[15:34] <balloons> pitti, can I get you to have a quick look at the conversion to autopkg for reminders: https://code.launchpad.net/~nskaggs/reminders-app/autopkgtest/+merge/226894. I'm confused by the error I'm getting
[15:48] <jibel> balloons, do you have the full output of the run in debug mode?
[15:48] <jibel> and do you still have the same error after your fix?
[15:50] <balloons> jibel, yes and yes
[15:50] <balloons> jibel, http://paste.ubuntu.com/7804047/
[16:01] <jibel> balloons, what is 'reminders' in your command lines 'adt-run -d reminders --click=com.ubuntu.reminders --- ssh -s /usr/share/autopkgtest/ssh-setup/adb' ?
[16:01] <jibel> balloons, is it a local checkout of your branch containing the tests?
[16:02] <jibel> balloons, what I did and worked: bzr branch lp:~nskaggs/reminders-app/autopkgtest && adt-run -d autopkgtest/ --click=com.ubuntu.reminders -s --- ssh -s /usr/share/autopkgtest/ssh-setup/adb
[16:03] <jibel> well, then the tests fail with 'ImportError: No module named 'requests_oauthlib'' but it's another story
[16:03] <balloons> jibel; ok that's good news I guess
[16:03] <balloons> well, except for the fact the tests still fail of course
[16:04] <balloons> jibel, trying again..
[16:04] <pitti> balloons: re
[16:04] <balloons> I wiped and rebranched
[16:04] <pitti> balloons: did you guys already figure it out?
[16:05] <pitti> balloons: so the branch LGTM
[16:05] <pitti> balloons: it may be that "reminders" is interpreted as an apt source name instead of a click path
[16:05] <balloons> pitti, well one part is figured out.. jibel is saying the tests still fail due to a depends.. I'm trying to reproduce
[16:06] <pitti> balloons: the / at the end (or any path before it) could make that differnce
[16:06] <balloons> and I'm getting apt repo errors; 404 Failed to fetch
[16:06] <pitti> balloons: incidentally in a branch of mine I have a better heuristics
[16:06] <pitti> balloons: probably missing apt-get update / too old image?
[16:07] <balloons> do you not apt-get update first?
[16:07] <jibel> balloons, I ran the test on 134
[16:07] <balloons> I'm on 132.. it's not that old :-) I'm updating now
[16:07] <pitti> balloons: utopi changes quickly, and thus apt indexes get out of date fast
[16:07] <jibel> it's terribly outdated ;)
[16:08] <balloons> pitti, yes I can understand that.. but it could be a problem even on the newest image
[16:09] <balloons> anyways, jibel python3-requests-oauthlib is listed as a depends, but you are saying it's not found when running. Do you see it being downloaded?
[16:09] <jibel> pitti, for click package does adt-run resolve the dependencies from the click package installed on the device or the local branch?
[16:09] <balloons> In mine, I don't see any of the python3 stuff being pulled
[16:09] <jibel> balloons, no it is not downloaded hence my question
[16:09] <balloons> this feels like my original issue.. I wonder if I specify a dependency that is already included first, if you are skipping them or ?
[16:10] <pitti> balloons: yes, I think I need to invent a "local user" apt-get update with temp indexes, if there is no root
[16:11] <pitti> balloons: did you actually install the .click with the updated deps?
[16:11] <balloons> yea, I'm only seeing adt-run: DBG: testbed reset: modified=False, deps_installed=[], deps_new=['autopilot-touch', 'ubuntu-ui-toolkit-autopilot']
[16:11] <balloons> adt-run: DBG: install_deps: deps_new=['autopilot-touch', 'ubuntu-ui-toolkit-autopilot'], recommends=False
[16:11] <pitti> jibel: the deps come from the .click; there is no general way to get them from the click source
[16:11] <balloons> whoa.. from the click? the app click?
[16:12] <balloons> the tests and the click have different depends
[16:12] <pitti> balloons: yes, as only that has a reliably formatted manifest
[16:12] <jibel> balloons, so you need to build the click from your source first then from with that click package
[16:12] <pitti> balloons: click sources often build it from an .in file, and the manifest template can be anywhere and not even a valid format
[16:12] <pitti> correct
[16:12] <pitti> for debs, the deb source is the primary object, as that has formally specified ways to get built and specify metadata
[16:13] <jibel> s/from/run
[16:13] <pitti> for clicks, only the click binary has that
[16:14] <jibel> pitti, would it be possible to specify a local manifest.json so the click package doesn't have to be rebuilt?
[16:14] <balloons> I'm slighty confused.. The application binary has some depends, which building the click will pull and package. However, the test specifically has other depends which the app does not have. How does building a local click help?
[16:14] <pitti> jibel: not right now, but I can make --override-control work for click manifests
[16:15] <pitti> balloons: because the manifest (which adt-run looks at) *only* is in the binary click
[16:15] <pitti> balloons: if you test the installed click, it will use the manifest from the installed click too (i. e. not your deps)
[16:15] <pitti> click sources are really underspecified :(
[16:15] <pitti> that's why adt-run doesn't support building them or taking manifests
[16:16] <balloons> ahh, ok.. that's... wild
[16:16] <pitti> --override-control sounds ok though, if you specify a syntactically correct manifest
[16:16] <balloons> makes sense.. the manifests often have buildtime vars in them as well
[16:16] <balloons> so passing one directly will fail
[16:17] <pitti> balloons: right, and there's also no canonical place where to put them
[16:17] <pitti> the build system can do whatever it wants
[16:18] <balloons> yep, sorry I was slow to pick up on what the issue was
[16:18] <jibel> balloons, correct for example ubuntu-clock-app has a click/manifest.json.in while reminders has manifest.json and both contains variables that are populated at build time
[16:18] <pitti> balloons: no worries; for now, just build the click locally and toss that to adt-run
[16:19] <pitti> wishlist bug appreciated for specifying a local manifest
[16:19] <balloons> it only matters in this case because things changed.. most of the time I should be able to get away with it
[16:19] <balloons> re-using the installed click that is
[16:19] <pitti> still, a local override seems useful for quickly checking a test with an extra restriction or dep or so, without having to rebuild
[16:20] <balloons> yea, you can craft a manifest yourself
[16:21] <balloons> ok, so fingers crossed then once this builds it will all work :-)
[16:24] <jibel> pitti, bug 1342818
[16:25] <pitti> jibel: merci
[16:25] <jibel> de rien
[16:25] <balloons> pitti, before you take off.. let's talk notes-app as well.. is it still causing you issues?
[16:26] <pitti> balloons: I haven't re-run them recently
[16:27]  * pitti got sidetracked with a few other urgent requests from folks
[16:27] <balloons> pitti, same same.. just remembering them
[16:27] <pitti> what was that giant meta-bug again?
[16:27] <pitti> awesome bar search fail
[16:37] <balloons> pitti, looks like it's still only grabbing those 2 depends :-(
[16:38]  * balloons investigates further
[16:39] <pitti> balloons: did you give it the local click now?
[16:39] <pitti> balloons: i. e. /your/path/foo.click, not "com.ubuntu.notes-app:
[16:39] <pitti> s/:/"
[16:39] <balloons> yea, just confirming it's installed, and checking the mainfest
[16:39] <balloons> pitti, ohh, I pushed it to the device
[16:40] <pitti> balloons: and installed with --all-users? that ought to work too
[16:40] <pitti> (that's more or less what adt-run does)
[16:40] <balloons> pitti, ahh that's it.. it's only for phablet
[16:40]  * balloons confirms root still has old version
[16:40] <pitti> balloons: the test runs as phablet too, so it's a bit strange that it doesn't use that version then
[16:41] <balloons> pitti, well let me try by passing the click as an arg
[16:42] <balloons> yep, it sees the depends
[16:43] <balloons> so I'm not sure if there's something we want to change here. It could be confusing, as most folks install clicks for phablet, not for root
[16:43] <pitti> balloons: oh, that could be the bug -- it runs "click info" as root, not as user
[16:43] <pitti> balloons: bug report appreciated
[16:43] <balloons> k, I will file :-)
[16:43] <pitti> balloons: I haven't tested this mode indeed
[16:43]  * pitti ^5s his favourite beta tester for autopkgtest :)
[16:44] <balloons> I'm just happy to have such a tool, and an upstream who can fix bugs :-)
[16:46] <pitti> bonsoir tout le monde ! c'est l'heure de courier
[16:46] <balloons> ciao pitti
[16:46] <pitti> and I'm just done replying to all IRC pings, so I better toddle off now :)
[16:46] <jibel> it's too hot to run now, I'll go in 1 hour
[16:52] <balloons> k, 2 bugs filed :-) https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/1342836 https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/1342838
[17:11] <balloons> jibel, so after everything, something weird is happening trying to run the reminders tests. It's just stuck on 'Tests running' and has been for 15 mins now. Nothing is happening on the device either
[17:12] <jibel> balloons, I'll try. How do I build a click package of reminders?
[18:09] <balloons> jibel, I built in a chroot
[18:09] <balloons> armhf chroot