Robert Bragg

I haven't had a chance to look at improving this to upstream but we're using this patch here currently: https://github.com/EmbarkStudios/winit/commit/9f47e7bad9bd98b8a12f8e920a9f650b63ad7c6a which is probably equivalent to what you have. One thing...

Interesting, thanks for experimenting to check the state of the device ID - yeah I had only figured that it intuitively seemed like it could make sense to check the...

> Another high-level thing that needs to be considered is how to handle sandboxing this, and how to expose that to the user. The main things we'd need to achieve...

Sorry for the delay with getting a chance to look at this. I agree that it should be reasonable to be caching `java.lang.String` class references and even more generally think...

> Thanks! I actually have a new concern - https://mostlynerdless.de/blog/2023/07/17/jmethodids-in-profiling-a-tale-of-nightmares/ says that a jclass becomes invalid if the class is redefined. I haven't confirmed it applies to this case, but...

My guess here is that you're building with the "invocation" feature enabled which isn't supported on Android. The code you're changing, relates to support for loading and starting a JVM,...

Even though it's not technically possible to load a new jvm on Android, maybe it'd be possible to have the invocation API effectively be a NOP on Android if you...

I'm not really sure about my working theory above, and I'm not really sure what to suggest here without more context. This crate doesn't have a build.rs and doesn't have...

Ah, I just realized this is with an older version of this crate when it used to have a `build.rs` This was removed some time ago here: - https://github.com/jni-rs/jni-rs/pull/293 and...

Heya @wuwbobo2021 and @argv-minus-one; sorry that this is quite a big dump of draft changes but it would be great to get some more eyes here to see what you...