mozilla :: #mobile

21 Apr 2017
03:40jhaohi all, can I set gecko's pref in fennec's native code?
03:45rnewmanjhao: see PrefsHelper
03:46jhaornewman: ah got it, thanks :-)
15:02ahuntIm noticing an increasing number of captive portals (airports, hotels, etc) seem to have the same SSL issues that sumo had. Only chrome/webkit on Android complain - which means the default log in to network prompt fails on Android, whereas firefox (or any desktop browser including chrome) dont have issues
15:03ahunt(We should pitch Geckoview to the Android team?)
15:27snorpahunt: heh.
15:27snorp"upgrade to the open web"
15:27snorpnot quite obnoxious enough
15:27snorpneed a better tagline
15:28mkaplyDoes the mobile playback window for audio use videocontrols.css (same as desktop?)
15:28snorpmkaply: it's a different css file IIRC
15:28snorpactually maybe not because I remember it being a complete clusterfuck
15:29mkaplysnorp: We aren't handling high res devices with our playback buttons.
15:29snorpmkaply: bummer
15:29mkaplyWhich is every device coming out now :)
15:29snorpthey seem ok on my pixel?
15:30mkaplyPixel XL
15:30firebot NEW, when play link audio, the icon is too small
15:30snorpmkaply: oh holy shit
15:30snorpthat's terrible
15:30mkaplyI wonder if softvision is testing higher res devices
15:42mkaplyvideocontrols.css is using SVG. You would think this shouldn't happen...
15:42mkaplyAlthough desktop controls look completely different
15:44mkaplyYeah, our controls don't seem like they are videocontrols.css
16:20rbarkeresawin: ping?
17:49esawinrbarker: pong
17:50rbarkeresawin: You have a pixel C correct? Could you test today's nightly. I has the new dynamic toolbar and I'm guessing it is busted on tablets (don't have one to test). Particularly in landscape mode.
17:50rbarkerI has -> it has
17:51rbarkeresawin: I'm guessing hiding/showing will display a red bar instead of the toolbar snapshot.
17:53esawinrbarker: works fine, only teh back button icon is broken
17:54rbarkeresawin: The back button icon disappears while hiding/showing?
17:54esawinno, it's broken by default
17:54rbarkeresawin: no sure I understand.
17:54rbarkernot sure
17:55esawinrbarker: I can post a screenshot
17:55esawinlooks like a round layer mask with the icon missing
17:56rbarkeresawin: I probably broke it when I fixed "the curve"
17:59rbarkeresawin: could you post a picture of what it should look like? I'm guessing it's is the curve code. I wonder if I can turn on the back button on my device....
18:00esawinalso, toolbar hiding is blocked while the page is loading, not sure if that's new.
18:01esawinok, that's not new.
18:03rbarkeresawin: that is newish, but was a security fix before the new toolbar landed.
18:04esawinrbarker: that's on release:
18:04rbarkeresawin: it was uplifted.
18:04esawinI mean the back button
18:05bkellydoes anyone here know how to flash android onto the sony xperia we got at the whistler all hands?
18:05rbarkeresawin: sorry, yeah. Thanks. I think I know what is wrong. Just need to figure out how to get the back button to show...
18:06katsbkelly: on the z3c? i have cyanogenmod flashed on it
18:07katsyou might be able to flash a stock android if you can find a sony z3c build
18:08bkellyok, thanks
18:09esawinrbarker: both, the curve and back button are somewhat redundant, if you ask me :)
18:09rbarkeresawin: It shouldn't be hard to fix, just need to be able to reproduce.
18:34njparksebastian: hmm, I think I'm missing something - is it that I can run ./gradlew testFocusWebkitDebugUnitTest jacocoTestFocusWebkitDebugUnitTestReport to collect jacoco reports?
18:34njpark(I created a PR:
18:53nine_milliis ndk11 required or can you use a later version?
19:22njparksebastian: woo never mind, got it, forgot to actually build the stuff :|
19:25rbarkerjchen: ping? I have a small patch for you to review, would you prefer MozReview or Splinter?
19:25jchenrbarker: either way
19:26snorpnalexander: do you know if there is something about our geckoview aar that prevents studio from introspecting it correctly
19:26snorpnalexander: builds work, but the IDE acts like it can't see types, etc
19:27rbarkerjchen: I guess MozReview since I should probably get used to using it :/
19:30rbarkerdon't tell snorp, I don't want to get fired ;-D
19:31nine_millihey i did an update to get jchen's changes and now i get an error when linking webrender
19:31nine_milli error: linking with `cc` failed: exit code: 1
19:33nine_milliit's complaining about -ldl -lpthread -lgcc_s
19:34esawinjchen: the lib handler handling is a mess, it needs a rewrite :/
19:35esawinjchen: we need locking at the handler cache level, not at the object level.
19:56katsnine_milli: can you pastebin more of your compiler output? not really much context to go on
19:57nine_milliaha i think i was missing libc-dev 386 version
19:57nine_milliseems to have passed that part now
19:57rbarkeresawin: could you try: I'm not sure that will work but...
20:00esawinrbarker: almost there:
20:02rbarkeresawin: thanks, the stupid PorterDuff code is broken for software rendering. >:(
20:04nine_millinow it's just missing -lgcc_s
20:05nine_millianyone know packagae that is in debian?
20:15rbarkeresawin: one more to try:
20:15rbarkerif that doesn't work, I'm going to have to do something ugly :/
20:19esawinrbarker: no change
20:19rbarkeresawin: thanks. :/
20:19rbarkeresawin: was sort of expecting it.
20:23njparksebastian: ahunt trial run with only focuswebkitdebug:
20:23njparkI guess we should run everything except release perhaps?
20:31nalexandersnorp: when it was setup in the past, AS definitely indexed it correctly. It's possible I busted things with
20:31firebotBug 1320310 FIXED, Local builds: Home screen icon disappears when switching between gradle/mach build
20:38ahuntnjpark: since were running tests for all branches, I wonder if its worth enabling coverage for all. That said, were hopefully not running different code in different builds (except for gecko vs webview)
20:39njparkahunt: only odd thing is that for the release branch, one unit test failed, that kinda killed the report generation
20:39ahunt(And I guess there are minor differences, such as webview debugging and screenshots being enabled for debug builds)
20:40ahuntnjpark: is that the DisconnectTest? If so - it only works on buddybuild because it relies on the adjust token being set (which is set by buddybuild)
20:40njparkahunt: yup it was, oh so if I feed this thing @ buddybuild, it'd work then
20:40ahuntyup, it only fails locally.
20:41ahunt(We didnt want to add special case code to disable it for tests, because if that breaks then)
20:41njparkahunt: I'll add a coverall token in bb, but I'm not really sure which gradle command is run by bb - I need to figure it out and append coveralls parameter in the end
20:43njparklocally i just do ./gradlew --info clean assembleDebug lint testFocusWebkitDebugUnitTest jacocoTestFocusWebkitDebugUnitTestReport coveralls
20:49rbarkeresawin: still there? I've got one more to try... uploading now...
20:49ahuntnjpark: you could maybe check if jacoco is listed as a dependency of build in "gradle tasks all?
20:52ahunt(I dont know if buddybuild just does a |gradle build| or something else though)
20:53njparkahunt: yeah... I thought there is this final gradle command that BB executes somewhere.. I should read up on how android tests are configured
20:53njparki don't think I declared it as a dependency
21:31ahuntJust wondering - afaict we tend to import the sources of external dependencies (other than the sdk/support libraries). Is there any reason we couldnt use aars for those too (Im guessing the issue there is wed have to add our own scripts to download them at build time, which I guess we dont want - except that gradle does that anyway)?
21:31ahunt(And add more tooltool archives with those aars for automation, etc)
21:32ahuntActually - where do our gradle builds even get their deps from in automation (Im guessing they cant just download whatever from the web)?
22:08snorpahunt: nalexander has a magic cache
22:10nalexanderahunt: we want to support building in TC (where downloading AARs from the web is trivial) and also in buildbot (where downloads pretty much have to be in tooltool). Hence, tooltool first.
22:10nalexanderahunt: as for AAR vs. source, generally we want source so we know what goes into our builds. That's philosophical, not logistical.
22:10nalexanderahunt: over time, I think we're going to have to trust more and more external sources. (We already do for the Android toolchain, Google Play Services, etc.)
22:11nalexanderahunt: you can look in mobile/android/configs/tooltool for the (opaque) tooltool details.
22:11ahuntAh I never considered the trust aspect.
22:12ahuntSo IIUC our gradle builds are done in TC, and gradle just grabs its deps on every run there (possibly cached)?
22:12ahunt*only in TC (whereas older builds are still in buildbot)
22:14esawinrbarker: looks good, is that using the patch?
22:15rbarkeresawin: yes.
22:15rbarkeresawin: I was able to test it in emulator...
22:21esawinrbarker: this also reproduces the issue with the old toolbar, an isomorphic fix ;)
22:24rbarkeresawin: what was the old issue?
22:28mcomellagrisha: I'm getting a 401 (unauthorized) when I hit the token server but I've basically copied the SyncAdapter code so I'm not sure why that is. My code is here:
22:29rbarkeresawin: updated commit message to include fix and pushed to inbound. Hopefully in tomorrows nightly. thanks...
22:29mcomellagrisha: I rm'd code that seemed unimportant to getting it working (e.g. backoff) and I'm using the iOS endpoints ( so those could be related
22:30mcomellagrisha: I verified I had the &quot;Authorization: BrowserID <key>&quot; header, as per I don&#39;t know that that key is valid however
22:31mcomellaWith &quot;unauthorized&quot;, it gives &quot;invalid-credentials&quot;
22:31esawinrbarker: I applied your patch on an old branch, which did not have your toolbar changes in, and it would reproduce the issue you just fixed with your new toolbar. makes sense since you just invert the path.
22:31rbarkeresawin: ah, yeah, I would expect that actually :)
22:33rbarkeresawin: it&#39;s because I switch Mode.DST_IN to Mode.DST_OUT to fix &quot;the curve&quot; which requires the path be inverted.
22:35nalexanderahunt: yeah, that&#39;s correct. Gradle builds are &quot;offline&quot; and we prepopulate the Gradle dependencies on the builders.
22:35nalexanderCallek: are you still online? (It&#39;s late Friday for you.)
22:44ahunt(I guess we could build the AARs ourself, but at that point theres not much difference to having them in tree. Although it does make testing upgrades simpler if you can just flip a switch in build.gradle locally to play with a new version / new library.)
22 Apr 2017
No messages
Last message: 2 days and 7 hours ago