Marijn Suijten

Results 1102 comments of Marijn Suijten

> `native_activity` is public so it could happen. and I think winit does actually call `native_activity` and it runs in a separate thread. Yeah I think that's the point we're...

We can possibly drop/`.finish()` in an `atexit` handler as well? EDIT: I doubt the `atexit` handler runs unless all threads are gone (and the entire process terminates), which is likely...

> It's funny how the smaller the PR the larger the discussion hahha Yup the bigger churn is never an issue, but smaller controversial lines like these need to be...

In hindsight I'm not too certain about the proposed changes above. The looper is explicitly associated with the current thread, and returning from `fn main()` means the looper must have...

@msiglreith Checking out `winit` it never calls `.finish()` for the user anywhere, probably leading to a stuck application. Crates using `winit` are responsible for using the `ndk_glue::main` macro directly themselves,...

@msiglreith Thanks for your view, and apologies for letting this sit unattended for a while. > @MarijnS95 I'm fine with these. Regarding `finish` vs `drop` I would prefer calling `finish`...

Note that you can link to lines in build steps directly: https://github.com/torokati44/ruffle-android/runs/5707487779?check_suite_focus=true#step:5:810 As you pointed to #20, `ndk-build` only bundles this library (or any library in general) when the linker...

I'm surprised. You even added `println!("cargo:rustc-link-lib=c++_shared");` to force linking against this lib yet it still doesn't show up. Perhaps the NDK linker is stripping this, but I don't see `-nostdc++`...

@zarik5 Thanks for reporting! I have definitely been thinking about calling it after `fn main()`: https://github.com/rust-windowing/android-ndk-rs/blob/49e8ed51220567dccac9fd1e8baa526e9cdac163/ndk-glue/src/lib.rs#L183-L202 We can perform and publish this as a breaking change as it might potentially...

We've had quite the discussion in #160 and automatically calling `.finish()` upon returning from `fn main()` seems undesirable at best; applications will have to call it themselves. I'll think about...