=== JoseeAntonioR is now known as jose [04:41] alesage: didn't this fix the import issues for you? [04:41] https://code.launchpad.net/~canonical-platform-qa/ubuntu-sanity-tests/remove_tests_import/+merge/248817 [04:42] it should have. === ePierre_ is now known as ePierre === chihchun is now known as chihchun_afk === vrruiz_ is now known as rvr === chihchun_afk is now known as chihchun === jpds is now known as Guest49936 === pgraner-afk is now known as pgraner [14:14] Hello everybody. === chihchun is now known as chihchun_afk [14:28] hi elopio [14:28] * dobey wonders what the "password" is for ssh into the failed adt-run vm [14:29] derp [14:29] ubuntu duh === chihchun_afk is now known as chihchun [14:49] dobey: hello. [14:50] yay progress [14:51] dobey: yay, I've just read your email [14:51] * dobey hopes those tests run correctly this time [14:51] also, i should probably set up an archive mirror on my server [14:52] I have no idea why I get only the build dir. But it seems you got it right. I will just ask you to put on the README the intructions to get the vm and to run the tests there. [14:52] sure [14:52] dobey: what pitti has recommended me is to use an apt cache. [14:52] I used to have squid-deb-proxy, now I need it again. [14:52] elopio: yeah [14:53] elopio: but i have this: /dev/md200 3.6T 88G 3.4T 3% /srv [14:53] might as well use the space :) [14:53] apt-cacher-ng FTW :) [14:53] :D [14:53] pitti: have you tried apt-cacher-ng when running the tests in the phone? [14:54] elopio: no, I didn't; you can't reconfigure apt on a phone, after all (as /etc is r/o) [14:55] pitti: I tried making it read/write, installing squid-deb-proxy, and then back to read only. It didn't work. [14:55] elopio: yeah, that can't work at all [14:56] elopio: at most, you can add an apt.conf.d/ snippet to use apt-cacher-ng or squid-deb-proxy or whatever from your *host* [14:56] and set up ssh port forwarding [14:56] so that you forward your host's port 3142 to the phone [14:56] I can try that. It's too slow for me, it takes like 20 minutes to set up the phone, and 10 running the tests. [14:57] $ cat /etc/apt/apt.conf.d/01proxy [14:57] Acquire::http::Proxy "http://127.0.0.1:3142"; [14:57] i. e. put something like this on the phone [14:58] and then ssh -R 3142:localhost:3142 phablet.. [14:58] (not sure if phablet-shell has an option to add ssh options, i. e. do port forwarding through adb) [14:58] elopio: it takes 20 mins to download the test deps? [14:59] it doesn't directly, but you can hack the script [14:59] pitti: yes. And sometimes it fails when it's about to end. That's the joy of living here. [15:23] elopio, functionally speaking https://code.launchpad.net/~canonical-platform-qa/ubuntu-sanity-tests/skip_sim_pin_if_nosim/+merge/248920 is ready [15:23] elopio, please review when you can [15:23] alesage, fgimenez ^ [15:25] brendand: on it. [15:26] fgimenez: I have built the selenium package in the ppa with debuild, but it doesn't contain the debian/patches. [15:26] can you give me a hand with that? [15:27] elopio: of course if i can, where is it? [15:28] fgimenez: https://launchpad.net/~canonical-platform-qa/+archive/ubuntu/selenium/+packages [15:29] elopio: ok thanks, i'll ping you back [15:30] fgimenez: I've made this a task inside the oxide card. === qwebirc76277 is now known as slickymasterWork [15:33] pitti: do you have any recommendations for putting a package from a ppa in the test bed? [15:33] elopio: hi. I tried to run autopilot3 with latest image. But seems a lot of error. http://paste.ubuntu.com/10159979/ [15:33] so far, we are copying the package into our branch, because we have only one dependency. That might be better than adding the ppa to apt sources, because that will require apt-get update. [15:34] elopio: Is it because the image and the code I'm using is outdated? [15:35] paulliu: if it's the latest image, it has unity from trunk, right? I see that trunk has the emulators module. [15:36] you could be missing some dependencies. Did you install unity8-autopilot? [15:36] elopio: yes. But I don't know why it loads failed. [15:36] elopio: yes. I've installed it. [15:37] elopio: you can do soemthing like --setup-commands 'add-apt-repository -y ppa:user/repo' -U [15:37] paulliu: phablet-shell, and then open the python3 interpreter. [15:38] try from there: from unity8.shell import emulators [15:38] elopio: but on a touch image that's of course only for a local test (nothing for production) [15:39] elopio: ah. error. [15:39] elopio: let me check. [15:41] pitti: and that -U brings us back to having a system that's different from the published image. [15:41] I guess the right solution is to update selenium on the archive. [15:42] fgimenez: fyi, we have selenium in a ppa because of https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=770232 [15:42] Debian bug 770232 in python-selenium "Please remove the non-free files and upload to main" [Serious,Open] [15:44] elopio: well, yes (but that's already done by adding a PPA) [15:44] elopio: ah ok thanks [15:45] pitti: right, but the ppa only has selenium, which doesn't affect the system and is used only for testing. It might be nice to be able to install packages from a ppa with the read-only partition. [15:46] elopio: so a complete thing might be: [15:46] elopio: --setup-commands 'mount -o remount,rw /; add-apt-repository -y ppa:user/repo' -U --setup-commands 'sync; sleep 2; mount -o remount,ro /' [15:47] what if instead of adding the repository, we wget the deb and dpkg -i ? [15:48] that setup will look uglier, but doesn't need the -U. [15:48] elopio: well, then just replace the -U with an apt-get install -y selenium [15:49] pitti: doesn't it require and apt-get udate? [15:50] elopio: ah yes, you have to do that [15:50] and that can cause later the read-only upgrade of things that don't work that way, like unity. [15:51] I think that coping selenium into our code base while we update it in the archive is not so bad. Will discuss during today's meeting. [15:51] s/coping/copying. [15:52] elopio: as long as you remount-ro after that, it shouldn't be too bad [15:52] but it's quite brittle of course [15:52] apt-get often fails, remounting r/o sometimes fails with "busy" (the sync+sleep seems to help, but obviously not guaranteed) [15:54] elopio: YAY! got some passing tests now :D [15:57] woohoo. [15:59] still some failures, but the app starts up correctly now [16:01] dobey: in addition to the weird launching and the ubuntu-touch-session script, did you have to do something else? [16:02] it would be really sweet if mir had a virtual frame buffer we could use, and we could have qemu images built with the ubuntu-touch seed and a mirvfb that was used as the display system in it [16:02] elopio: there's still a couple things that need to be done i think [16:02] but i'm not sure exactly what yet [16:03] https://bugs.launchpad.net/mir/+bug/1239872 [16:03] Launchpad bug 1239872 in Mir "[enhancement] Add a virtual framebuffer mode/driver for testing" [Medium,Triaged] [16:03] for qemu touch images, I suppose we can request that to ci. [16:03] why? [16:04] ci isn't using qemu for running the tests [16:04] and the qemu images are easy enough to create locally (well, i created mine locally, but i don't recall exactly how i did it right now) [16:04] 1:04:41.269 INFO testresult:44 - OK: pay_ui.tests.test_pay_ui.PayUITestCase.test_add_credit_card_completed [16:05] whee [16:05] thouse pasting that line made irssi go a bit nuts [16:07] something weird is happening [16:07] there are 7 test_foo methods in the test_pay_ui.py, but trv on the subunit output file only shows two tests [16:11] i think autopilot itself is crashing [16:12] http://pastebin.ubuntu.com/10160415/ [16:12] elopio, pitti: ^^ that trace seems to be coming from within autopilot itself, and not my tests [16:13] yeah, I've seen this occasionally too [16:13] python-testtools or whatever [16:13] any idea how to fix it? i think it's preventing autopilot from completing the whole test suite [16:14] last time I've seen it was from bug 1417966, but when I fixed that it just went away [16:14] bug 1417966 in autopkgtest (Ubuntu) "adt-run does not work with autopilot on emulator" [High,Fix committed] https://launchpad.net/bugs/1417966 [16:15] alesage: https://code.launchpad.net/~paulliu/unity8/notification_helper/+merge/249211 [16:15] but I didn't investigate that actual crash (and no cookies for using something as legacy as latin1 by default..) [16:16] paulliu, I'll have a look at thx [16:17] pitti: is the locale not being set properly in the emulator perhaps? [16:17] dobey: adt-run defaults to C.UTF-8, and services in the emulator will probably have en_US.UTF-8 or so [16:18] dobey: for sure nothing uses latin1 (ubuntu never supported non-UTF8 locales in any release) [16:18] pitti: until very recently the default locale was unset, which led to it being something like C.ANSI-something [16:19] pitti: we hit that plenty in builds on launchpad with the u1 packages === roadmr is now known as roadmr_afk [16:29] dobey: that's a crash when trying to append the failed results. It's reported in testtools, I think. [16:30] but your test has an error too, before that crash. [16:42] elopio: yes, but it's only running 2/7 tests [16:55] ok, so --setup-commands ro-apt is unusable [16:56] dobey: pitti: it's reported here: https://github.com/testing-cabal/testtools/issues/111 [16:57] ah, thanks === roadmr_afk is now known as roadmr [17:03] is there any way we can work around that somehow? [17:11] bah [17:11] i don't remember how exactly i created the qemu images [17:14] dobey: I don't know. Would have to dig into testtools' details. If you want to see the results of the rest of the tests, I would just skip the one causing the crash. [17:14] and, well, there's also the option to fix it, of course :) [17:14] elopio: i don't think there is a specific test causing that [17:15] dobey: if no tests fail, autoplot won't go into that code branch. It won't try to append the failed details, which is what causes that exception. [17:15] elopio: what seems to be happening is that sometimes a couple of the tests time out, and autopilot crashes before the test finishes, and it doesn't run the tests after it. [17:15] elopio: well i can't just skip all the tests :) [17:16] and i don't know why the tets are timing out [17:16] dobey: I prefer to run them one by one, either skipping all except one, or passing the test id to the run command. [17:17] elopio: the problem then is if they all pass when run individually, it doesn't help when running them all at the same time, so i like to avoid creating such situations :) [17:19] dobey: yeah, it's your choice for which method sucks less :) [17:19] * dobey wonders what the easiest way to skip tests is [17:19] @skip('Foo') ? [17:19] dobey: Error: "skip('Foo')" is not a valid command. [17:20] I also find comfort on those situations thinking about the 100 tests in system settings. [17:20] @unittest.skip('Foo') [17:20] elopio: Error: "unittest.skip('Foo')" is not a valid command. [17:20] weird. [17:20] @makemeasandwich [17:20] alesage: Error: "makemeasandwich" is not a valid command. [17:20] self.skipTest('foo') inside the test or the setup. [17:21] i thought the decorators were available in python now [17:21] yeah, you can use both. [17:22] and not just twisted or such [17:31] ok, the decorator doesn't work [17:54] and need to find a solution for running the tests on a device/emulator/etc running with mir [18:21] who's the best person to chat at about touch image testing? got a nexus 4 at long last! [18:24] wxl, awesome to hear it [18:24] flash the devel-proposed channel and go break stuff! [18:24] balloons: yeah, so i'm anxious to get going with testing and then eventually move on to development [18:25] balloons: so it's all exploratory testing? i note there are testcases on the iso tracker [18:25] there is usually something specific to test as well, if you sub to ubuntu-phone [18:25] yeah i've already done that so i guess we'll see what happens there [18:27] weird; I removed those testcases [18:27] balloons: well there's no linked images :) [18:31] wxl, there isn't supposed to be [18:31] it's not managed there [18:32] hence my surprise the daily builds were showing up; I removed them.. odd [18:32] * wxl shrugs [18:32] wxl, you can look at https://wiki.ubuntu.com/QATeam/TouchTesting and https://wiki.ubuntu.com/Touch/Testing [18:33] balloons: thanks! i didn't notice that second one so that's super helpful [18:35] why oh why is this timing out [18:42] ah, of course, i know why it's timing out :( [18:47] awful credentials === elfy is now known as hobgoblin [19:23] hmmm, maybe not :-/ === hobgoblin is now known as elfy === chihchun is now known as chihchun_afk [19:56] Letozaf_, hello! [19:57] doug5, hello howzit ? [19:57] balloons, hey! [19:58] Letozaf_, good, you?!? [19:59] dobey, ah... I have a bad cold, have an headache ... but no so bad :P [20:00] oops doug5 not dobey sorry :P [20:00] :D [20:02] ok, so the buy button isn't being clicked :( [20:07] Letozaf_, you're working too much :) [20:08] dobey, naaah! when you like something you never get enough :P [20:08] correct :) [20:09] oopss it's the cold I keep on getting dobey instead of doug5 LOL === roadmr is now known as roadmr_afk [20:25] howdy Letozaf_ ! back from the trip? [20:26] doug5, hello to you as well [20:26] balloons, yes :D [20:26] balloons, howzit ? is your wrist fine now? [20:27] Letozaf_, yep totally fine, pretty much 100% at this point [20:27] balloons, so you're brand new :P [20:27] Letozaf_, something like that hah! [20:28] :) [20:28] so . . . how's the phone? Better when you aren't testing it? [20:28] balloons, the phone is great, very responsive and fast, better than the Nexus 4 [20:28] balloons, well I also like testing with it :D [20:29] balloons, you kind of get bored if you do "normal" things with it, have to hack to have fun :D [20:30] balloons, Ubuntu Phones are for hackers :D otherwise you would get "normal" ones, this is special :D [20:30] Letozaf_, very true [20:33] balloons, hello! [20:33] balloons, you know what? once you get used to the Ubuntu Phone if you go back to Android you feel weired and uncomfortable, you wish to go back to Ubuntu Phone asap [20:34] balloons, :P [20:36] elopio: any idea how to figure out why autopilot is clicking the mouse in the wrong location? [20:37] dobey: are you clicking in a scope? [20:37] elopio: no it's the pay ui app [20:38] dobey: I was trying to click the uninstall button and found out that I was trying to click a button that was out of view, I had selected the wrong one. [20:38] dobey: are you using click_object ? [20:39] let me see [20:39] elopio: yes [20:39] item = select_single(); click_object(item) [20:40] the button is definitely not out of view. it's in the failed screen shot [20:40] dobey: things that come to mind: selecting the wrong object, or that the button is hidden by something else that's catching the mouse click. [20:41] but i think the screen shot might be scaled down and i can't figure out how to get it out so i can open it in gimp and find the exact coordinates where it was clicked [20:42] dobey: corey did a package for that. Let me see... [20:43] dobey: https://github.com/cgoldberg/subunitdetails [20:44] oh, hrmm, the coordinates are correct it seems [20:45] so if the coordinates are correct, where is the mouse click going :( [20:48] elopio: any idea how to debug why the click isn't happening if the coords are correct? [20:48] dobey, sounds alot like a timing issue [20:49] as a quick check, add a sleep before grabbing the object and clicking it. When it works you'll have confirmed it :-) [20:49] i already tried adding a sleep before the call and it didn't help [20:50] when you grab and click an object that *looks* like it's on screen the object properties often are incorrect [20:50] i guess i'll add one between finding it and clicking it and see if it helps [20:51] dobey, it's certainly possible it's something else, I just know timing definitely causes what appears to be autopilot clicking in the wrong place. The next thing I would try is to print the object and see what the properties are (and see if they line up with where the object actual is) [20:52] balloons: yes, i already verified the coordinates are correct from the object tree. i was just confused about what the values for the rect were actually referring to [20:52] (some rectangle implementations use two corners instead of width/height for example) [20:53] dobey, you can also manually move the mouse and click the coords as you read them from the object [20:53] my guess is it will click in the same "wrong" spot :-) [20:56] it clicks in the correct spot [20:57] and adding a sleep didn't help [20:58] Letozaf_, so I missed the discussion around file manager.. need something? [20:58] also doug5 did you get connected on the scientific calc tests? [20:58] it's too bad i can't just do button.activate() [20:59] dobey, so it clicks the right spot if you read it manually? If so, we can check the ap source for determining click object.. let's see [20:59] balloons, yes I cannot find out why tests fail on this mp: https://code.launchpad.net/~ajalkane/ubuntu-filemanager-app/sdcard-less-is-more/+merge/248444 [20:59] that is one thing that was nice about gtk+/a11y stuff [20:59] balloons: i don't know what you mean by reading it manually [20:59] dobey: no, that's precisely the idea. You have to go through the display and the input services to do the click with autopilot. [20:59] dobey: do you have any other clicks working? could it just be that your qemu is not receiving the input? [21:00] balloons, looks like the problem is with patching home as it happens only when launching tests [21:00] balloons: i printed the tree and checked the globalRect: value, and calculated the center of that rectangle, and it's the exact value where the log says the mouse is clicking [21:00] balloons, not if you launch filemanager on desktop alone without tests [21:00] balloons, look at this http://91.189.93.70:8080/job/generic-mediumtests-vivid/964/artifact/filemanager.tests.test_places.PlacesTestCase.test_go_to_root_must_open_the_root_directory.ogv [21:00] dobey, yes it uses globalRect first, then center_x, center_y, and finally the x, y, w, h [21:00] elopio: i doubt that's it. before trying to tap that button, it selects a text field and enters a value in it with no problem [21:01] balloons, anyway I am still trying to find out the reason of this problem [21:01] dobey, I mean if you issue a pointer.move(x, y), then pointer.click() === roadmr_afk is now known as roadmr [21:01] with x, y based on the properties you read [21:02] I assume that's what you mean when you said you got it to work [21:02] Letozaf_, ohh, I see the labels are blank. That the issue? [21:03] i never said i got it to work. i said the values were correct based on the data available :) [21:03] dobey, ahh.. I would play with move and move_to_object then [21:04] balloons: seems like that gnome-sys-mon testcase is a new copy of yesterday's [21:04] balloons, yes [21:05] * dobey turns up some basement jaxx [21:05] instead of just changing the original - added one to get rid of lubuntu references [21:05] balloons, tests fails because of that as there is no Object name 'Standard' and properties {'objectName': 'placeDevice'}. [21:05] balloons, for instance [21:05] elfy, yesterays? [21:06] https://code.launchpad.net/~saqman2060/ubuntu-manual-tests/gnome-system-monitor_on_Lubuntu/+merge/248990 === pgraner is now known as pgraner-afk [21:06] sorry i haven't got to that :( [21:06] that's where the _lubuntu one is coming from - just grabbed his branch - has both files :) [21:10] ok, doing a manual click doesn't help either [21:11] elfy, ahh [21:11] wxl, ideally we like things to be agnostic when they are common across flavors [21:12] wxl, but please add your comments to the review :-) it'll be nice to have more than just elfy and me bantering hehe [21:12] wxl: and no worries I watch for things on the manual side [21:12] balloons: understood. makes it possible to start testing metapackages (e.g. lubuntu-desktop) in an agnostic fashion sometimes in the future. [21:12] dobey, right, I assume it clicks the exact same location [21:12] okie dokie [21:13] which means the coords are not correct or something else is amiss [21:13] balloons, also this mp tests fail for the same problem: https://code.launchpad.net/~nikwen/ubuntu-filemanager-app/compressed-tars/+merge/249250 [21:13] wxl, well things specific to lubuntu should assume lubuntu, but i don't think gnome-system-monitor counts in that regard [21:14] balloons: well, that too! [21:14] wxl: for a lot of the specific xubuntu things - you'll see them in a folder of their own on the branch [21:14] are the screenshots that get crammed into the subunit data, and which cause the testtools crash, getting scaled down when attached? [21:15] what *IS* flavor specific? anyone can install anything from the repos [21:15] aaah yes - but then that's not default and hardly likely to end up being tested [21:16] wxl, right we try and focus on default stuff.. So pcmanfm would be lubuntu only for instance [21:16] even so - we still tried as much as possible to make them agnostic [21:16] so lxsession would be an example of something lubuntu-specific [21:16] right right [21:16] right [21:16] * dobey thinks it would be easier to build a budget flow bench and controller than to do what he's doing [21:16] dobey, are you using trv to read subunit files? it's really nice [21:17] and yes, you should save the subunit output to a file [21:17] don't console out it [21:17] yes [21:17] wxl: but even though pcmanfm is only used by you - if the testcase is agnostic - then anyone else can use it if they needed to [21:17] but the screen grabs look scaled in that [21:17] makes sense elfy [21:18] makes it simpler to write as well.. and in case the UI changes, heh, means you are changing silly steps in the test [21:18] *aren't changing [21:18] if it says FAMILY instead of lubuntu or xubuntu - then it looks right [21:18] elfy: can you link me to the xubuntu testcase folder if you have it handy? [21:18] else i can go look it up [21:18] http://bazaar.launchpad.net/~ubuntu-testcase/ubuntu-manual-tests/trunk/files/head:/testcases/packages/Xfce/ [21:18] Letozaf_, that all makes sense. I assume trunk still passes though right? [21:18] thx [21:20] balloons, let me check just to be sure, but I think so [21:20] * balloons pulls [21:24] * Letozaf_ is running trunk tests [21:25] so Letozaf_ if you wish, you could try it without patching home, but instead making a temp file inside of it.. Or run the test on the device (no patching there) [21:25] there, nice comment left :) [21:26] balloons, I am seeing the same error running trunk tests, I see the folders without labels, running tests on my desktop [21:28] ok, using a different set of coordinates didn't help either :-/ [21:29] balloons, I will try as you said, that is, running the tests on the device but anyway on desktop I will need to try without patching if I want to fix them [21:29] Letozaf_, ok so it's pre-existing and that makes sense. I suppose it wasn't an issue in the past [21:29] whee, and no screen shot this time either :-/ [21:30] Letozaf_, what I would do is remove the patching. Then change things so you change directory to a new temporary folder you create before each test. Does that make sense? [21:30] it was setup this way at one point, but moved back [21:30] hrmm, but a couple other tests tap on a hyperlink label, and they pass just fine :-/ [21:31] balloons, ok makes sense I will try what you say [21:31] Letozaf_, although actually, I'm wondering how trunk is failing too. I'll have jenkins run it [21:32] here's a trunk run; http://91.189.93.70:8080/job/ubuntu-filemanager-app-ci/453/ [21:33] balloons, even if I saw the directories without labels this is the result of running trunk tests on sandbox on my desktop: [21:33] Ran 42 tests in 522.855s [21:33] FAILED (failures=1) [21:33] I: autopilot tests done [21:33] XIO: fatal IO error 11 (Resource temporarily unavailable) on X server ":5.0" [21:33] after 8483 requests (8483 known processed) with 0 events remaining. [21:35] so why isn't this working :( [21:43] balloons, mmm ... on Jenkins tests passed [21:45] balloons, oops no... wrong build [21:47] balloons, yes it passed, now it has finished [21:48] balloons, I will re-run trunk on my PC and check better [21:50] Letozaf_, it just makes me suspicious of new code, since trunk does pass [21:50] dobey, sounds like it's not an autopilot thing based on what you said [21:51] balloons: what could it be then? it works fine when i run it on my phone and tap on the screen with my finger :-/ [21:52] balloons, I will check better, anyway I am also curios to find out why tests fail on my desktop :-P [21:52] balloons, I will be back tomorrow, it's getting late now [21:52] balloons, 'night [21:52] Letozaf_, sure.. sounds like some fun hacking is in your future [21:52] ciao Letozaf_ [21:53] dobey, if click_object and move, click both click the same place, and according to you the object's properties are correct and it's clicking in the right place, then ??? something with qemu? [21:54] balloons: well other things get clicked ok, so i doubt it's qemu [21:59] and changing it to click the cancelButton works [21:59] so something else is definitely wrong [22:01] either an invisible gremlin is stealing the event, or autopilot is getting the coordinates wrong somehow [22:01] dobey, fair enough.. you are seeing what is going on, not I so ;-) I guess my point is, it sounds like autopilot is clicking exactly where it's supposed to click. Why the object is lying to you about where it' slocated is ? [22:01] well it's clicking exactly where it thinks it is supposed to click [22:01] dobey, ohh.. so you aren't sure it's in the wrong place? if it's an event stealing thing, that too could be interesting [22:02] well, i'm sure it's the right place, given the data from the print_tree() [22:02] but i am not sure the data is correct, i guess [22:02] since clicking on all these other things works fine [22:03] right.. so every other object reports correctly, just not this one. I guess I'm confused by your statements of event stealing [22:05] that does make more sense.. in other words the click isn't captured. you could check that by having it click more than once [22:10] it waasn't a statement [22:10] it was a suggestion of something that is possible [22:10] i just can't imagine what would be stealing the event [22:13] apps can do that.. there was an app I worked on at one point that did it [22:13] I think it was a UITK bug actually.. indeed it was [22:14] a special scenario the app created, which caused the first click after opening the page to get eaten [22:14] still since it works everywhere but in qemu, well :-) [22:16] eh? it works in qemu for all the other buttons in this qml [22:18] dobey, the answer is obvious. Autopilot hates you :-) [22:19] you should join #ubuntu-autopilot and tell veebers to remove the #don't work for dobey code [22:19] * veebers perks up on his name mentioned [22:20] btw veebers things are still moving along on the AP import. I'm working on the quality section;probably have something to show you tomorrow [22:20] balloons: awesome news ^_^, you're not currently waiting on anything from me are you? [22:20] just the autopilot release :-) did it go in, I didn't check [22:20] I saw it went to proposed [22:20] balloons: also, let me know if there is anything I can do to help in this current convo [22:21] balloons: oh rats, I forgot to put it on the backlog *blush* I'll work on that, but it might be delayed (sorry) [22:21] veebers, dobey has historically had horrible times with ap, and is experiencing some interesting behavoir atm [22:21] I'll let you two talk if you wish.. [22:21] * balloons goes to walk the dog [22:21] veebers, no worries, just keep in touch on it landing! [22:21] balloons: will do [22:22] dobey: without knowing the whole story, I would suggest ensuring you're selecting the correct object that has the correct globalrect (and not some sub-support object for instance) [22:23] also, when using the -v (and -vv) flags autopilot will report the coords of where it is clicking, which might be useful double checking it's happening in the right place [22:23] and finally, ensure that the object that you are attempting to click is ready for the input (and not currently moving for instance (and the coords would be wrong)) [22:24] ^_^ [22:26] http://pastebin.ubuntu.com/10164717/ [22:26] that is the code that does the select and tap of the object [22:27] and yes, the coordinates are correct. just checked again what they are in relation to the button which does get clicked, and is immediately next to this button [22:27] both buttons are same width/height, and 16px between them [22:39] veebers: the big problem is that i have no idea how to even debug this [22:40] dobey: pretty much the first thing I do to debug issues like this is to put a break point in the code and poke around once it's paused [22:41] (e.g. if I do the tap a 2nd will it work, if I wait for a second will it work) [22:41] veebers: i am not sure i can do that here. i'm running the tests under adt-run with qemu [22:42] so i don't think a break in pdb will result in adt-run spawning sshd and letting me access the vm [22:42] no, adding a sleep, or calling it multiple times, doesn't work [22:42] it's like the event is not making it to the button for some reason [22:44] dobey: ah right, so when I do that I scp the code onto the device and run it manually there [22:44] that does sound odd, are you able to tap the button yourself during the test? [22:44] no [22:45] * dobey wonders if it's something to do with screen resolution [22:46] dobey: as in, when the test is running and when the test attempts to tap the button, if you manually tap it for it does the button work? [22:46] veebers: you mean physically, or by calling pointer.click() at the specified location? [22:47] dobey: sorry yes I meant manually with your own finger :-) [22:47] calling pointer.move() then pointer.click() didn't work [22:47] well it works as installed on the phone. i can't do that with a virtual X display inside a qemu [22:47] where the test is running [22:48] and running tests on a phone or in the emulator, is very non-trivial right now, because this isn't a normal app [22:49] dobey: ah right (sorry I'm not completely up to play on how your running it etc.) [22:50] dobey: Can you see the screen when the test is running? [22:51] yeah, it's the pay-ui which is a very special thing, so can't ubuntu-app-launch it, and can't start random processes that connect to mir [22:51] i cannot [22:52] i can see the attached failed screenshot in the subunit data when it fails, is all [22:54] dobey: Ah I see. Yeah, the screenshot is what I was going to suggest taking a look at [22:55] yeah the screenshot isn't helpful [22:55] i can see the button is there :) [23:04] dobey: heh right :-P Is there a cursor on screen? Also, might be worth trying the video recording argument (I really don't know if that will work or not with the qemu/virtual X, but worth a try) [23:05] cursor isn't visible, but that's normal for screenshot to not show cursor [23:06] and i don't think there is a reasonable way to pass additional arguments to autopilot through adt-run [23:08] dobey: hmm, I'm a little stumped. It would be ideal to be able to run the test locally or on a device as it gives us a bunch more debugging options [23:11] well on a device is probably not going to be doable until we can have a solution to how to run the code under both mir and x without too much trouble [23:11] ugh, now the downloading the debs bit is being slow for me [23:55] oh well, way past time to have gone [23:55] later