[01:30] how does snapcraft determine the version number from a git repository? my version number is way off [02:06] koala_man: it doesn't by default. You can set the version to "git" and it'll use the commit hash as the version string. Otherwise it uses whatever you set in the yaml. You can also use a version-script to code the rules for your application's source [02:07] Caelum: doesn't `go build` work? [02:27] diddledan: ok.. so where does v0.3.0 come from in https://build.snapcraft.io/user/koalaman/shellcheck/177997 ? [02:30] weird, it looks like it's picked up v0.3.0 from git via a tag somewhere, but then appended a commit hash on it (right at the bottom of the log): [02:30] https://www.irccloud.com/pastebin/sftr0qcu/ [02:31] yes, but why did it pick 0.3.0 and not e.g. the latest 0.4.7 or the earliest 0.1.0? [02:31] that I'm not sure about [02:32] perhaps the commit history on master doesn't match any of those other tags? [02:33] so 0.3.0 is the latest tag that is in sync with master [02:33] in sync - I mean has a shared most recent commit in the tag with a commit that exists in master [02:34] might have to wait till the european daytime to get some more concrete answers from sergiusens when he awakens [02:34] or kyrofa if he's still awake? [02:39] I'm not entirely sure what you mean but `git log master` definitely shows other tags === verterok` is now known as verterok [05:13] morning [06:19] PR snapd#4908 closed: [RFC] cmd/snap-confine: attempt to detect if multiarch host uses arch triplets === chihchun_afk is now known as chihchun [06:57] good morning [06:57] mborzecki: I see that yesterday evening was fruitful [06:58] zyga: hey, yes [06:58] can you please prepare a backport for 2.32 as well [06:59] yup, already have it, checked it on ubuntu too [07:00] great, what's the PR for that? [07:00] we could prepare another release today [07:00] PR snapd#4952 opened: [2.32] cmd/snap-confine: attempt to detect if multiarch host uses arch triplets [07:01] the media test is failing on fedora [07:01] looks like a real bug [07:02] zyga: might need the same treatment as media-sharing test on fedora [07:02] zyga: idk if you recall, that's /media vs /run/media [07:02] + test-snapd-removable-media.sh -c 'ls /media/testdir' [07:02] ls: cannot access '/media/testdir': No such file or directory [07:02] yes, I know exactly [07:02] I had to code that difference [07:02] :/ [07:02] ok, looking at what's there [07:03] I _really_ wish fedora shipped a /media -> /run/media symlink [07:03] but ... well [07:03] i fixed media-sharing ~3 weeks ago, but iirc removable-media test was opened long before that [07:03] mborzecki: were there any changes to the nvidia multiarch fix since I read that [07:04] zyga: just one commit, i dropped some debug messages per niemeyer's request [07:04] great [07:04] ok, brb for coffee (I woke up waaay to late today) [07:04] zyga: oh and maybe a commit with some comments (as suggested by jdstrand) [07:04] when wife and kids don't go to school [07:04] I'll review it shortly [07:12] moin moin === pstolowski|afk is now known as pstolowski [07:21] mornings! [07:21] kalikiana: pstolowski: morning [07:23] hey kalikiana [07:37] zyga: re. the safe bind mount branch, I had some new thoughts about simplifying it but am not sure of the security concerns [07:38] zyga: what if we did a bind mount from /proc/self/fd/$source_fd to /proc/self/fd/$target_fd? [07:39] it'd avoid the need for the stash directory, and it ensures the source and destination are anchored at the expected locations [07:45] jamesh: hmm, that's interesting [07:45] did you try it? [07:45] giving it a go. [07:45] jamesh: that's pretty interesting I must say [07:45] would be awesome if it works [07:46] it'd also handle bind mounting files [07:46] since there's no more fchdir() [07:49] yes, iff this works it's pretty interesting [07:49] I wonder what happens to /proc/pid/fd in this scenario [07:49] open a directory with O_DIRECTORY (and with or without O_PATH) [07:49] inspect the fd in /proc [07:49] mv the directory around, replace with symlink or whatever [07:50] inspect the fd in proc [07:57] jamesh: I also found /proc/pid/root as useful thing but kernel folks broke it because OMG security :/ [07:57] jamesh: it would allow mounting from one ns to another [07:57] (it did until it was changed) [07:59] mborzecki: 4952 reviewed [07:59] mborzecki: question: what do we need to do to remove nvidia as compile time option entirely, and make it fully dynamic [08:01] zyga: imo we'd need to interpret data coming os-release, pick the right location based on known paths [08:02] what kind of choices would we make? can you walk me through this please? [08:04] zyga: using a simple Python example, it mounted the expected directory after a move [08:04] that's super promising! [08:05] so the mount followed the file descriptor rather than the name [08:05] jdstrand: ^^ FYI, I think jamesh found the holy grail of mounting [08:06] zyga: choosing the location where the libraries are, i.e /usr/lib/ & /usr/lib32 on arch, /usr/lib/ on debian/ubuntu, /lib64(?) on fedora (although i think this might be close to arch), historically some drivers would install under their /usr/{libdir}/nvidia{-maybe-suffix} path, that also varies from distro to distro and release to release [08:06] zyga: here's the test program I used: https://paste.ubuntu.com/p/zvpjCcmcjm/ [08:07] zyga: don't recall what suse does, it's probably close to fedora (or the reference layout of systemd distros) [08:08] mborzecki: I see, thanks [08:08] mborzecki: I'm weighting this against need for nvidia-xxx snap [08:08] and the pros/cons [08:11] zyga: I tried extending it to do the read only remount, but that fails. Presumably because dest_fd references the directory underneath the mount point rather than the mount point itself [08:11] after the mount we can open another fd [08:12] zyga: do you mind if i fix https://github.com/snapcore/snapd/pull/4952#discussion_r177972590 in master once we merge the release branch back? [08:12] PR #4952: [2.32] cmd/snap-confine: attempt to detect if multiarch host uses arch triplets [08:12] mborzecki: not at all [08:13] ok, thx [08:13] I'm looking at fedora failure now [08:13] removable media? [08:13] yes [08:13] ok [08:14] btw. just restarted the build on 4952 for the 2nd time now, first one failed fetching stuff from github/gopkg.in, now it failed fetching the snaps from store [08:15] noise][, hey, is the store operational again? [08:17] zyga: the rpm build is doing a test even if I don't have a %check section, do you know how to not do that? [08:17] what kind of test? [08:18] it's running: + /usr/lib/rpm/golang.sh test github.com/snapcore/snapd/... [08:18] ah [08:18] looks like something built into the rpm macro for golang itself [08:18] Caelum: maybe it's golang rpm stuff? [08:18] yeah [08:18] no, no idea how to turn it off, you'd have to look at expansion of the macro [08:19] ok thanks! [08:19] one question [08:19] what's the original macro in the spec file? [08:19] * zyga is rusty on that [08:24] Caelum: are you sure it's %gobuild that is doing that [08:24] I see %gotest in %check [08:26] I commented it out and it still happens [08:26] that's the thing I don't get [08:27] oh I figured it out, I need a %check section with some command [08:28] hmmm [08:28] well, not sure [08:28] maybe ask in #opensuse-buildservice, people there may be able to help more than I can [08:28] I tried to expand %gobuild with 'rpm -E' but it didn't expand [08:30] mborzecki: question about /media vs /run/media [08:30] it fails becasue snap-confine on fedora bind mounts /run/media [08:31] now my idea is as follows [08:31] inside snap execution environment we could make /media == /run/media [08:31] so [08:32] snap authors will have easier way to reason about how things work [08:33] zyga: sounds good [08:33] zyga: will both locations be avaialble? [08:34] yes [08:34] zyga: ok [08:34] yeah, I think this is easier than hacking the test [08:34] which should not depend on the host [08:34] right [08:34] media-sharing test will need an update then [08:35] no, actually it should work just the same [08:35] my point is that host's native $MEDIA_DIR just becomes both /media and /run/media inside the execution environment [08:36] in core18 we could perhaps make /media as symlink into /run/media [08:39] zyga: yes, but the test tries to figure out what host it's running on and chooses /media or /run/media, if we make those equivalent, it would make sense to update the test to either always look in /media or compare both locations (maybe the latter) [08:39] it does so on the outside [08:39] that's all fine [08:39] we'll only make things equivalent on the inside [08:40] PR snapd#4931 closed: configstate: give a chance to immediately recompute the next refresh time when schedules are set [08:40] hmm [08:40] though there may be one complication [08:40] ah, no [08:51] PR snapd#4953 opened: configstate: give a chance to immediately recompute the next refresh time when schedules are set (2.32) [08:53] zyga: does #4932 needs again a master merge or a rerun ? [08:53] PR #4932: interfaces/content: add rule so slot can access writable files at plug's mountpoint [08:53] pedronis: I think master is broken again, let me look [08:53] (I'm fixing master now) [08:53] yes [08:53] econnreset is fixed in one PR [08:53] and removable media is in progress [08:54] so don't rerun things for now please [08:55] PR snapd#4954 opened: store: Sections and WriteCatalogs need to strictly send device auth only if the device has a custom store (2.32) [08:56] ok [08:57] zyga: so I can run %gobuild userd but I have no idea what to do with it afterwards, it doesn't get installed anywhere, do you know? [08:57] %foo is a RPM macro [08:57] yeah [08:57] well it does a go install [08:57] it can be expanded with rpm -E '%foo' [08:57] it installs into $GOPATH/ [08:57] but I don't know what the macros do nowadays, they changed recently [08:57] there is a userd and userd.a [08:58] .a is an archive [08:58] yeah [08:58] but what do I do with userd, where do I put it [08:59] mborzecki: look at https://github.com/snapcore/snapd/pull/4955 please [08:59] PR #4955: cmd/snap-confine: establish equivalence of /{,run/}media [08:59] PR snapd#4955 opened: cmd/snap-confine: establish equivalence of /{,run/}media [08:59] Caelum: hmmm, hold on, what is userd? [08:59] pedronis: ^ this PR should fix master [09:00] but it cannot land alone because of econnreset bug with store being wonky [09:00] there's a directory in git called userd, I was guessing this is the userd daemon you wanted me to add [09:00] no no, that's not what I meant [09:00] the userd is part of the "snap" binary [09:00] you want to package the systemd / dbus service files [09:00] that's why I hinted at missing files compared to the debian package [09:00] another point of refence is the fedora package [09:00] you can compare the list of files there [09:01] lastly we have packaging for opensuse in packaging/opensuse-42.3 [09:01] (inside the git tree) [09:01] that last packaging is what we run tests agains [09:01] oh I see [09:01] you can use that as another reference, ideally they would be identical [09:02] (though sometimes it is not easy as test builds have some extra requirements) [09:02] but have a look at those all and see what they do differnetly [09:02] *differnetly [09:02] ok I know what to do now [09:02] let me know if you get stuck on anything [09:02] thank you :-) [09:04] I'm running a run of fedora 27 on linode now, let's see if my fix is sufficient [09:07] zyga: you are aware that one of the tests is now skipped on fedora? [09:08] if it's about that you need to unskip it [09:08] pedronis: no, I wasn't [09:08] which test was that? [09:09] zyga: https://github.com/snapcore/snapd/commit/0848363411c07c47580a6d64594911b3a5511595 [09:09] thanks === ackkk is now known as ackk [09:23] tests passed (before re-enabling) [09:23] all of main passed [09:23] running that single test after re-enabling it now [09:30] zyga: do you have a moment for quick HO? i'd like to talk about udev [09:30] yes [09:30] sure, I'll join the standup HO [09:31] pstolowski: ready [09:32] you are missing a show with my daughter who tries her best to annoy me today [09:46] zyga: currently snapd-generator is installed to /usr/lib/snapd/snapd-generator but should it be installed to /usr/lib/systemd/system-generators/snapd-generator instead? [09:46] or is that a unit file? [09:53] Caelum: one sec, on a cal [09:53] call [09:55] yeah it looks like that's what I need to do [09:56] yes [09:57] it should be where systemd looks for generators [09:57] pstolowski: a bit old but perhaps still relevant https://people.redhat.com/nhorman/papers/netlink.pdf [09:58] pstolowski: question [09:58] pstolowski: have we closed the previous socket correctly [09:58] it _feels_ like we did not [09:59] zyga: yes, I close the socket [10:02] zyga: ok, the doc you linked seems to be pretty clear about setting pid to own pid [10:11] PR snapd#4947 closed: spread: disable StartLimitInterval option on opensuse-42.3 [10:13] PR snapcraft#2042 opened: Vendoring lxd [10:14] pstolowski: https://tools.ietf.org/html/rfc3549 is also interesting [10:35] zyga: I have all the missing files now, except for the userd unit file, where do I get that? It's not in ubuntu. [10:35] Caelum: it should be made by "make install" in data/ [10:35] let me look [10:36] look at data/dbus [10:36] if you make install -C data DESTDIR=... you should get something useful out [10:36] those are the two io.snapcraft.* things there [10:37] the dbus services? I got those. [10:38] so which file is missing still? [10:38] I didn't know that's what the userd service was [10:38] ah [10:38] yeah [10:38] that's it [10:38] commit and I'll build and try it out :) [10:43] mborzecki: thinking about media [10:43] I think I will only do one thing [10:43] on /run/media systems /media will be a fixed alias [10:43] on /media systems /run/media will not [10:44] because host /run is shared and there's no guarantee that /run/media exists [10:44] and we should not attempt to create it just because you are using snaps [10:44] (especially that on the outside /run/media woul be empty) [10:44] we can rely on /media in ways we cannot rely on /run/media [10:44] what do you think? [10:46] hm, so basically, no /run/media if there was none on the host? [10:48] yes [10:48] I pushed the branch over, have a look [10:49] jamesh: hey, if still there, so do you plan to use the /proc/pid/fd idea? [10:49] jamesh: I can discuss this with jdstrand and jjohansen today [10:49] I actually totally love it [10:49] zyga: yes. I'm integrating it into my branch and updating tests [10:49] zyga: I added a note about it to the PR too [10:50] sweet, thank you! [11:00] zyga: I figured out my problem with tests, you can't comment out rpm macros with # they still get expanded and run [11:01] ah :D [11:01] yes [11:01] I ran into that too, I recall now [11:01] what's the issue with the tests? [11:01] are they failing because of DNS? [11:02] no they're fine, I was just trying to disable them while testing builds because they take a really really long time [11:02] ah [11:02] ok [11:02] but there is one test there that randomly fails like 30% of the time [11:02] when I see it again I'll let you know [11:02] zyga: i'll rebase the PR with plain make thing [11:03] mborzecki: thanks [11:03] mborzecki: I had a look already [11:03] but not in detail === chihchun is now known as chihchun_afk [11:20] zyga: aand force pushed [11:20] thanks [11:21] I want to unbreak master and then work on my stuff [11:21] but I promise to build and test your PR today [11:21] pedronis: we need a 2nd review on https://github.com/snapcore/snapd/pull/4955 [11:21] PR #4955: cmd/snap-confine: make /run/media an alias of /media [11:21] please merge if you agree === pstolowski is now known as pstolowski|lunch [11:28] it needs a review by jamie, no? [11:29] pedronis: hmmm, yes, it should get one [11:29] I'll wait for his review to merge [11:33] zyga: shall we land https://github.com/snapcore/snapd/pull/4952 ? [11:33] PR #4952: cmd/snap-confine: attempt to detect if multiarch host uses arch triplets (2.32) [11:33] hey zyga [11:34] hey [11:35] zyga, beta is it gonna change? [11:35] or I can promote it? [11:35] cachio: I think it's fine to promote [11:35] cachio: we will do .2 but that's unscheduled still [11:35] to candidate? [11:36] yes [11:36] to candidate [11:36] zyga: well I updated the package, having laptop overheating issues here on linux [11:38] Caelum: ouch, sorry to hear that [11:39] I will pull and test [11:39] mborzecki, morphis: does anyone want to do a 2nd review on opensuse package refresh? [11:40] zyga: i can take a look [11:40] obs link? [11:40] https://build.opensuse.org/request/show/592282 [11:42] I'm building locally now [11:42] cachio: +1 from me [11:43] heh reddit witch hunt https://www.reddit.com/r/linux/comments/87ssdw/why_i_use_flatpak_for_3rd_party_apps/ [11:51] Caelum: the userd service works fine now [11:52] I'll reboot my suse box and play some more [11:56] Caelum: installation of snaps without sudo also works now [11:57] I will try a few more snaps now [11:59] Caelum: 42.2 build fails [11:59] I think you can hardcode the systemd generator location [11:59] [ 176s] File must begin with "/": %{_systemdgeneratordir}/snapd-generator [12:03] Hi there, attempting to install. Here's the error that I am getting: error: cannot install "anbox-installer": classic confinement requires snaps under /snap or symlink from /snap to /var/lib/snapd/snap [12:04] wadadli: hey [12:04] wadadli: some distributions chose not to ship the /snap directory and on such distributions snaps that are using classic confinement cannot function [12:04] you can make a symlink from /snap to /var/lib/snapd/snap to fix this issue on your machine [12:05] wadadli: what version of ubuntu are you running? [12:05] Fedora [12:06] :/ [12:06] leftyfb: that's expected and supported [12:06] wadadli: just make that symlink and you will be fine [12:06] zyga: yeah, but they originally asked for help in #ubuntu [12:06] no luck despite running ln -s /var/lib/snapd/snap /snap [12:06] wadadli: note that anbox is not the full snap, just the installer [12:06] wadadli: what are you getting now? [12:07] w8, doesn't anbox require kernel modules too? [12:07] same here. [12:07] same thing** [12:07] (not to install, but to run) [12:07] mborzecki: I don't remember, I think so [12:07] PR snapd#4952 closed: cmd/snap-confine: attempt to detect if multiarch host uses arch triplets (2.32) [12:07] wadadli: same as in? [12:08] cannot install "anbox-installer": classic confinement requires snaps under /snap or symlink from /snap to /var/lib/snapd/snap [12:09] zyga: I've pushed the changes to remove the stash dir to https://github.com/snapcore/snapd/pull/4868 now [12:09] PR #4868: cmd/snap-update-ns: add secure bind mount implementation for use with user mounts [12:10] wadadli: are you using gnome software to install anbox? [12:10] nope, anyways got the installer to install. [12:11] now the installer complains that Fedora isn't supported. [12:12] wadadli: that's what I expected [12:12] wadadli: it requires some kernel modules AFAIR, morphis would know [12:12] jamesh: thank you [12:13] wadadli: the snap distributes just the installer, not the full application [12:13] Would have been fun just to try. :) [12:17] Any suggestions for running android apps on Linux? [12:18] wadadli: sorry, not a clue [12:18] maybe use ubuntu and anbox [12:18] or talk to morphis about what would it take to work on fedora [12:18] maybe the issue is just temporary [12:20] Caelum: let me know if you run into any issues [12:20] pedronis, mborzecki: trivial cleanup PR https://github.com/snapcore/snapd/pull/4956 [12:20] PR #4956: cmd/snap-update-ns: rename i to segNum [12:20] PR snapd#4956 opened: cmd/snap-update-ns: rename i to segNum [12:31] mborzecki: more thoughts on the media issue [12:31] mborzecki: remove the config option [12:32] mborzecki: if /run/media exists it is aliased as /media and bind mounted [12:32] mborzecki: otherwise /media is just bind mounted to /media and that's it [12:32] mborzecki: one less compile time choice [12:32] mborzecki: the only other one would be nvidia [12:32] mborzecki: WDYT? === pstolowski|lunch is now known as pstolowski [12:33] zyga: sounds good, but it's probably worth double checking that for instance debian does not have both /media and /run/media, each being a different thing :) [12:33] yeah, I plan to check [12:34] HMMM [12:34] iirc this depends on how udisks2 was compiled [12:34] fedora 27 doesn't have /run/media [12:34] mborzecki: yes [12:34] mborzecki: it's a bit insane and sad we have this difference [12:35] opensuse tumbleweed has /run/media [12:35] haha [12:35] try installing udisks2 in fedora [12:36] drat, /run/media is created dynamically [12:36]