16:02:49 #startmeeting 16:02:49 Meeting started Fri Jan 25 16:02:49 2013 UTC. The chair is ogra_. Information about MeetBot at http://wiki.ubuntu.com/meetingology. 16:02:49 16:02:49 Available commands: #accept #accepted #action #agree #agreed #chair #commands #endmeeting #endvote #halp #help #idea #info #link #lurk #meetingname #meetingtopic #nick #progress #rejected #replay #restrictlogs #save #startmeeting #subtopic #topic #unchair #undo #unlurk #vote #voters #votesrequired 16:03:03 hey everyonw, this is the nexus7 status meeting 16:03:10 hi 16:03:17 i'll give a quick summary of what happened this week ... 16:03:41 so thanks to jani who wrote a prototype in GO, we now have auto-rotation working ! 16:03:52 as well as the ambient light sensor 16:04:10 (i took the prototype and implemented it in shell, as an interim solution) 16:04:35 togather with the autorotation, an xorg fix landed that makes touch input in portrait mode work 16:04:51 if we are to integrate with gnome-system-daemon we need changes to the kernel so we expose an input interface as that daemon expects 16:05:05 we may get it this cycle or we may just stick with the shell daemon, not sure yet 16:05:23 right, the question is if we will keep g-s-d in this form on the later tablet setup 16:05:23 I have been looking into it the past few days though and I'd like to get it done properly 16:05:44 right, the shell daemons are really just interim solutions 16:05:53 no idea. What is the later tablet setup - related to ubuntu phone? 16:05:57 g-s-d is being discussed 16:05:57 i found it important for testing the features to have it in asap 16:06:02 we want to reduce its scope for sure 16:06:12 so we can see how the UI reacts to rotation etc 16:06:15 so if it's as easy to implement those feature in other places feel free 16:06:37 seb128, well it is certainly easier but at the cost of extra footprint that a separate daemon brings 16:06:52 something we keep ignoring and that got us into this bloated mess in the first place :) 16:06:59 right 16:07:10 well, we should perhaps implement those feature in unity itself for example 16:07:13 well, i wouldnt mind a small daemon in C 16:07:24 we were supposed to have a push for reducing memory footprint this cycle, at leats not add extra stuff if we can't make it lower :) 16:07:26 depending on the feature though 16:07:42 the issue is that the higher number of process, the higher cost in context switches, wakeups, memory usage, etc 16:07:47 janimo, as long as the new extra stuff is smaller i dont care :) 16:08:15 janimo, btw, did tvoss talk to you ? 16:08:16 this daemon is by far not among the high footprint daemons but everything adds up 16:08:20 nope 16:08:31 he seems to work on an API for sensor stuff 16:08:49 nice, I'll follow up after the meeting 16:08:55 so if there is something in the works we could use, that might help 16:09:04 i pointed him to you ... hm 16:09:11 (days ago) 16:09:27 well I discovered g-s-d covers accel and screen rotation, but only with a paritcular type of accelerometer in the kernel 16:09:32 there's no unifying interface 16:09:39 right 16:09:50 seems better to build on that codebase 16:09:53 though the question is, will we stay with g-s-d at all for sensors 16:09:54 rather than to write a new one 16:09:58 also discovered by accident while poking in sysfs that my x86 laptop has an accelrometer 16:10:05 which also does not work with g-s-d 16:10:23 ogra_, no immediate plan to replace g-s-d, and we can easily move bits of code out of gsd to some other source later if needed 16:10:32 ok 16:10:32 ogra_, it's not like the work would be wasted 16:10:40 I would improve g-s-d as a start 16:10:45 seems the less work path 16:10:49 well sensors are not that separate from settings, it makes sense imo to have light sensor integrated in backlight settings and other screen related stuff 16:10:50 right, making g-s-d actually generic deinitely helps 16:11:05 definitely 16:11:12 so from what limited knowledge I have g-s-d seems the place for it - if we plan on keeping g-s-d that is 16:11:19 right 16:11:30 we do plan to keep it in the near futur 16:11:37 i'm just careful because i heard we might not ... but what seb128 said, it helps in general 16:11:48 we might move some bits out to small "on-demand" services 16:11:55 ++ 16:12:04 once upstart can handle it :) 16:12:07 I hope those are not written in python though 16:12:10 like if "rotation" can be an upstart daemon we could span the helper from an upstart script 16:12:17 and take it down again 16:12:18 yeah 16:12:23 thats what i do atm 16:12:30 upstart daemon->signal 16:12:33 but for the session side it uses Xsession.d 16:12:36 rotation does not seem like an on-deman thing though, I expect people want it on all the time with a tablet 16:12:47 instead of upstart ... 16:12:56 well, if the rotation triggers an upstart signal we can bind a script to it 16:12:59 janimo, in android i dont :) 16:13:19 that would avoid having the code running all the time to poll for the rotation event 16:13:20 if i read a book and lie on the side in bed i dont want it rotating all the time 16:13:30 * seb128 always turn off autorotate 16:13:35 it's more annoying than useful 16:13:42 seb128, well, the polling is the biggest issue atm, yeah 16:14:05 the daemon is like 20 lines of shell :) 16:14:17 wouldnt eat much if it wouldnt have to poll every second 16:14:37 in the current state its one of the biggest CPU consumers though 16:14:49 we could fix that by doing the poll in the kernel and firing off uevents when things change 16:14:55 can we make xorg/the kernel to send a signal? 16:14:56 yeah 16:15:00 needs some coding but can be done - not sure if it's the right place though 16:15:01 kernel 16:15:29 it might be worth dropping an email on ubuntu-devel list? 16:15:31 Apparently gsd reads an ID_INPUT_ACCELEROMETER_ORIENTATION property via gudev 16:15:34 to get input from kernel people? 16:15:36 seb128, but we also need to add code to interpret the raw data and keep track of orientation in the kernel then - moving the shell code to kernel C code 16:15:53 lool, rhight, which only very few accelerometers seem to trigger 16:16:00 If I understand what you folks are saying: 1) this doesn't work on N7 2) we would like to change this from poll to event-based 16:16:03 seems a bit crazy - but this sensor does not have an orientation even itself unless loaded with proprietary firmware which we do not have 16:16:23 so on 1), do we know what to do to set this property right? 16:16:34 lool, right, but 0.5) we only have the free part of the driver 16:16:34 lool, I am looking into this now. The nexus7 sensor is dumb 16:16:43 and 2), do we have a way to trigger something in the active session based on udev events? 16:16:45 and only provides raw accel data 16:16:50 lool, which doesnt give us 100% of the features 16:17:03 I think there's an udev bridge for upstart, can we use it? 16:17:09 lool, g-s-d listens to udev events from accel devices 16:17:22 right, we just need the events 16:17:25 so that part is covered if we make our kernel side similar to what is already handled 16:17:34 yeah 16:17:36 janimo: ah so it's not poll, it's already uevent based 16:17:37 so an input device with uevents, polling 16:17:46 lool, right. 16:18:07 our current driver doesnt even provide a device 16:18:10 so we only miss a correct kernel path for this 16:18:12 when the currenlty handled (asus-laptop pega I think) accel fires off an ACPI even saying rotation changed 16:18:15 so these bits need adding 16:18:17 the driver emits an uevent 16:18:39 a udev helper is called that reads the raw data and calculated orientation - again, as uevents cannot carry metadata 16:18:44 and exports that as env -var 16:18:56 udev property I mean 16:19:01 and g-s-d reads it 16:19:13 that sounds like a really slow process 16:19:23 compared to what we do atm 16:19:41 lool, the rewason I am not sure about the kernel part is that there is no standard kernel way of exposing accels, g-s-d uses one of the existing approaches 16:19:54 i wonbder if we couldnt have that with less players involved 16:20:01 but for instance it does not work with my acer accelerometer either, that too does things a bit differently 16:20:12 * xnox realises it's meeting time. 16:20:38 ogra_, well g-s-d as our shell script could just read raw data and handle everything, but would need to have a lot of sensor specific code 16:20:43 and would need to do detection 16:20:51 udev handles detection of device at least 16:20:51 and would need to poll 16:21:23 right, i just wonder if that cant be done more direct ... udev->xorg 16:21:38 so the g-s-d stuff seems ok to me, just that restrict all sensor to the input subsystem (nexus is on IIO does not even provide an input device) 16:21:46 and have g-s-d only intercept if the user actually switches it off 16:21:56 and a specific devnode with input_struct layout as a joystick (X, Y, X values) 16:22:06 yeah 16:22:31 it is doable, I'll keep at it for a few more days 16:22:33 well, lets not go to much into detail 16:22:48 we'll see what you come up with ;) 16:22:57 xnox, want to give an update ? 16:22:59 * xnox also thought that udev is ought to be generating the events from accelorometer and then g-s-d or whatever can piggy back on top. Gives us the most flexibility to consume those events. 16:23:08 and take into account the galaxy nexus too which has a sensor from the same family but different 16:23:21 same but different ... yay 16:23:28 I'd rather we not have to add code to every sensor if upstream decides they are all going to be IIO and not input devices 16:23:29 I finished usb-creator support for flashing nexus7 tablets using .img + .bootimg files. 16:23:35 wohooo 16:23:35 so maybe changing g-s-d is better long term 16:23:44 janimo, yeah 16:23:51 Currently it only "recognises" nexus7 ID_VENDOR_ID and ID_MODEL_ID 16:24:03 thats enough for a start 16:24:12 we can add a big database over time :) 16:24:14 so if there are other devices please send me their VENDOR_IDs / MODEL_IDs 16:24:19 janimo, maybe drop a summary email to ubuntu-devel@ about the rotation/g-s-d/kernel topic? might be better for discussion 16:24:24 xnox, nexus devices you mean? 16:24:30 once we switch to nexus4 and nexus10 :) 16:24:43 seb128, ++ 16:24:44 seb128, I will follow up to the original thread once I have a bit more info 16:24:47 janimo: yeah, i do want ids of the nexus range to be in the usb-creator. 16:24:53 janimo: but I don't have them. 16:24:58 janimo, thanks 16:25:00 xnox, so galaxy nexus phone too? 16:25:02 [action] jani to mail ubuntu-devel about accelerometer plans 16:25:02 * meetingology jani to mail ubuntu-devel about accelerometer plans 16:25:20 janimo: yes please. 16:25:42 janimo: we don't have images for those devices yet, but I'd like usb-creator to know about them already. 16:26:01 (.img + .bootimg style images that is) 16:26:03 xnox, one thing I remember is that both nexus 7 and galaxy report different USB IDs depending on whether they booted in recovery mode or regular android 16:26:14 note though that the phone installation functions differently from our installation procedure 16:26:20 janimo: currently I trigger on two ID_MODEL_IDs 16:26:21 ogra, when appropriate, I have a short topic. 16:26:31 (i assume we can adjust that though) 16:26:50 kyleN, well, we dont have an agenda atm. feel free to just start when its quiet 16:27:21 i.e. now :) 16:27:24 ogra_: yeah. =/ given enough demand we can add additional types of flashing. the *.bootimg and *.img backend took only have a day to code in the end. 16:27:42 Before the holidays there was some discussion of using valgrind to find memory leaks and thereby improve Ubuntu for resource constained devices. 16:27:53 I worked on apport-valgrind, which was released into raring as part of apport. 16:28:01 nice ! 16:28:09 This allows you to run valgrind for an executable (to find memory leaks) while first populating a temporary directory with available debug symbols for the executable, which makes the resulting valgrind log file more complete and useful (because more symbols names are filled in, making the stack traces clearer) 16:28:25 Alex Chiang submitted a needed patch for this that landed in valgrind. 16:28:32 and in Ubuntu/R 16:28:42 This is a first step. There are many aspects to this. I started a blog to cover some of this: http://kylenubuntu.blogspot.com/ 16:28:58 you need to blog about that, that extends far beyond nexus7 16:28:58 I expect to add more blog posts soon on this topic. 16:29:01 DONE 16:29:01 and sounds helpful everywhere 16:29:17 time you get onto planet ! 16:29:23 I was going to say 16:29:37 nobody is going to read those if they are not on planet 16:29:37 do I need a rocket ship to get there? ;) 16:29:45 no offense to your blog popularity ;-) 16:29:47 just ubuntu membership 16:29:55 BTW, nexus 7 testers please http://people.canonical.com/~jani/linux-image-3.1.10-8-nexus7_3.1.10-8.21_armhf.deb 16:30:05 ack 16:30:11 note that this will break your touchscreen :P 16:30:15 for me this deb fixes sound on boot without having to do a suspend. I'd like other opinions too 16:30:58 well log in via ssh and install http://paste.ubuntu.com/1570120/ over /usr/bin/acceld :) 16:31:49 janimo, is that need ? i would just drop the "Rotatet "right"" from xorg.conf 16:31:51 *Rotate 16:32:16 anyway, the current nx7 looks pretty good already ... we have one big showstopper bug left though 16:32:38 bug 1068994 16:32:40 bug 1068994 in ubuntu-nexus7 "button1 gets stuck after a while" [Critical,Confirmed] https://launchpad.net/bugs/1068994 16:32:47 * ogra_ pokes the bot 16:33:09 ogra_, it needs to correlate xrandr names to xinput rotation matrix. The latter get shifted with the kernel 90 degree rotation 16:33:13 so it is needed 16:33:20 and after the fix for portrait mode i must say that i clearly suspect compiz to be at fault, not xorg 16:33:56 if the bug occurs to me, i can still hit the ubuntu key on onboard to switch apps, and i can interact withj the app content 16:34:10 what doesnt take any input are window frames, panel and launcher 16:34:19 all therr are compiz 16:34:22 *three 16:34:48 so i start to suspect compiz itself has an issue here 16:35:27 what we are also still missing is bluetooth 16:35:33 is cyphermox around ? 16:36:53 the quantal image has some hacks to push the binary firmware into the driver when BT starts ... we dont have that in raring atm (and i'm not sure we want it) 16:37:33 so it would be good to know the plans here, can we get it to work without the binary bits, if not, ho do we integrate them properly 16:37:37 *how 16:37:54 seb128, could you forward that question ? 16:38:08 ogra_, sure 16:38:10 (probably also better discussed by mail= 16:38:24 yeah, sounds like something for ubuntu-devel@ as well 16:38:32 what else did i miss ? 16:38:57 is anyone working on collecting the valgrind datas from the community? 16:38:58 do we have anyone from QA around ? 16:39:06 plars, ? 16:39:06 what about optimizing processes running? 16:39:17 oh, definitely 16:39:27 who? 16:39:40 I want to start organizing those better 16:39:49 so I want to get everybody in touch 16:40:08 how about we make a task table and people can pick from it ... 16:40:15 Gema and Paul Larson both mentioned activities related to collecting smem output in QA labs 16:40:25 we have started doing a table with lool 16:40:25 unclaimed tasks will then be forcefully assigned 16:40:28 we need to publish it 16:40:33 great 16:40:35 we have also some blueprints 16:40:51 lool, yeah, i know plars wroks on nexus stuff 16:40:59 (thus the ping) 16:41:08 ogasawara: hi, I'm here 16:41:09 i implemented preseeding support for him last week 16:41:12 grr 16:41:14 hehe 16:41:15 ogasawara: sorry 16:41:27 ogra_: yes, I'm here, but in about 30 different irc windows at the moment :) 16:41:49 plars, so anything you want to tell us about nx7 work ? 16:42:15 seb128, for collecting stuff from the community it sounds like balloons might be the right guy 16:42:26 ogra_: just that I got your email about preseeding and I am anxious to try it, but I was off yesterday and haven't gotten to it yet today 16:42:58 ogra_: assuming that works, we should be able to automate installs on the nexus in the lab and have automated tests running pretty soon 16:42:59 ok, i'm not sure it works 100% yet, please bug me if there are issues 16:43:09 ogra_, ok, I think I will try to draft a "status update/summary/how we organize a bit better going forward" email early next week 16:43:18 ogra_, I will run the draft through you, lool and some others before sending 16:43:25 ogra_: if you have an example preseed or script for munging it into the image already, that would streamline things considerably 16:43:30 seb128, great, lets turn that (or parts of it) into our meeting agenda 16:44:07 plars, heh, no, i dont, i used to abuse tobin as a preseed stoirage in the past :) 16:44:50 oh, btw, do we have anyone from the kernel team here ? 16:45:02 cking used to work on power management on the nx7 16:45:19 and it got really really bad since we switched to raring 16:45:48 ogra_, it's the same kernel though right? 16:46:11 from battery values <50% on i get a battery popup every second until the device shuts down (3-4h long)+ 16:46:29 janimo, nope, it was re-based on the later android release 16:46:37 ah right 16:46:50 i got these issues since the first raring kernel landed 16:46:59 so a quantal kernel has no such issues ? 16:47:08 hey guys. 16:47:14 could be a raring userland policy bug as well 16:47:16 ogra_, has a bug been filed so we can track that? 16:47:24 percentage and time are totally off all the time and i get weird power messages on charging or after a few hours of usage 16:47:30 cking, sure 16:47:50 janimo, yeah, could be, i also noticed that upower shows an additional usb charger 16:47:59 i dont think that was there in quantal 16:49:35 cking, bug 1093543 16:49:36 bug 1093543 in ubuntu-nexus7 "Battery status behaves erratically" [Medium,Confirmed] https://launchpad.net/bugs/1093543 16:49:43 (sorry, tootk a moment to find it) 16:50:24 ogra_, thanks, i'll look at that one 16:50:35 great 16:50:46 i think we talked about it before btw 16:50:56 but didnt find any cause 16:51:12 ok ... 16:51:23 anyone from the community around with questions etc ? 16:51:59 doesnt look like 16:52:13 doesn anyone want to bring something up before i close ? 16:52:26 going once 16:52:26 ogra_, do we know if the pointer/grab lock issue is being worked or making any progress? 16:52:37 it's pretty much a blocker if you want to use the nexus 16:52:39 seb128, tjaalton sent it upstream 16:52:49 seb128, its not as bad as you think 16:52:51 is it just sitting there or did somebody pick itup? 16:53:12 i use the nx7 all day, if you use it carefully you can get through with only having the issue once a day 16:53:17 lol 16:53:21 i think it rots upstream 16:53:22 "if you use it carefully" 16:53:24 seb128: upstream knows about it, pinged him earlier this week but no new patches to try 16:53:26 ogasawara, are you using a mouse? 16:53:30 like explain to users to not use the indicators? ;-) 16:53:33 there was quite some activity in the beginning 16:53:37 but that died 16:53:44 janimo, indeed not :) 16:54:08 tjaalton, did you try with the latest fixes, its really only compiz that freezes 16:54:11 "if you make sure not to use the touchscreen on your tablet, you're fine" 16:54:19 ogra_: hm, no 16:54:23 tjaalton, i can interact with everything else on the desktop just fine 16:54:27 oh 16:54:29 sweet 16:54:44 just not panel, launcher and *window frames* 16:54:57 lol, because who needs to use a panel or launcher? ;-) 16:55:03 but firefox content, terminal, onboard etc work fine 16:55:35 seb128, well, if i just read a long article in FF, all i need is the grab'n drag extension to work ;) 16:56:01 and switching apps works fine via onboard (_hit the ubuntu key, pick the number of your app) 16:56:37 tjaalton, so i think it would be worth taking a second look with focus on compiz 16:57:51 3 min left ... anything else ? 16:58:00 ogra_: update the lp bug and I'll ask peter again where the bug is. the previous logs we had did show that the grab state was fooked, but.. we'll see 16:58:25 tjaalton, well, there were a bunch of issues with input fixed in the last xorg upload 16:58:48 so i would suspect we might have cleared the xorg side now 16:58:52 oh the patch bryce put back in? yeah sorry about that, it was a bad merge by me :) 16:59:14 mistakenly dropped a patch that fixed the rotation 16:59:25 well, i think bryce and danien van vugt added patches 16:59:27 when merging new bugfix release 16:59:31 *daniel 16:59:43 xserver only got this one back 16:59:50 also a point release before that 16:59:54 well, todays raring works pretty well for me 17:00:03 apart from the compiz hang 17:00:21 no other input issues here 17:00:27 ok, time is up 17:00:33 going once 17:00:38 going twice 17:00:41 #endmeeting