mozilla :: #mobile

20 Apr 2017
00:39rnewmanahunt: no, everything will be fucked if we don't have a profile
00:40rnewmanand if we can't create a profile dir, either we have a bug (creating a folder that already exists? race?), or everything is fucked
00:41ahuntYeah, my testing confirmed that - other code just falls over if theres no profile (which means dependent code can be simplified / avoid null-checks/etc)
13:45WillPittengerGoogle may build ad-blocking into Chrome | PCWorld ( )
13:54blyHi. Firefox on my non-neon Android autoupdated today, and I'm now getting the message "This Firefox won't work on this device (armeabi-v7a. 25). Please download the correct version." I'm expecting to downgrade back to 52.0.2 and lock updates, or is there a more "correct version" out there?
13:55sebastianbly: bug 1305815
13:55firebot FIXED, Mechanism to exclude non-NEON versions
13:55sebastianbly: Which device are you using? I think we tried to block them from getting the update in google play?
13:56blysamsung galaxy 10.1 with a custom rom
13:56blyso it's just the message that's wrong, and there is no other version?
13:57sebastianbly: Yeah, it's a generic "device unsupported" message that is not really helpful in your case
13:57blyalright, makes sense. thanks
13:58sebastian@snorp: Did we block neon devices in google play? Should we add this device too?
13:58snorpsebastian: uh, we were supposed to, but maybe it didn't happen?
13:59snorpwtf is galaxy 10.1
13:59snorpthe tablet?
13:59snorpfrom like 2008?
13:59bly2011 I think
13:59sebastianIt might be possible that it's already blocked by android version and this case the custom rom has made it available agian :)
13:59snorpoh it's probably tegra2
13:59blythat's right
14:00snorpyeah, that's unsupported now
14:00snorpso the message is correct
14:00sebastian^-- mh, this list includes galaxy tab 10.1
14:01sebastianbut maybe the custom rom is cheating
14:01snorpcould be
14:01snorp(likely, in fact)
14:36franksnorp: did you see my comment on bug 1357786
14:36firebot FIXED, Send actual CPU architecture in core ping
14:36frankI'm concerned that the previous way we sent arch and the new way will get mangled in analysis
14:37frankwill the new strings be sufficiently different than the old that we will be able to know which is which?
14:37snorpfrank: main difference will be the presence of aarch64
14:38snorpfrank: surely we have the ability to filter on submit date or app version?
14:38snorpI just don't think it's a significant enough change to really worry about it
14:38snorpwe clearly were not using the arch value for anything useful before
14:39franksnorp: good point :) but if we want to in the future, then we want to be sure that it is the *correct* values
14:39snorpfrank: is bumping the version something we already do?
14:39frankyes, pretty regularly
14:39snorpah I see there is a version
14:39snorpok lets bump the version
14:40frankthanks :)
14:40snorpfrank: is it ok to just do that in the fennec submission code, or I need to do it everywhere else too?
14:40franksnorp: submission code + core ping docs
14:40frankserver-side I'll take care of
14:41snorpapparently fennec is already sending version 8
14:41frankyes, should be on version nine soon
14:41frankbug 1334110 is bumping to 9
14:41firebot NEW, Data review for adding campaign ID to core ping
14:41snorpfrank: core-ping.rst is what I need to edit?
14:41franksnorp: yup
14:41snorpthat one says 7
14:42snorpfrank: desktop/fennec need to send the same version?
14:42frankthere's a conflicting change in bug 1334110
14:42frankdesktop doesn't send core ping
14:42snorpah, ok
14:42* snorp races nevin to landing
14:50Aryxhi, did the system requirement for v53 change? a user with android 4.0.3 reported that they got updated (on an Asus Transformer TF101) and the bottom only shows that Firefox doesn't work with that device and that they shell download the appropriate version. where do they find the appropriate version?
14:52sebastianAryx: The toast is generic and misleading. it's likely that there's no appropriate version (if the user is hitting bug 1305815)
14:52firebot FIXED, Mechanism to exclude non-NEON versions
14:53Aryxthank you
14:54blyAryx: I uninstalled and installed the 52.0.2 apk on mine
15:00sebastianfrank: What's the difference between android_events and android_events_v1 again?
15:00franksebastian: non-versioned tables are a view on the maximum versioned table
15:01frankso in this case they are the same
15:01* frank needs to check though that there is no android_events_v2
15:01sebastianah, okay
15:05sebastianI'm a data analytics expert :D
15:10frankit is a thing of beauty
15:11frank"If picasso was a data analyst"
15:20bkellysnorp: what does it mean if I get: Failure [INSTALL_FAILED_UPDATE_INCOMPATIBLE]
15:20bkellywhen I do ./mach install
15:21snorpbkelly: you have an installed version that is signed with a different key
15:21snorpif it's a local build, just do 'adb uninstall org.mozilla.fennec_bkelly'
15:21bkellyI tried deleting it from the app launcher screen thing
15:21snorpor it could be something else I guess, but that's usually the problem
15:22bkellysnorp: yea, I think so... I switched computers
15:22snorpyeah, that's it then
15:37bkellykind of annoying you have to log in with a google account in order to get updates on android
15:41snorpyeah, I have a work gmail account for that
15:55bkellysnorp: is there any tricks to speeding up the package&&install steps?
15:56snorpbkelly: buy a faster machine
15:56bkellyI have one of the new fast machines
15:56ahuntIsnt the main bottleneck for install the app size?
15:56bkellyI guess I will just combine it with the build step
15:56ahunt(and USB size)
15:57bkellypackage&&install takes a little over a minute for me
15:57bkellyincremental build took 30 seconds
15:57nalexanderbkelly: it's hard to improve. Fennec is an outlier; it's a huge APK, with a big C++ blob.
15:57bkellyI guess there is no delta updating the app on the phone
15:58nalexanderbkelly: there is, but not with the generic tools, and not for native .so libraries.
15:58bkellyits ok, I was mainly just asking
15:58snorpfor platform work it's always going to suck
15:58snorpfor frontend stuff you can use hot patching or whatever it's called
15:58snorpdirty dalvik VM tricks
15:58nalexandersnorp: Instant Run.
15:58snorpyeah, that.
15:59nalexanderbkelly: snorp:, but I don't have good ideas (or any time)
15:59firebotBug 1291371 NEW, [geckoview] Allow to load GeckoView libraries from /mnt/sdcard or another APK for developer speed
15:59nalexanderThat's GV specific, but the same idea applies -- get libxul onto the device faster. It's easy to do badly and hard to do well.
16:00ahuntUnrelated: I noticed that newer Android versions allows loading libraries direct from apk (without prior extracting), Im guessing that might not work with our current compression approach though?
16:00ahunt(Or does that happen at install time?)
16:00snorpahunt: does it really?
16:00snorpbut yeah, our compression is a problem
16:00ahuntsnorp: chromium/webview do that
16:00* ahunt looks for the docs
16:00snorpahunt: neat
16:01bkellysnorp: so, do we know why it takes so long to foreground fennec when a notification is clicked?
16:01bkellywhen its running and just in the background (like looking at homescreen)
16:01snorpbkelly: nope
16:01snorpshould be fast
16:01bkellyits like 5+ seconds to foreground fennec on my nexus 5x running aurora (or whatever aurora is these days)
16:02snorpeven if fennec not running, you should see something immediately
16:02bkellyno other apps running
16:02snorpwe can take a look
16:02bkellysnorp: I'll file a bug
16:02snorpbkelly: thanks
16:02sebastianahunt: grisha: Android meeting anyone? :)
16:03bkellysnorp: what category do you think this would be?
16:07nalexanderbkelly: sounds similar to the issue jchen fixed (?) where Gecko event queues would starve themselves.
16:07nalexandersnorp: do you remember that one?
16:07nalexanderIt led to slow push message processing.
16:07firebotBug 1358178 NEW, fennec takes 5+ seconds to foreground app when click on push notification
16:07bkellynalexander: well, I see the push notification fine... that is not delayed
16:07bkellyI also see twitter and telegram notification clicks sometimes never foreground firefox
16:07nalexanderbkelly: it was in the tap -> decrypt in Gecko path.
16:09bkellynalexander: not sure I know what that is
16:10nalexandergrisha: do you remember that ticket ^
16:13ahuntsnorp: Im still looking for where I saw the actual docs, but heres where they make use of it:
16:13ahuntI might have imagined seeing other docs tbh
16:15bkellyI got a logcat with some interesting things... I'll put it in the bug
16:19nine_millihey jchen, do you know if your selectbox fix for geckoview has been released yet?
16:19snorpahunt: ah, I think they just do stuff similar to what we do
16:19jchennine_milli: it landed in nightly yesterday
16:19snorpcopy cats
16:20nine_milliok thanks ill try to update
16:20jchennine_milli: you need to look at geckoview_example to see how to implement the new promptdelegate in your app
16:21bkellymaybe the delay is because it thinks its already focused: W/InputMethodManagerService( 550): Window already focused, ignoring focus gain of:$Stub$Proxy@1953ffe8 attribute=null, token = android.os.BinderProxy@22eb9ffe
16:21snorpbkelly: so the log seems to indicate that it's starting gecko, but wtf?
16:21ahuntsnorp: I found the original docs I mentioned, under opening shared libraries directly from an APK -
16:21snorpit has to be running already to receive the push
16:21ahuntI have no idea what we do in fennec though
16:21jchennine_milli: actually multiple select doesn't work yet. need to land one more patch for that
16:22snorpbkelly: if you do 'adb lolcat -v time' your log will have timestamps
16:22bkellysnorp: I will do that now
16:22snorpbkelly: useful for things like seeing where delay is :)
16:22nine_milliim just using single select
16:22snorpwe may not have enough logging really
16:24firebotBug 1358178 NEW, fennec takes 5+ seconds to foreground app when click on push notification
16:25bkellyI have to run to meet someone for lunch, but will look at this more when I get back
16:25bkellyhopefully you will have it solved by then, thoguh!
16:43snorpahunt: might be worth loading libs that way for GV
17:13njparksebastian: so if you're not already working on it, I can look into adding coverall in focus repo
17:39sebastiannjpark: Okay, go ahead! :)
17:40njparksebastian: might take a bit as I need to steal your knowledge from telemetry repo :)
19:48SeburoHi. Can I have some help with a SUMO forum question? Can "multiprocess" be enabled for Fx Android using a about:config setting?
20:05nalexanderSeburo: it can be enabled, but nothing will work.
20:05nalexanderSeburo: Fennec's implementation assumes it's single process (chrome and content). snorp's team is sorta-kinda working on changing that.
20:06Seburonalexander: Thanks. We have a user that has tried and it kills Fennec to where they need to reinstall it. I think I will tell them to stop trying and to wait for it to be released in the future.
20:07nalexanderSeburo: yeah, they'll be terminally screwed. No way to go back without being able to edit the prefs file :(
20:07nalexandersnorp: kbrosnan: can you think of a way to flip a pref in a release Firefox when it's wedged? ^
20:07nalexanderShort of a text editor and root, I can't.
20:07snorpnalexander: no way AFAIK
20:08nalexandersnorp: we should probably lock that pref for Fennec release.
20:08nalexanderIt's a footgun.
20:08snorpnalexander: eh, lots of footguns in there honestly
20:08snorpthere is a warning for a reason...
20:08snorp(maybe not in fennec, I forget)
20:08nalexanderAlthough... users who twiddle about:config without knowing what they're doing should expect sore feet.
20:10SeburoFully agree. Slightly worrying that they appear to have done this multiple times on different devices with different versions of Fx....(!). Thanks for your help, appreciated as always. :-)
20:10kbrosnannalexander: snorp adb run-as (syntax that i forget) coupled with push/pull on the pefs.js file?
20:11snorpkbrosnan: that doesn't work on release builds
20:11nalexanderkbrosnan: I think run-as requires android:debuggable.
20:11nalexanderThanks, folks!
20:14ahuntnjpark: jfyi, Ive landed the PR that causes focus builds to fail if any medium/high priority issues are found by findbugs. (We still have a bunch of low priority issues, none of them look particularly interesting but I might try to tackle them eventually.)
20:15mkaplyDoes anyone out there have one of the new really high res devices?
20:15mkaply(G6 or other)
20:15njparkahunt: awesome! I've been ridden with multiple small tasks today, but tomorrow I'll take a shot at coverall
20:15mkaplyTrying to figure out if there is a problem with our audio playback icons on those devices
20:16nalexandermkaply: you can create emulators with whatever configurations you want.
20:16mkaplynalexander: Duh. Good point.
20:16njparkahunt: oh btw, when you're in taipei, if you can let them know about code coverage stuff, that'd be awesome too
20:17* mkaply wonders if the phone will be higher resolution then my computer :)
20:17ahuntmkaply: I have the N6P which apparently has a similar dpi to the G6
20:18ahunt(Hmm, is that really true? The N6P has higher DPI than the pixel XL?)
20:18mkaplyahunt: Can you check
20:18mkaplySee if the play and volume buttons are too small
20:20mkaplyso no
20:20mkaplyWe got this image
20:21mkaplyI don't know what phone it is yet though
20:21ahuntThe G6 is slightly higher dpi though (564 vs 518 for N6P)
20:22ahuntIIRC sebastian has the Pixel XL, which is aparently 534 (I got confused by the normal pixel, which is 441)
20:23ahuntnjpark: yeah, its already on the agenda (Im guessing theres not a huge amount of timezone overlap between toronto and Taipei?)
20:24njparkahunt: yeah.. I think it's 11-12 hr diff, so in b2g we used to meet at 9pm here where they were at 8am. I really wonder whether they have their own qa team, if so would like to say hi sometime
20:26njparkbut feel free to forward me an invite if it happens between 8-12 taipei time
20:34mkaplyThat was weird. Tried to install Firefox 53.0 on the Pixel XL emulator and it said it wasn't compatible with the phone? Are you not allowed to do app installs on the emulator?
20:36nalexandermkaply: you are. Are you sure you have the ARM (not x86) emulator?
20:36mkaplyI thought I did arm
20:37mkaplyNope, you called it.
20:37mkaplyThat's the default for performance
20:46nine_milliis there anything like the geckoview widget in android for the desktop versions of firefox?
20:56Gautam-Msnorp: Hi!
20:56snorpGautam-M: hello
20:58Gautam-Msnorp, So I was trying to use bindings for JNI stuff, you suggested me that to access a class like java.util.Locale I need to create a text file then reference it in mozbuild.
20:59Gautam-MAfter doing so, the build failed. I think I missed some step in between? :)
20:59snorpGautam-M: what's the error
21:00Gautam-Msnorp: 'UnsortedError: An attempt was made to add an unsorted sequence to a list. The incoming list is unsorted starting at element 3. We expected "!Locales.cpp" but got "!MediaCodec.cpp"\n'
21:00snorpGautam-M: ah, it needs to be in alphabetical order]
21:00snorpin the
21:02Gautam-Msnorp: Ah, okay. I should have noticed that. Let me try building again. Thx!
21:17esawinjchen: incoming linker r? because g.landium is on pto.
21:17snorpuh, why do we let glandium go on pto?
21:18snorptactical error
21 Apr 2017
No messages
Last message: 156 days and 18 hours ago