697 can be merged now [11:41] which should unblock 8533 [11:42] pstolowski: thanks for commenting there [11:42] pedronis: I agree with pawel's comments [11:43] pedronis: I don't think we need a distinction between suite/task in the name, a common prefix plus something relevant to the tool function is good for me [11:43] PR snapd#8697 closed: packaging: build cmd/snap and cmd/snap-bootstrap with nomanagers tag [11:43] pedronis: if you pick a name for session tool I'll do a test mass patch to see how it looks like [11:44] zyga: I'm not against the prefix, my issue is that I really care about distinguishing things that affects state across tasks vs not [11:45] hmmm [11:46] PR core20#67 opened: tests: use SNAPCRAFT_PRIME dir instead of hardcoding "prime" [11:46] so things like retry-tool vs things like lxd cleanup? [11:48] pedronis: under this classification, session tool does not affect state across tasks, correct? [11:51] zyga: well session can survive if don't clean them up, no? [11:51] pedronis: sure but the point is to pair session prepare and restore [11:52] so it affects state across task, unless you use it just right [11:52] the classification is based on ideal usage, it's about what gets touched [11:52] I think virtually everything has this property [11:52] mkdir does [11:53] zyga: mkdir is a tool [11:53] sorry [11:53] mkdir is not a tool [11:54] so what is it? [11:57] zyga: for analogy, it's the difference between a helper function vs a methond on the suite struct in go [11:57] ok [11:57] we have some stateless helpers attached to suite sometimes [11:58] I think attaching this to suite is misleading as it reflects a particular way we use it, I look at it more like a context manager in python, it's an enter/exit pair with user code in between [11:58] zyga: we don't have a with statement [11:58] pedronis: ijohnson: https://github.com/snapcore/snapd/pull/8716 is updated now [11:58] as such it could be used with the same vailidity, at the level of a task or the level of a suite [11:58] PR #8716: o/devicestate: refactor current system handling [11:59] pedronis: sure but the pairing nature is similar to one, it's just an analogy [11:59] zyga: anyway it seems you are objecting to suite. , not to the dinstiction [11:59] yes, and I understand your point better now [12:00] so that glancing at the tool name gives some context as to what is the possible impact [12:00] yes [12:00] I don't know if I have a better name suggestion but I see the goal [12:03] zyga: pstolowski: how much is short names a goal for you? [12:04] pedronis: I'm okay with a non-short name [12:04] time to revisit https://github.com/snapcore/snapd/pull/7414 [12:04]