=== KindTwo is now known as KindOne [05:23] morning [06:17] mvo: hey [06:21] hey mborzecki [06:24] mvo: can you take a look at https://github.com/snapcore/snapd/pull/8686 ? [06:24] PR #8686: o/devicestate: cleanup system actions supported by recover mode [06:24] mborzecki: sure [06:28] mvo: thanks! [06:30] PR snapd#8686 closed: o/devicestate: cleanup system actions supported by recover mode [06:30] yay [07:00] morning [07:01] zyga: hi, could you complete the comment #8123? is the only thing stopping it from landing [07:01] PR #8123: interfaces/network-control: bring /var/lib/dhcp from host (approach b) [07:01] Good morning [07:01] pedronis: certainly, in a moment === pedronis_ is now known as pedronis [07:10] mvo: should we land #8682 and change the pattern after? or do you prefer it changed first? [07:10] PR #8682: tests: port interfaces-password-manager-service to session-tool [07:11] Smak conflict there, lets resolve it [07:14] pedronis: either way is fine [07:14] pedronis: let's land as it will make tests better [07:14] landed [07:15] PR snapd#8682 closed: tests: port interfaces-password-manager-service to session-tool [07:16] anyway there's the other one that needs some tweaks [07:16] still [07:27] mborzecki: hi, I re-reviewed #8689 [07:27] PR #8689: o/devicestate: raise conflict when requesting system action while seeding [07:27] pedronis: thanks, i'm merging master and will address those in the next batch of patches [07:33] mborzecki: zyga: the fedora failure here is interesting: https://github.com/snapcore/snapd/pull/8692 [07:33] I will push a new test-snapd-with-configure{,-core18} to edge, should not affect anyting but if you notice anything strange please let me know and I can revert [07:33] PR #8692: configcore: add nomanagers buildtag for conditional build [07:34] looking [07:34] mborzecki: zyga: the cleanup is confusing snap.rootfs_* for a snap [07:34] a bit unclear why there's one around but never the less [07:44] hmmm [07:44] the snap.rootfs_ pattern is unusual to persist [07:44] something has failed earlier and left it there [07:44] btw. https://github.com/actions/cache/issues/208 is fixed [07:45] mvo: ^^ [07:45] oooo [07:45] good :) [07:45] we should be able to revive the issue with caches [07:45] s/issue/PR/ [07:45] 2020-05-20T06:18:51.4079959Z `-/tmp/snap.rootfs_9l42OX /dev/sda1[/tmp/snap.rootfs_9l42OX] ext4 rw,relatime,seclabel shared [07:45] mborzecki: holly cow [07:45] mborzecki: let me update my PR [07:47] yeah, this is curious [07:49] I can only say this [07:50] at some point snap-confine -> snap-* startup sequence failed [07:50] this is the only possibility of a leftover /tmp/snap.rootfs_ mount point [07:50] I have two simple ideas: we can add an invariant check to detect those and fail early [07:50] probably a good idea in general [07:51] pstolowski: mvo: I think #8692 can be either force merged or it needs another run [07:51] and we can also change the temporary directory name to contain the snap name [07:51] PR #8692: configcore: add nomanagers buildtag for conditional build [07:51] but probably not so useful because we use test-snapd-sh a lot so it may be less revealing than a failure at a specific test [07:52] I'll finish with conflicts and then look at this [07:52] pedronis: I can force merge, let me look at the failures [07:54] PR snapd#8692 closed: configcore: add nomanagers buildtag for conditional build [07:56] mvo: thanks [07:56] for the first time I resolved a mount-ns test conflict manually [07:57] pstolowski: there are some comments in #8697 [07:57] PR #8697: [RFC] packaging: build cmd/snap with nomanagers tag [07:58] pedronis: yes, i've been adressing them (almost done), except for your last comment [08:00] I resurrected 8508 let's see how that goes [08:01] pstolowski: thx [08:03] PR snapd#8631 closed: image,cmd/snap,tests: add support for store-wide cohort keys [08:25] hmmm [08:25] mup: ping [08:25] zyga: Roses are red, violets are blue, and I don't understand what you just said. [08:25] so mup responds [08:25] is github down? [08:26] no... [08:26] weird [08:26] I cannot git-push [08:28] dns works [08:28] wth? [08:28] mborzecki: can you git push anything to gh? [08:29] or maybe even git fetch [08:29] zyga: yup, but had some problems fetching/pushing yesterday afternoon too [08:29] I cannot add comments either [08:30] * zyga reboots [08:32] rebooted [08:32] git push hangs [08:32] ssh hangs on [08:32] debug1: Connecting to github.com [140.82.118.4] port 22. [08:33] what does github.com resolve on your machine mborzecki ? [08:33] zyga: 140.82.114.3 atm [08:33] hmmm [08:33] same result [08:33] sigh [08:33] what a morning [08:35] PR snapd#8702 opened: overlord/configstate: add log-control option [08:42] gah [08:42] any ideas [08:50] it works :) [08:52] zyga: 8578 looks ready - does it need jdstrand first or can I merge? [08:52] checking [08:52] zyga: hrm, I guess it does need him? [08:52] yeah, I think he wanted to see it [09:06] mvo: should we turn all the ubuntu-20.04-* in our tests task.yaml to ubuntu-2* ? [09:11] brb [09:16] pedronis: I think so [09:16] pedronis: not sure if zyga replied to my suggestion yet, I don't see a downside [09:16] (but maybe I'm missing something?) [09:16] mmm, mvo the one on tests to use 2* [09:17] I didn't but I think it's a good idea [09:17] zyga: cool, then we are in agreement [09:17] but perhaps as a separate pass to do everything in one go [09:17] sure [09:22] pedronis: i've updated https://github.com/snapcore/snapd/pull/8689 also with some additional checks for non-uc20 [09:22] PR #8689: o/devicestate: raise conflict when requesting system action while seeding [09:23] pedronis: i suppose it's ok to assume that there's no explicit system mode in non-uc20 [09:26] mborzecki: various methods will say we are in run mode? [09:27] zyga: mvo: yes, as a separate pass, there's a bunch, also there are core20 nested tests that I think really want 20.04 precisely [09:27] pedronis: mmm, indeed, I think a simple guideline is that if a test is precise and targets one thing it can stay that way, if it targets subsequent LTSes we can expand that to include 2* [09:37] pedronis: it's in the contxt of system requests on uc16/18, we don't support that and might as well error out early in such case [09:37] mborzecki: yes, I'm actually looking at that code [09:37] we don't do that anywhere else [09:37] I'm not sure it's a good check [09:38] mborzecki: which reminds me that the code should use m.SystemMode() and not systemMode much more [09:38] pedronis: it will be blocked later on anyway, so it's ok to drop that [09:40] or maybe we should change things to just set systemMode to run [09:40] and make SystemMode a straight accessor [09:41] pedronis: yes, it's a bit confusing, otoh knowing when it's != "" gives us a little more insight as it's set only based on modeenv [09:42] yes, what something has to give, either we refuse to operate on modeenv without mode [09:42] early [09:42] or something [09:42] right is all a bit not well defined [09:45] mborzecki: we never write modeenvs without mode, right? [09:46] pedronis: i don't think we do, but looking at the modeenv code there's no sanity check for Mode != "" [09:47] and tests seem to disagree now that i added such check, heh [09:49] fortunately it's just the test mocks [09:51] mborzecki: yea, I think we want the read in boot to fail if no mode [09:51] an deal with it [09:51] mborzecki: can be a follow up [09:59] mvo: thanks for the new test, I added some comments in #8351, I see the newline issue only in the full diff view [09:59] PR #8351: config: apply vitality-hint immediately when the config changes [10:01] pedronis: aha, thanks! I misread your comment about the gadget service, I will add that too [10:09] #8689 needs 2nd reviews [10:09] PR #8689: o/devicestate: raise conflict when requesting system action while seeding [10:20] mborzecki: #8661 is unblocked now, can you do another pass? (I will look myself but later) [10:20] PR #8661: configcore: add "service.console-conf.disable" config option === ergiusenss is now known as sergiusens [10:45] PR snapd#8703 opened: tests: detect signs of crashed snap-confine [11:05] PR snapd#8123 closed: interfaces/network-control: bring /var/lib/dhcp from host (approach b) [11:20] PR snapcraft#3132 closed: go v1 plugin: warn when using go.mod and go < 1.13 [11:44] PR snapcraft#3133 opened: go v1 plugin: go.mod support for 1.11 and 1.12 [11:49] pedronis: thanks for merging === alan_g is now known as alan_g_ [12:11] cachio: I pushed a fix for PR 8700 but we need to check with zyga if the latest spread is in use [12:11] PR #8700: tests: fix the issue where all the tests were executed on secboot system [12:11] cmatsuoka: thanks, I'll check that now [12:11] cmatsuoka, checking [12:13] cmatsuoka, I see the test is failing with [12:13] panic: cannot seal the encryption key: cannot seal and store encryption key: cannot add EFI secure boot policy profile: cannot compute secure boot policy digests: the current boot was performed with secure boot disabled in firmware [12:13] updating spread [12:13] cachio: yes, it should work after zyga updates spread [12:14] cmatsuoka: that's an example of the errors that we need to think now to imrpove [12:14] *how [12:14] our our defense it's much better than random errors from software [12:15] but we do need to think about how to make a pattern for such errors that is easier to understand [12:15] yes, at least in this one most levels add some info (except the first two that feals redundant) [12:15] the one I saw the other day was worse [12:16] that's true [12:16] at least it's not a backtrace :) [12:16] pedronis: in this case the first three are ours and we can clean them up, I'll check that [12:16] cmatsuoka: ah, interesting [12:25] PR snapd#8704 opened: cmd/snap-preseed: improve mountpoint checks of the preseeded chroot (1/2) [12:27] pedronis: the tpm-sealing-in-secboot PR already gets rid of the second level [12:33] cmatsuoka: done [12:33] zyga: thanks! [12:34] I restarted the tests, let's see what happens... [12:38] zyga: I was playing with the pi4 here, and interestingly it keeps the cpu frequency at the maximum value even idle and with the ondemand governor [12:38] cmatsuoka: on ubuntu or raspbian? [12:38] changing it to conservative made it drop the frequency [12:38] it's the current core20 beta [12:38] I see [12:39] feels like a bug [12:39] how do I check the freq? [12:39] I can check core18 on mine [12:39] zyga: yes, please do, I'll check classic too [12:40] hmm, interesting, switching back to ondemand now keeps the freq low [12:41] not sure if that's the right measurement [12:41] zyga@pi4-1:/sys/devices/system/cpu/cpufreq/policy0$ sudo cat cpuinfo_cur_freq [12:41] 1500000 [12:41] root@ubuntu:/home/cmatsuoka# cat /sys/devices/system/cpu/cpufreq/policy0/cpuinfo_cur_freq [12:41] 600000 [12:42] I'm on "ondemand" [12:42] perhaps a kernel bug? [12:42] try to switch to powersave and go back to ondemand [12:43] cmatsuoka: raspbian on a pi0w has the same behavior [12:43] 1GHz [12:43] humm interesting [12:43] I don't have another pi4 to confirm raspbian semantics there [12:43] I have a large heatsink with two fans so it's not a problem [12:43] but perhaps it does save some power [12:44] I could connect it to my power supply and measure later if you are interested === KindTwo is now known as KindOne [12:45] I'm using a small heatsink with no fans so it tends to heat up a little bit [12:46] zyga: yeah, measuring could be interesting -- at least we'll know how well cpufreq works power-wise [12:46] cmatsuoka: I need to look how to do that, I have a hacked micro cable but not a hacked usb-c cable === techalchemy_ is now known as techalchemy [13:25] stgraber: I think lxc file push needs confuses --mode [13:25] an octal input is doing funny thing [13:28] jdstrand: some low hanging fruit: https://github.com/snapcore/snapd/pull/8578 [13:28] PR #8578: interfaces: add system-packages-doc interface [13:29] zyga: yes, I'm pivoting to PR reviews later today once I finish the store reviews which are almost done [13:29] super, thank you :) [13:29] it's not urgent [13:29] just low-hanging [13:29] * jdstrand nods [13:37] stgraber: s/needs// -- i meant to say that it seems it is confused by what --mode= is supposed to be, giving octal values does funny stuff [13:38] What's the difference between #snappy and #snapcraft? [13:39] zyga: with our without leading 0? [13:39] without [13:39] I think decimal mode is quite unusual so I didn't pass it [13:41] court_jester: snappy is where most snapd developers talk [13:41] court_jester, they kind of overlap ... but the fcus here is snapd development, Ubuntu Core development and such [13:41] court_jester: snapcraft is where the snapcraft developer hang and where loads of people ask questions [13:41] while the snapcraft channel is for developing snapcraft but also of packaging etc [13:41] snapd is the package manager, snapcraft is the package builder [13:42] s/of/for/ [13:42] mvo: should I drop the bitcacher now? [13:42] take 8 of them and make it a bytecacher ;) [13:44] zyga: yes please [13:44] over the parallel port [13:44] mvo: ack [13:45] done [13:45] ta [14:02] gah [14:02] I need to reboot again [14:02] suspend during the standup nuked X input [14:04] re [14:05] cmatsuoka: https://www.raspberrypi.org/forums/viewtopic.php?f=63&t=274595&sid=05bfd24c66e0458db63482ae20121427 [14:05] ^ also a case for that eeprom update we talked about [14:05] very tempted to buy a pi4 for testing this [14:05] ah intersting [14:06] pair pi4 with a SSD [14:06] also case for ubuntu core *installer* that you use to write from sd to other storage [14:07] I only wish pi4 firmware was open source [14:13] short break, bbiab [14:13] * ogra recommends https://www.amazon.co.uk/dp/B078SVRH4B/ref=twister_B07XTSVFS4?_encoding=UTF8&psc=1 for the pi4 [14:13] it flies with writable on the SSD (and system-boot on SD) [14:15] nice! [14:15] ogra: thanks! [14:15] oh, and usb-c too [14:16] I personally would prefer something that I can fuse into one big enclosure, so that it's not a tangled mess of wires [14:16] but I think that's very hard with pi design [14:18] well, you might find cases to print on thingverse [14:20] yeah, that's true [14:21] I wish WD kept making those pi-specific usb drives [14:21] just updated to SSDs now [14:21] I still have the original set and it is amazing, apart from being a hdd [14:21] (and the notion of native USB-HDD is fuN) [14:23] i wish the pi foundation would add an M.2 PCIe socket to the next pi4 interation :) [14:24] that saves all case issues by simply mounting it to the board itself [14:24] yeah [14:25] with pi4 having pci-e anyway, I think it's doable [14:25] and could be something to put on that bottom side :) [14:25] i really love the nanopc-t4 for that ... but still having driver issues (it occasionally dies even with BSP kernel snap) [14:25] (but if it doesnt die ... OMG it is fast !) [14:25] :) [14:27] "tangled mess of wires" is something I'm getting increasingly familiar with [14:27] zyga: 2020-05-20T12:53:27.3457005Z invariant-tool: system is corrupted - in https://github.com/snapcore/snapd/pull/8703/checks?check_run_id=692478727 maybe that is helpful? [14:28] PR #8703: tests: detect signs of crashed snap-confine [14:29] mvo: yes! [14:29] looking [14:29] fantastic [14:29] thank you [14:30] I think it's a red herring here as old snap-confine did not clean some things up [14:30] but it surely works [14:30] and I wonder why it only failed here [14:30] ah, that's the PR introducing it [14:30] :D [14:30] I'll adjust the test [14:30] thanks === benfrancis6 is now known as benfrancis [14:45] cmatsuoka, https://paste.ubuntu.com/p/WQ7ZcDKmD5/ [14:45] I see thie error now [14:45] I'll try to update the fedora image [14:45] this error is strange [14:46] ah, I think you're trying to create the instance with shieldedInstanceConfig empty [14:47] does it make sense? [14:50] cmatsuoka, I am just sending the parameters list empty when secboot is not enabled [14:50] as I do with all the systems [14:50] hmm [14:50] 2020-05-20 14:49:17 Cannot allocate google:amazon-linux-2-64: cannot allocate new Google server for amazon-linux-2-64: invalid value for field 'resource.shieldedInstanceConfig': ''. Shielded VM Config can only be set when using a UEFI-compatible disk [14:50] I think the thing we did did not do the thing we hoped it does [14:50] cmatsuoka: ^ [14:50] does this indicate the spread patch is buggy [14:50] cachio: and only fedora didn't like it? [14:51] zyga, mmm, also amazon-linux [14:51] this is from upstream spread now failing on our tests [14:51] zyga, yes [14:51] it feels like we need to revert something [14:51] semms to be a gce thing [14:51] I'll break for lunch and let you figure it out, I can happily deploy updated spread in minutes [14:51] it seems the setting only works if the image is using UEFI [14:51] which is sensible [14:51] spread patch was not tested properly [14:52] I know the simple fix for spread , but perhaps I can solve it from gce [14:52] zyga, agree, I just validates that in some systems [14:55] re [14:56] cachio, cmatsuoka: please get a patch reviewed and ping me for deployment [14:57] zyga: ack [14:57] zyga, I'll update the images first [15:00] zyga, I think the main problem is how we create the images, but seems to be also a problem in gce, because it doesn't like the configuration parameter even if it goes empty [15:00] so for fedora we send the parameter empty and gce does not like the parameter [15:01] I need to fix that on spread [15:01] zyga, but first I'll update our images to support uefi configurations [15:01] cachio: perhaps fixing spread is faster [15:01] and that is a trivial change [15:01] there's no telling what happens if we move to uefi boot [15:02] maybe something else breaks [15:02] zyga, we are not moving to uefi oot [15:02] cachio: I changed your code a little bit and I'll run a test here [15:02] so what does the image update entail? [15:03] zyga, the change I am doing is telling to gce that the image now is compatible with uefi, but it does not mean we are going to use uefi [15:04] zyga, will just allow us to send uefi parameters [15:04] while the instance is created [15:04] so are there two knobs in GCE: 1) uefi supported 2) really use uefi to boot? [15:04] yes [15:04] weird :) [15:04] ok [15:04] uefi suported you set when you create the image [15:05] boot with uefi you set it when you start the image [15:05] so you can choose [15:05] either if you want to boot with uefi or not [15:05] the same image [15:06] lets make a try with this change, and then make the update in spread [15:11] sounds good [15:13] brb [15:14] mborzecki: I'll update opensuse today [15:14] mborzecki: maybe next week we could chat about suse security's group idea [15:19] zyga, fedora 31 is updated and now boots well [15:19] I'll run the entire suite to validate there is not side efect [15:20] zyga, udpating amazon linux now [15:42] ta [15:44] cachio: wdyt about something like https://pastebin.ubuntu.com/p/vtNg3Nx5Gk/ ? [15:45] mborzecki: probably for tomorrow 8508 is ready but sadly all spread tests worked so we can't test the re-run in a meaningful way [15:45] mborzecki: still ready for review [15:45] cachio: it also addresses some of gustavos comments on your original patch [15:47]