Android Support #96

saikrishna321 · 2017-03-07T11:56:37Z

Is there any ETA for android support as react-native apps support both iOS and android

rotemmiz · 2017-03-07T12:24:08Z

No ETA yet, we are in the process of finishing and stabilizing detox for iOS. Android will follow.

saikrishna321 · 2017-03-16T10:28:49Z

@rotemmiz any specific tool detox would depend on like espresso ?

rotemmiz · 2017-03-16T11:42:36Z

The plan is to use espresso, yes

jeduden · 2017-03-22T13:45:04Z

@rotemmiz is there a rough implementation plan ? would you consider accepting pull requests helping to get android support working faster ?

rotemmiz · 2017-03-22T15:41:21Z

There's a plan indeed. I think it will be best if I write a bit about our roadmap in the documentation.

In a nutshell, we want to use a similar architecture as the one we implemented on iOS:

use Espresso as the native driver (we'll need to figure out how to integrate with espresso, since it runs only with instrumentation context)- there's already a rough native skeleton implemented. the invocation mechanism is already implemented in react-native-invoke.
wrap android and adb for emulator control - like we did with fbsimctl.
extend Device.js to AndroidDevice and AndroidEmulator, implementing all of its functionality.
The hardest part will be creating a custom IdlingResource for react native bridge, like we did in iOS. Since java can't swizzle functions, we'll need to use a different approach.

rotemmiz · 2017-03-22T15:41:45Z

And of course, we love PRs :)

DanielZlotin · 2017-03-22T16:27:39Z

for future reference: I managed to make this almost work, look here

Although I gave up on this when trying to test application lifecycle, as Espresso runs as part of the app under test's process, so stopping one will stop the other.
I ended up using uiAutomator with great results

rotemmiz · 2017-03-22T16:33:41Z

I'm a bit afraid of trusting uiautomator , I had experience with it in MagnetoTesting, and it was flaky, we had to pet our UI tests constantly.
The problem you raised might not be an issue as the test runner in detox is external to the application and to espresso's lifecycle and resides in a different process.

DanielZlotin · 2017-03-23T08:27:47Z

if you look at the current tests in https://github.com/wix/react-native-navigation/tree/v2/AndroidE2E/app/src/androidTest/java/com/reactnativenavigation/e2e/androide2e you'll see there's zero flakiness, it's just a little different kind of thinking than espresso

talarari · 2017-04-02T16:46:32Z

also its worth noting that uiAutomator lets you interact with stuff outside the app.
so pulling down the notification curtain and such would work.

simonracz · 2017-04-12T09:35:58Z

Hey, I plan to work on 2. and 3. from @rotemmiz 's list.

Is anyone else working on this?

Few notes:

android is deprecated, was split to sdkmanager and avdmanager
platform_tools can be installed now separately, which is nice
adb and avdmanager seems to be enough alone, they cover the current use cases

simonracz · 2017-04-12T09:41:22Z

Which one of the following would be preferred?

Call adb directly similarly to how fbsimctl is called.
Use adbkit for adb commands.

The 2. option adds another dependency, might not cover everything, but on the upside is already tested. I would prefer 1. at the moment, but really don't mind.

talarari · 2017-04-12T12:13:25Z

@rotemmiz - can you elaborate on this please?

The hardest part will be creating a custom IdlingResource for react native bridge, like we did in iOS. Since java can't swizzle functions, we'll need to use a different approach.

LeoNatan · 2017-04-12T13:41:05Z

@talarari What I did for iOS was to strategically swizzle RN methods so that I could react to what is happening for internal logic, such as starting a timer or the JS thread runloop.

Method swizzling is the process of replacing a method implementation at runtime. Since Objective C is dynamic, there is a great flexibility regarding this. Java does not allow method swizzling, so it would be much more difficult to implement a similar approach.

LeoNatan · 2017-04-12T13:46:05Z

Some approaches exist by hot-patching the vtable:

http://vaioco.github.io/art/art-part-ii-hooking
http://vaioco.github.io/art/art-part-iii-arthook-framework//
https://github.com/vaioco/art-hooking-vtable

@rotemmiz

rotemmiz · 2017-04-12T23:35:09Z

Thanks for the references @LeoNatan, I like where this is going :)

@talarari , @simonracz , I will post our roadmap for detox soon, along with how we thought about implementing support for Android.

simonracz · 2017-04-20T14:11:49Z

Just to keep the ball rolling.

I have a proof of concept code for a custom Android test runner. It is actually a subclass of MonitoringInstrumentation.

It gets the detoxServer and detoxSessionId arguments from cmd line (or from Gradle settings). Can be started like adb shell am instrument -w -e detoxServer ws://localhost:8001 -e detoxSessionId 1 com.example.detoxsample.test/com.detox.DetoxRunner.

It opens a websocket to the detox server, logins to it and then starts listening to commands. From this point it can call, .e.g Espresso methods using reflection as it has the permission to poke the app in any way it wants.

Some open problems for me:

Espresso's main selling point is it's automatic detection of idleness. For this, custom async code must notify Espresso about it's work. E.g. okhttp-idling-resource This must be added for react native.
Espresso needs AndroidJUnitRunner. Subclassing this was a not an option. So, I have to add it's functionality to the custom test runner, which is a bit messy.
React Native transforms testids to view tags on Android. This is okay for Espresso, but won't work for UIAutomator. (UIAutomator is needed for testing notifications.)

ptomasroos · 2017-04-20T14:47:06Z

Great news. Looking fwd to a follow up!

rotemmiz · 2017-04-20T15:12:42Z

@simonracz this is great!! can you please share the code ?

regarding your points:

it would be easiest to add okhttp-idling-resource to react-native, but I want to start trying the patch art vtable (this would be very useful for other features, like the RN bridge idling-resource).
Did you have a look at my initial skeleton for Android ?(https://github.com/wix/detox/blob/master/detox/android/detox/src/main/java/com/wix/detox/DetoxManager.java). It uses instrumentation, and when the single instrumentation test is initiated, detox adds a looper to that thread, forcing it to wait for network requests. I'm not sure this is the right solution yet, but it saves the need to run a custom test runner.
I agree, although the issue with UIAutomator should be in a lower priority, we can send a PR to RN that adds testIDs to content-description as well.

simonracz · 2017-04-20T20:52:06Z

@rotemmiz I'll upload the code tomorrow morning.

Yes, I did, but completely glanced over the Looper. Were you able to test this approach? It seems simpler than mine.

I am staring at the AndroidJUnitRunner code for half an hour now and can't remember why I made the conclusion that it can't be subclassed for this purpose. So, please ignore my earlier comment about that.

simonracz · 2017-04-20T22:42:20Z

One thing that goes for the custom test runner is that that is the only way to run code before Application#onCreate().

simonracz · 2017-04-21T08:54:02Z

I've uploaded the code here.

talarari · 2017-04-21T11:44:00Z

Regarding point 3:
we use accessibilityLabel as well as test id.
React native transforms it to content description

simonracz · 2017-05-01T12:54:57Z

I delved deep into RN's Android port while keeping in mind Detox's iOS support.

At first, I thought of using Espresso's internal QueueInterrogator class for idle detection. It would work fine, as while RN's core classes both have Java and mirror C++ parts, threading is done in a JVM aware style, only in the Java layer. The four relevant MessageQueues to be observed can be found here. The Loopers could be caught through the ReactContext for them. This is a valid approach, but I've found a better one.

RN's Android port seems to be more mature than the iOS port. They have a vertically implemented addBridgeIdleDebugListener function in CatalystInstance. They also have an idle detection utility that is based on this here.

The waitForBridgeAndUIIdle method could be used by Detox. This still needs to be wrapped for Espresso, but that's not an issue.

The only thing that is missing, is that it does not monitor the Timers the same way as Detox does. It just waits 2 frames of Choreographer, then marks it idle. (Choreographer is similar to CADisplayLink on iOS.) However, the timers are orthogonal to this.

Waiting for the Timers "the Detox way", could be done by getting hold of the mTimers variable using reflection, and checking whether it's empty or not. So, there is no need to attach to method calls, like on iOS.

I'll work on a proof of concept code and will check the idle detection manually with Detox's test. I'll share the code afterwards. Any feedback is welcome.

rotemmiz · 2017-05-01T16:30:30Z

Hey @simonracz, please check your Twitter ;)

AlanFoster · 2017-05-25T15:00:03Z

Adding to the discussion; I wonder if there will be an issue in providing full android test support without testID being available within android? 🤔

itsmepetrov · 2017-05-25T17:52:12Z

BTW, the long discussion about testID in Android: facebook/react-native#9942

simonracz · 2017-05-25T19:42:09Z

Hey, @itsmepetrov
I am familiar with that thread. Pretty sad. That's my main reason, why I don't really consider PRs to React Native.

Test ids are transformed to view tags on Android. Espresso can use view tags. UIAutomator can't.

I think tags will be okay for most of the cases.

simonracz · 2017-05-31T14:50:03Z

@AlanFoster You can find an RN view with testID='hello_button' the following way.
Espresso.onView(withTagValue(is((Object)"hello_button"))).check(matches(isDisplayed()));

This should enable other users to use android more easily as the current status configuration can be quite confusing. Build with the help of @AppsThatMatter Relates to #96

vocampos · 2017-10-13T19:28:45Z

I'm going through the problem below when running the tests.

My Settings (package.json)

"detox": "^5.8.1",
"react": "15.4.2",
"react-native": "0.41.2",
"mocha": "^4.0.1",


 "detox": {
    "configurations": {
      "android.emu.debug": {
        "binaryPath": "android/app/build/outputs/apk/app-debug.apk",
        "build": "pushd android && ./gradlew assembleDebug assembleAndroidTest -DtestBuildType=debug && popd",
        "type": "android.emulator",
        "name": "Nexus_5X_API_25",
        "session": {
          "server": "ws://localhost:8099",
          "sessionId": “appName”
        }
      }
    }
  }

Test (firsTest.spec.js)


describe('Example', () => {
  beforeEach(async () => {
    try {
      await device.reloadReactNative();
    } catch (error) {}
  });

  it('should show world screen after tap', function() {
    element(by.text('Say World')).tap();
    expect(element(by.text('World!!!'))).toBeVisible();
  });
});

Command executed:


 detox test --configuration android.emu.debug --loglevel info

Result

"after all" hook:
Error: Timeout of 120000ms exceeded. For async tests and hooks, ensure "done()" is called; if returning a Promise, ensure it resolves.

Any suggestion?

simonracz · 2017-10-13T20:11:27Z

Hey @vocampos ,

Could you please run an adb logcat in a terminal while you run your tests?
My guess is, that you will see something like “XY is busy” in the logs.

Please share those lines with me.

Cheers

simonracz · 2017-10-17T10:35:43Z

The --debug-synchronization [ms] feature for Android has just been pushed to the master branch. It'll be in the next release.

Example usage

detox test -c android.emu.debug --debug-synchronization 500

This is the first tool you should consult, in case your tests are extremely slow or even timeouts. It reports back the busy IdlingResources that are holding back the test call.

vocampos · 2017-10-17T17:06:42Z

@simonracz ,

I ran the command according to its orientation, but it did not work.

When analyzing the adb logs, I have the below error:

--------- beginning of crash 10-17 14:54:38.384 3639-3677/? E/AndroidRuntime: FATAL EXCEPTION: com.wix.detox.manager Process: com.app, PID: 3639 java.lang.RuntimeException: ReactInstanceManager is null at com.wix.detox.ReactNativeSupport.waitForReactNativeLoad(ReactNativeSupport.java:159) at com.wix.detox.DetoxManager.start(DetoxManager.java:64) at com.wix.detox.Detox$1$1.run(Detox.java:133) at android.os.Handler.handleCallback(Handler.java:751) at android.os.Handler.dispatchMessage(Handler.java:95) at android.os.Looper.loop(Looper.java:154) at com.wix.detox.Detox$1.run(Detox.java:136) at java.lang.Thread.run(Thread.java:761)

Any suggestion?

Thank you

vocampos · 2017-10-17T19:54:15Z

The error is happening for detox.init not finished.

before(async () => {
    console.log('Start ');
    await detox.init(config.detox);  // not finished
    console.log('End');
});

1) "before all" hook: Error: Timeout of 120000ms exceeded. For async tests and hooks, ensure "done()" is called; if returning a Promise, ensure it resolves.

simonracz · 2017-10-18T04:38:42Z

Hey @vocampos

Thanks for the extra info.

You just hit an error of the type that should never happen ™. :)

I've just noticed that you use RN 0.41.2. We never tested the Android version below 0.44.0. So, any bad thing could happen there. That's my current guess.

If you can reproduce it with later RN versions too, please let me know.

Regardless of that, I'll take a look whether I can support 0.41.2 too.
But, please be aware that under 0.45.0 we already know, that some things can't be supported, e.g. syncing with native animations.

skv-headless · 2017-11-02T12:16:51Z

How to match by type on android?
For example on ios by.type('RCTTextInput') is working find. On android I tried TextInput, ReactTextInput, EditText, ReactEditText nothing seems to work.

simonracz · 2017-11-02T12:48:47Z

For Textinput you probably have to search by android.widget.EditText.

My guess is that the full classname was the missing part for you.

Let me know if it still doesn’t work.

vasyas · 2017-11-03T15:07:02Z

Is Genymotion supported atm?

ptomasroos · 2017-11-04T07:22:40Z

@vasyas #96 (comment)

ilonashub · 2017-11-09T13:04:52Z

@simonracz hey, I have the same issue with running detox on Android. We got it working in iOS.
We use RN 0.49.
When I run detox on Android - I get a white screen, and in the console I get stuck in the "before all" hook:

 1) "before all" hook

  0 passing (3m)
  1 failing

  1) "before all" hook:
     Error: Timeout of 200000ms exceeded. For async tests and hooks, ensure "done()" is called; if returning a Promise, ensure it resolves.

I tried to look for stuff in the adb logcat, but didn't see any error with “XY is busy” like you mentioned.

Do you have any idea what might be the problem? Thanks!

simonracz · 2017-11-09T13:31:07Z

Hey @ilonashub ,

It's hard to guess what went wrong.

Few random things, I would check.

If your app is a hybrid app, then, is your detox android test (not the js, but the java test) starts the correct Activity, which starts RN?
You should see something in the logs. In case there are no errors there at all, how far did your app managed to get? (It's very strange that you see only a white screen.)
You can also start the detox server manually with detox run-server. I would check whether the android emulator also logs there, not just the js test runner.

I think it would be best to open a seperate issue, and please attach a chunk of adb logcat that you think might be relevant.

Simon

simonracz · 2017-11-09T13:37:54Z

Hey @vasyas ,

@rotemmiz worked on this, but I think Genymotion is currently not supported. See here.

We should have an option, which just let's you use "whatever is visible via adb".

I plan to look at this. Was a bit busy recently with non-detox stuff.

vasyas · 2017-11-09T15:21:45Z

@simonracz thanks.
I had already connected to Genymotion using the way you said ("whatever is visible via adb").
But now I'm getting a different message, see #386. Could you comment what else needs to be done to fix it?

Probably I can make it a pull request.

ipinzari · 2017-11-16T19:19:24Z

I m having a hard time getting detox working with Android, it can't connect with the Emulator. Is there a more detailed android Get Started where it goes step by step how to configure?

rotemmiz · 2017-11-16T19:49:04Z

Not yet, they will be out when Android support gets ironed out. We are currently working on that.

ipinzari · 2017-11-16T19:58:39Z

do we have an estimated date?

…

On Thu, Nov 16, 2017 at 11:49 AM, Rotem Mizrachi-Meidan < ***@***.***> wrote: Not yet, they will be out when Android support gets ironed out. We are currently working on that. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#96 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ATQZ0sWz6aQPSnVBDut4I7UqFL9Pr69Aks5s3JG7gaJpZM4MVY4e> .

-- *Ion Pinzari* QA Automation Engineer | SkySlope t: 916-225-6318 e: ipinzari93@gmail.com <ipinzari93@gmail.com> | w: SkySlope.com <http://www.skyslope.com/>

rotemmiz · 2017-11-16T20:04:52Z

It's an ongoing open source project effort. Android support is usable, and will continue to get better as we put lots of efforts in it.

You are welcome to open a specific issue with exact information regarding your errors, it might help us pinpoint issues we didn't yet find.

There will never be "estimate dates", especially to vague goals !

Karniej · 2018-01-18T11:04:43Z

Anyone got Detox + Jest working on Android with
"detox": "^7.0.0-alpha.0"
jest": "^22.1.0",
"babel-jest": "^22.1.0",
?
I am struggling with configuring it for more than a day today and nothing really helps. I just want to know if someone is actually using it.
I got it running on iOS pretty quickly...

Well I have moved from Genymotion to Android Studio 3 and it helped...

llostris · 2018-01-19T00:35:49Z

I'm also having issues setting it up for Android (using detox 7). When I run detox it uploads and starts the app, but then it hangs on the first, white screen (app is not responding)...

rotemmiz · 2018-02-05T08:23:18Z

Android support is already pretty mature now, I suggest any problem will be handled in a separate issue.

Closing this issue.

rotemmiz mentioned this issue Apr 11, 2017

Fixed demo-react-native for Android. #112

Merged

DanielMSchmidt mentioned this issue Oct 5, 2017

Add documentation for the usage with Android #316

Merged

simonracz pushed a commit that referenced this issue Oct 5, 2017

add documentation for the usage with android

64a34f9

This should enable other users to use android more easily as the current status configuration can be quite confusing. Build with the help of @AppsThatMatter Relates to #96

vocampos mentioned this issue Oct 13, 2017

Tests hanging on before_all hook #155

Closed

vasyas mentioned this issue Nov 9, 2017

Implemented AttachedAndroidDriver device for connecting to Genymotion #397

Merged

sibelius mentioned this issue Dec 1, 2017

Detox 6.0.0 does not work with react native 0.50.1 #440

Closed

rotemmiz closed this as completed Feb 5, 2018

wix locked and limited conversation to collaborators Jul 23, 2018

Android Support #96

Android Support #96

Comments

saikrishna321 commented Mar 7, 2017

rotemmiz commented Mar 7, 2017

saikrishna321 commented Mar 16, 2017

rotemmiz commented Mar 16, 2017

jeduden commented Mar 22, 2017

rotemmiz commented Mar 22, 2017

rotemmiz commented Mar 22, 2017

DanielZlotin commented Mar 22, 2017

rotemmiz commented Mar 22, 2017

DanielZlotin commented Mar 23, 2017

talarari commented Apr 2, 2017

simonracz commented Apr 12, 2017

simonracz commented Apr 12, 2017

talarari commented Apr 12, 2017

LeoNatan commented Apr 12, 2017

LeoNatan commented Apr 12, 2017 • edited Loading

rotemmiz commented Apr 12, 2017

simonracz commented Apr 20, 2017

ptomasroos commented Apr 20, 2017

rotemmiz commented Apr 20, 2017

simonracz commented Apr 20, 2017

simonracz commented Apr 20, 2017

simonracz commented Apr 21, 2017

talarari commented Apr 21, 2017

simonracz commented May 1, 2017 • edited Loading

rotemmiz commented May 1, 2017

AlanFoster commented May 25, 2017

itsmepetrov commented May 25, 2017

simonracz commented May 25, 2017

simonracz commented May 31, 2017

vocampos commented Oct 13, 2017

simonracz commented Oct 13, 2017

simonracz commented Oct 17, 2017

vocampos commented Oct 17, 2017 • edited Loading

vocampos commented Oct 17, 2017

simonracz commented Oct 18, 2017

skv-headless commented Nov 2, 2017

simonracz commented Nov 2, 2017

vasyas commented Nov 3, 2017

ptomasroos commented Nov 4, 2017

ilonashub commented Nov 9, 2017

simonracz commented Nov 9, 2017 • edited Loading

simonracz commented Nov 9, 2017 • edited Loading

vasyas commented Nov 9, 2017

ipinzari commented Nov 16, 2017

rotemmiz commented Nov 16, 2017

ipinzari commented Nov 16, 2017 via email

rotemmiz commented Nov 16, 2017

Karniej commented Jan 18, 2018 • edited Loading

llostris commented Jan 19, 2018

rotemmiz commented Feb 5, 2018

LeoNatan commented Apr 12, 2017 •

edited

Loading

simonracz commented May 1, 2017 •

edited

Loading

vocampos commented Oct 17, 2017 •

edited

Loading

simonracz commented Nov 9, 2017 •

edited

Loading

simonracz commented Nov 9, 2017 •

edited

Loading

Karniej commented Jan 18, 2018 •

edited

Loading