juj

A SINGLE_FILE build expressing the failure can be seen in https://oummg.com/dump/buffer_map_sync.html

What is odd here is that the error occurs in the next run of the requestAnimationFrame event loop..

Ah actually it is not the next run of the requestAnimationFrame event loop, but the current call. So I need the `WebAssembly.promising()` call. I find this patch fixes the issue:...

For now I removed use of `emscripten_set_animation_frame_loop()` as it is incompatible with JSPI. In wasm_webgpu bindings I added a custom function `wgpu_request_animation_frame()` that works in a manner that is JSPI-aware,...

I observe that calling `WebAssembly.promising()` generates temporary garbage, so to avoid calling that function each frame to wrap the same function pointer, I wanted to optimize the wrapping to only...

The more I poke at https://oummg.com/dump/buffer_map_sync_breaks_after_a_few_seconds.html RuntimeError, I get a feeling like that would be a browser bug. I can't see any reason why execution of such a requestAnimationFrame loop...

I reported the bug in Chromium bug tracker at https://issues.chromium.org/issues/364667545. The difference between the working and failing runs is as small as 

Revised the title of this bug to focus on the Emscripten side challenge, namely that none of the callback-taking functions work with JSPI, unless one does a custom wrapper like...

> it's not really an API that universal scripts ought to be using anyway Looking at [MDN: document.currentScript](https://developer.mozilla.org/en-US/docs/Web/API/Document/currentScript) there are no markings of the API being deprecated or obsoleted. This...

I have a little bit of experience for creating WPT tests from the interaction we had before. To get going, I added a WPT test at https://github.com/web-platform-tests/wpt/pull/48536 . Of course...