[04:17] Bug # changed: 1663615, 1667359, 1681328, 1682550 [04:20] Bug #1667408 changed: docker container command XXXX could not be invoked === Lin-Buo-Ren-alt1 is now known as Lin-Buo-Ren [04:25] https://forum.snapcraft.io/t/how-to-create-a-lxd-container-for-snap-development/4658 [04:25] RFC [05:13] morning [06:25] hey [06:25] * zyga will be a little late due to DST change [06:25] zyga: hey [06:29] hey ho [06:29] so this week will be, hopefully, more at ease than the last one [06:30] * zyga -> breakfast [07:13] good morning === pstolowski|afk is now known as pstolowski [07:14] morning guys! [07:16] o/ pstolowski [07:23] Hey hey [07:40] PR snapd#4929 opened: cmd/snap-confine: apparmor: allow creating prefix path for gl/vulkan [07:40] Looking [07:42] there's probably a nicer way to write this rule, would appreciate suggestions [07:46] Not in a major way [07:46] It is fine [07:47] oh, ok ;) [07:47] zyga: can you take a look at https://github.com/snapcore/snapd/pull/4908 too? [07:47] PR #4908: [RFC] cmd/snap-confine: attempt to detect if multiarch host uses arch triplets [07:49] zyga: a thought i had during the weekend, maybe we should dump all the figuring out of nvidia dirs from snap-confine and do it in snap command instead, only pass the paths we detected in some predefined environment variables [07:49] Yes, that is the goal [07:50] I think we have all the means now [07:50] something SC_NVIDIA_LIB=/usr/lib/..., SC_NVIDIA_METHOD=symlink, or SC_NVIDA_LIB=/usr/lib/nvidia-340xx, SC_NVIDIA_METHOD=bind, simiarly for 32bit variants [07:50] No no [07:50] Just use the mount profile [07:51] No need for more variables [07:51] .fstab files? [07:52] Yes they have all the bits needed now [07:52] I wanted to do this for quite some time now [07:56] Move the existing system over [07:56] Let snapd handle this [07:56] Rip the c code out [07:58] Nvidia mode gets added to system key [07:58] Later on it just becomes nvidia snap [07:59] zyga: yes, this will work [08:00] meanwhile, 4908 fixes nvidia not working at all with 18.04 :/ [08:03] running new gnome, they seem to half-respect LC_DATE, it currently shows 'poniedziałek, 26 March 2018' [08:04] mborzecki-ubuntu: oh? maybe it's IPC'd over from gnome-something-something [08:04] mborzecki-ubuntu: looking at 4908 now [08:04] * zyga got back from a walk [08:07] how did the failure look like? [08:08] zyga: which failure? [08:09] the 4908 fix for nvidia on ubuntu [08:10] zyga: well, no 3d with nvidia on 18.04 [08:10] zyga: some details https://forum.snapcraft.io/t/nvidia-gl-libs-access-broken-on-ubuntu-18-04/4440 [08:10] let me read that [08:11] so I'm puzzled, you said that your triplet branch made it work on 18.04 [08:11] or was that still on arch? [08:12] zyga: arch fixes are already merged [08:12] and what happens on ubuntu with that multi-arch branch? [08:12] zyga: triplet is for ubuntu, but uses the same thing as is used on arch [08:13] does it crash, does it just not do anything? [08:13] zyga: without the branch there's no 3d, eg. ohmygiraffe hangs, glxinfo shows swrast [08:13] right [08:14] that's expected now that 18.04 uses libglvnd [08:14] zyga: with the branch, if nvidia libs are in /usr/lib/, they'll get symlinked, otherwise /usr/lib/nvidia-xxx is bind mounted (if it exists) [08:14] right [08:14] and what happens on 18.04? we get the symlinks, right? [08:14] zyga: yes [08:15] and? [08:15] zyga: and it works [08:18] oh [08:19] I must be sleepy then, I misunderstood you [08:19] zyga: heh, no worries, I might be sleepy too and talking garbage :P [08:22] hi, a 2nd review of #4928 would be appreciated (should solve at least of most tests/main/layout failures) [08:22] PR #4928: tests: disentangle etc vs extrausers in core tests

[08:23] good morning [08:23] hey mwhudson [08:24] pedronis: yes, thank you for pointing that out, pstolowski, mborzecki-ubuntu can you please look at 4928 [08:25] mborzecki-ubuntu: did you see this https://forum.snapcraft.io/t/refresh-scheduling-on-specific-days-of-the-month/1239/22 [08:25] looking [08:26] pedronis: yes, that's the 5minute ensure interval playing tricks, i recall asking about that back when i was double checking that it works, but the behavior (and the code) looked so specific that i did not view it as a bug [08:27] well, it's an easy fix [08:28] zyga: hm pretty much any pr to master touchich apparmor profiles has a problem with temp profile files left behind [08:29] mborzecki-ubuntu: I can prepare a PR if you are busy with nvidia debugging [08:29] pedronis: that'd be great if that's not a big problem for you :) [08:29] mborzecki-ubuntu: can you reproduce that with -debug? [08:30] something must have changed to cause this [08:32] zyga: i have 3 prs for master, 4875, 4908, 4929, each touching apparmor profile and all fail tests/main/snap-confine, i was able to get a debug shell in 4875 using the same -seed, but i was only able to confirm there were temp files in /etc/apparmor.d and those files were not in snapd state snapshot [08:32] interesting [08:32] perhaps the key to trigger it is to change the apparmor profile [08:32] that would trigger the code that updates it [08:34] pedronis: btw. that's just calling EnsureBefore() at the right place? [08:34] yes [08:34] the most annoying bit is writing the unit test [08:34] anyway I started already [08:35] pedronis: great, thank you! [08:45] Bug #1758849 opened: Snap not able to enable ssh after core upgrade [08:50] * zyga tests the community update to the opensuse package [09:05] `VK_ICD_FILENAMES=/var/lib/snapd/lib/vulkan/icd.d/nvidia_icd.json` snap run graphics-debug-tools-bboozzoo.vulkaninfo makes it find nvidia icd now [09:09] PR snapd#4930 opened: skip test that requires internet when not present [09:16] PR snapd#4931 opened: configcore: give a chance to immediately recompute the next refresh time when schedules are set [09:17] mborzecki: ^ [09:18] we probably need to revisit validateRefreshSchedule , it has grown convoluted [09:18] pstolowski: still looking at 4928, any questions? [09:29] pedronis: yes, i've a qustion [09:29] pedronis: just got sidetracked.. give me a sec [09:33] pedronis: commented [09:34] zyga: hi. Now that jdstrand seems mostly happy with the "secure bind mount" implementation, what's next as far as integrating it? [09:35] pstolowski: yes, the tail/head usage is preexisting, we could use grep (at least for root and test), the issue is that afaict the user called ubuntu, is not called like that, it depens on the cloud [09:36] so it's a win some/lose some kind of situation [09:37] pstolowski: I suppose we can use a mixed approach, grep for root, and sanity check for test (but copy the two users, the "cloud" one as well) [09:38] pstolowski: seems like a follow material though [09:38] *follow up [09:38] pedronis: yes, 'test' user would be worth checking for [09:38] pedronis: ok [09:39] I'll do a 2nd PR this afternoon [09:41] I'm squash merging it, because we probably want it in 2.32 if we have more development to do there, or if we get in trouble for this problem with the validations [09:42] pstolowski: thank you for the review [09:42] PR snapd#4928 closed: tests: disentangle etc vs extrausers in core tests

[09:43] pedronis: just to be sure, each EnsureBefore(0) causes one iteration of this loop https://github.com/snapcore/snapd/blob/master/overlord/overlord.go#L257 right? [09:44] mborzecki: not each, if you call EnsureBefore(0) many times before the loop has a chance to run, they coalesce to one request [09:44] that's why it's time based [10:00] zyga mvo Morning [10:00] hey flexiondotorg [10:00] mvo is off this week [10:00] Last week before I left for the US I mentioned that snapd 2.32 caused some serious regressions on 18.04. [10:01] Is there a new release to resolve that? [10:01] I'm still experiencing problems. [10:02] niemeyer: ^ [10:02] pedronis: i'm a bit slow today, where are the calls coalesced into one? iirc calling time.Timer.Reset(0) makes it expire immediately, so if the call is made from a handled in one of the .Ensure() calls in overlord/.. the loop will wait (from what i saw those are run in the same goroutine), but if a call like this was made from daemon api, then the loop could turn (different gouroutines), or am i missing [10:02] something? [10:03] zyga: hm got these 2 files: etc/apparmor.d/snap.core.4321.usr.lib.snapd.snap-confine /etc/apparmor.d/snap.core.4321.usr.lib.snapd.snap-confine.pXJnJ6rsK7wX [10:07] mborzecki: do you have a debug session? [10:07] zyga: jest [10:07] zyga: notice this: [10:07] can you stat both of the files [10:07] -rw-r--r-- 1 root root 21297 Mar 26 09:33 /etc/apparmor.d/snap.core.4321.usr.lib.snapd.snap-confine [10:07] and diff them [10:07] -rw-r--r-- 1 root root 21297 Mar 26 09:26 /etc/apparmor.d/snap.core.4321.usr.lib.snapd.snap-confine.pXJnJ6rsK7wX [10:08] so the garbage one is slightly older [10:08] this host executed tests/main/snap-seccomp around 9:26 [10:08] I wonder if we don't do some error checking [10:09] aha [10:09] (have this special patch to spread that logs machine names running each test) [10:10] there is nothing in that test that looks like it would touch snap-confine [10:10] and we only make that file when snapd starts up [10:11] actually, and when it installs new core snap [10:11] zyga: hmm 2018-03-26 11:26:12 Executing google:ubuntu-16.04-64:tests/main/snap-seccomp (1/230) (google:ubuntu-16.04-64 (mar261114-742037))... [10:18] mborzecki: we move reset only if the next is before the current next, but yes usually calls from api.go will provoke many ensures [10:19] pedronis: thanks for confirming :) [10:22] zyga: sorry, for some reason I only got your email now [10:22] zyga: about opensuse package [10:22] hey! [10:22] haha, I'm running "osc build" now :) [10:22] welcome Caelum :) [10:23] I fixed the spaces in the PR, I'll update the patch [10:23] thanks, I'll test it and see what is going on there [10:24] we have some packaging in the git tree [10:24] but it's aimed at our CI loop [10:24] it would be great if you could see how to unify the. two [10:24] Caelum: and again, thank you so much for doing this! [10:24] what was that about the udev rules, just remove those files? [10:24] mmm, which one was that? let me look [10:24] hey, I was only doing this so I could run spotify :D [10:26] I think we don't need to ship the repair service at all [10:26] it's a feature that is only for core devices (distribution that is using just snaps for everything) [10:26] here we can always update snapd with the regular rpm [10:26] ah, I see -%{_udevrulesdir}/80-snappy-assign.rules [10:26] * zyga looks at that file [10:27] so that file we do need [10:28] it's there so that devices that are udev-tagged to a specific snap are actually added to the device cgroup of that snapo [10:28] what made you remove that? [10:29] I didn't touch that [10:29] looking at https://build.opensuse.org/request/show/590994 [10:29] I see -%{_udevrulesdir}/80-snappy-assign.rules [10:31] oh I must have removed it because the dist didn't install it [10:31] oh, I see [10:32] but we still need it, it's also a little bit more complex as it was renamed recently [10:32] let me see what I get in the built rpm now [10:34] do you want me to update to 2.32 ? I didn't do that yet because it wasn't listed as a release [10:34] yes, I think we want that as it brings in a lot of fixes and new features [10:35] but [10:35] let's see if we can get 2.31.x out first [10:35] note that we want to do 2.31.2 probably [10:35] as it's a minor bug fix release and should be the same for packaging [10:35] ah ok, I will update this [10:39] zyga: Regarding snapd in 18.04. [10:40] I've just tested core snap from edge and *all* strictly confined desktop snaps fail to execute. [10:40] `failed to create prefix path: /tmp/snap.rootfs_wo6vzw/var/lib/snapd/lib/vulkan/icd.d: Permission denied` [10:41] mborzecki: ^ is that covered by your recent work? [10:41] or do we need a new PR to update the profiles to match? [10:41] zyga: flexiondotorg: #4929 fixes that [10:41] PR #4929: cmd/snap-confine: apparmor: allow creating prefix path for gl/vulkan [10:42] mborzecki: Thanks. When is that due to land? [10:42] zyga: well ran into a snag already, the vendor releases have the go libs bundled, but the github tarball does not [10:43] Caelum: oh, I see, we didn't make a .vendor tarball? [10:43] OpenGL apps on 18.04 using core from stable either fail to execute or attempt to regression to swrast. [10:43] I think I had a similar problem and I solved it by creating the /var/lib/snapd/lib directory [10:43] flexiondotorg: opengl is well known and not fixed yet [10:43] flexiondotorg: hopefully as sonn as jdstrand gives it a green light (and the build is green) [10:43] Caelum: no, the issue here is much more complex [10:44] flexiondotorg: proposal for fixing opengl is in #4908 [10:44] PR #4908: [RFC] cmd/snap-confine: attempt to detect if multiarch host uses arch triplets [10:44] Caelum: I'll try to get that corrected [10:44] zyga: awesome, because I have no idea how I'd get the go libs otherwise :) [10:45] Caelum: ./get-deps.sh does that [10:45] but I think we can get the vendor tarball done [10:47] is it standard for go to use real tabs? I don't know much about go yet [10:51] Caelum: I think this is what the default "go fmt" formatting does [10:52] ah [11:03] zyga: also you said in the original email, which I only got now for some reason, to remove snap.mount and snap-repair units, should I do that? [11:07] yes [11:15] PR snapd#4932 opened: RFC: Content-interface: add rule so slot can access shared directory using plug's namespace [11:15] zyga: jdstrand: when you have a moment, I'd appreciate your opinions on this: https://github.com/snapcore/snapd/pull/4932 [11:15] PR #4932: RFC: Content-interface: add rule so slot can access shared directory using plug's namespace [11:16] greyback: ack, I'll read this and get back to you [11:16] thank you [11:18] Caelum: so, I know how to make the vendor tarball now, I will make one for 2.31.2 now [11:18] Caelum: and then we can look at 2.32.1 later today [11:18] 2.32 has a small known issue that we must address first [11:20] Caelum: I will let you know when it is available and I will test your package and approve it in the end [11:20] Caelum: perhaps with small tweaks if needed [11:27] zyga: awesome thank you! [11:39] PR snapd#4933 opened: tests: copy or sanity check core users using usernames [11:40] pstolowski: ^ [11:43] * cachio afk [11:55] pedronis: thanks [11:57] +1 === pstolowski is now known as pstolowski|afk [12:04] * zyga curses at backspace-goes-back-in-history in firefox [12:04] and erases *all* form input [12:10] PR snapd#4933 closed: tests: copy or sanity check core users using usernames === pstolowski|afk is now known as pstolowski|lunch === zarcade_droid is now known as Guest76762 === nessita_ is now known as nessita [12:18] pedronis: hey [12:18] pedronis: mborzecki figured out why we have the garbage filenames in snap-confine profile === matteo| is now known as matteo [12:19] pedronis: I think you will need to review that change, it's not big by any standards but it's curious [12:20] so far it's looking good, i'll start another spread run and open a PR [12:22] mborzecki: then we need to rebase/merge the vulcan PR so that it passes [12:24] ok, given that the standup has moved, I will have walk now but can review after the standup [12:24] pedronis: sounds good, thank you [12:26] PR snapd#4934 opened: cmd/snapd: make sure signal handlers are established during early daemon startup [12:26] pedronis: ^^ [12:28] zyga: go to "about:config", search for backspace, and change to 2. Disable backspace=back in FF [12:29] greyback: +1 on your PR [12:29] I added it to 2.32 [12:29] we will do a 2.32.x release soon [12:30] zyga: cool thanks [12:30] * greyback waves https://github.com/snapcore/snapd/pull/4545 seductively [12:30] PR #4545: interfaces/x11: allow X11 slot implementations [12:32] greyback: for 2.33 I'm afraid [12:32] zyga: fair enough [12:34] greyback: can you check if I answered all the open questions that are addressed at me [12:36] zyga: you did, thanks [12:44] thank you jdstrand [12:44] jdstrand: thanks [12:45] np [12:45] zyga: #4939 first, once merged rebase #4929 on top and we should be good, right? [12:45] PR #4929: cmd/snap-confine: apparmor: allow creating prefix path for gl/vulkan [12:45] and hi [12:45] yes [12:45] i meant #4934 [12:45] PR #4934: cmd/snapd: make sure signal handlers are established during early daemon startup [12:46] * kalikiana lunch [12:46] mborzecki: I noticed that there was a 'conflicting profile for snap-confine.XXXXXX. I restarted the test, but it seems like there are temporary files still on disk when the test runs... [12:46] jdstrand: the PR mborzecki just mentioned (4934) fixes that [12:47] ok cool [12:47] hence the rebase dance [12:47] nice [12:51] PR snapcraft#2010 closed: Release/2.40+18.04 [12:53] pedronis: do you think we should include the fix for //deleted mounts for 2.32.1 [12:53] pedronis: it was just test code but could save us some issues in autopkgtest === Son_Goku is now known as Conan_Kudo === Conan_Kudo is now known as Son_Goku [13:00] zyga: yes [13:01] I squashed thinking it might need backporting [13:01] pedronis: thank you [13:21] Issue snapcraft#2025 opened: Proper error on held packages [13:30] Issue snapcraft#2026 opened: Implement a passthrough feature [13:33] Issue snapcraft#2027 opened: Properly crawl tree of remote parts [13:33] Issue snapcraft#2028 opened: When asking to release to a branch that's too long, a traceback is printed that gives no hints as to the source of the error [13:36] Issue snapcraft#2029 opened: Proper error for apt hash sum mismatches [13:36] Issue snapcraft#2030 opened: Properly invalidate the cache if sources change [13:39] Issue snapcraft#2031 opened: Inconsistent results when building, cleaning stage and rebuilding [13:42] Issue snapcraft#2032 opened: Can not use NodeJS 8 [13:42] Issue snapcraft#2033 opened: LP: #1752344 [13:48] mborzecki: I have a fix for the atomic file thing [13:48] zyga: pr? [13:48] Issue snapcraft#1682 closed: Refactor grammar to be compatible with Python 2 [13:48] Issue snapcraft#1683 closed: Make grammar processing available in pypi [13:48] Issue snapcraft#1684 closed: Vendor grammar logic into snapcraft [13:50] PR snapd#4935 opened: tests: disentangle etc vs extrausers in core tests (2.32) [13:51] Issue snapcraft#1708 closed: Implement new design for travis CI UX [13:51] mborzecki: i saw you mention you'd update snapd in arch recently. On my manjaro system I'm only seeing 2.30. Do I need to do some voodoo to get updated? [13:52] popey: build snapd from aur https://aur.archlinux.org/packages/snapd [13:53] mborzecki: is that the 'recommend' way going forward to get the latest version? [13:54] Issue snapcraft#1670 closed: Change file migrations such that files move from stage to prime rather than coming from the part each time [13:57] Issue snapcraft#1690 closed: Design quilt support [13:57] Issue snapcraft#1691 closed: Implement quilt design [13:57] Issue snapcraft#1692 closed: snapcraft quilt tutorial [14:00] popey: yes, at least on arch [14:00] popey: manjaro maintains a set of prebuilt packages from aur, but i have no clue when they get updated [14:01] mborzecki: ok, what's the command to install it? Because that's not the way we have documented how to install snapd on arch.. https://docs.snapcraft.io/core/install-arch-linux [14:01] ah okay [14:01] maybe we need a guide for vanilla arch and one for manjaro? [14:01] oh duh, we have one for manjaro [14:01] https://docs.snapcraft.io/core/install-manjaro [14:01] thanks mborzecki [14:03] zyga: so #4934 is timing out... [14:03] PR #4934: cmd/snapd: make sure signal handlers are established during early daemon startup === pstolowski|lunch is now known as pstolowski [14:07] mborzecki: fun [14:07] mborzecki: I need to break for a moment now but I will propose a fix for the atomic issue when I'm back [14:07] zyga: yeah, fun fact, passing when i run it from spread directly (using google backend obviously) [14:25] zyga: i'll update tests/main/snap-confine to check rules vulkan [14:25]