Delay frame processor initialization until view creation on Android

[jthure]

What

This PR fixes java.lang.NullPointerException in FrameProcessorRuntimeManager. on Android phones

Although I never was able to reproduce the issue myself, looking at the crashlogs from other users, this should fix it

Changes

Moved the creation FrameProcessorRuntimeManager to happen in onCreateViewInstance instead of inside CameraViewModule initialisation. This should make sure that the react context is fully initialised, which apparently is not the case right now

Inspired by @pke in #946 (comment)

Tested on

Not properly tested yet. I'm having trouble building and running the example app from this repo. I have tested it in another app which uses this library. Any help to test and verify the change is greatly appreciated

Related issues

Fixes #946

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
react-native-vision-camera	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	May 4, 2023 9:19am

[pke]

Thanks for creating this PR!

[mrousavy]

Hey! Thanks for creating this PR, but I don't think that this is the right approach to this problem. This just swallows the error and silently fails when you try to use FPs on the JS side later on.

Also, this elevates the internal setup from internal (CameraViewModule.kt) to external (CameraViewManager.kt; aka the CameraViewModule's caller) code.

[mrousavy]

In V3 I have a different solution here which is called from JS - so this is when the JS bundle is 100% ready. That's probably what we want.

Sorry for closing this PR please don't take it personal ❤️

[pke]

Hmm... this patch was written against current main, which is not yet V3. Why not allow this backport until V3 is production ready?

[mrousavy]

Because of this:

This just swallows the error and silently fails when you try to use FPs on the JS side later on.

the FP runtime will not be installed and will just not work. You can disable FPs at build time if that's what you want

[jthure]

Hey! Thanks for creating this PR, but I don't think that this is the right approach to this problem. This just swallows the error and silently fails when you try to use FPs on the JS side later on.

Also, this elevates the internal setup from internal (CameraViewModule.kt) to external (CameraViewManager.kt; aka the CameraViewModule's caller) code.

Okay sure, I get what you mean. Do you have another approach in mind on how to fix it? Before v3 is ready and adopted?

Edit: Actually, I'm not sure I fully agree. It doesn't just swallow the error. The error should not happen since the fp init happens at a later stage, right?

[mrousavy]

Do you have another approach in mind on how to fix it?

If you do not need Frame Processors, just disable them at build time.

If you do, then I am not sure - I think the solution in V3 is much more solid since it does init from JS side - but this is a big change

Actually, I'm not sure I fully agree. It doesn't just swallow the error. The error should not happen since the fp init happens at a later stage, right?

No, it is wrapped in a try/catch now and in the catch there's no handling - aka it silently fails.