Guilherme Bernal

Turns out LLVM produces some really repetitive code when not optimizing. The generated wasm code for `String::Grapheme::codepoints` has 34151 local variables without `--release`, but only 2 locals when optimizing. This...

> So does that mean the workaround in [#4516 (comment)](https://github.com/crystal-lang/crystal/issues/4516#issuecomment-306226171) somehow does not work when targetting WASM? Yes, it still generates too much code. Also, compiling the specs in release...

I did a significant refactoring of this PR to make is safer and more organized. Relevant changes: - Created the `Crystal::Asyncify` module. It holds the logic of stack switching and...

It's fixed! # What was going on? By default the main stack comes after data and grows down. Nothing prevents it from overflowing and overwriting something in the data section....

Not any entire spec file. But there are some I can enable by adding flag checks inside the spec files. Should I? Channels is a good example. Most tests pass,...

Hi @z64! I'm having very little time to work on this lately, so things are going slowly. The current status is that the code done and has been refactored in-line...

Nice! I'll be going for a PR introducing it to the stdlib. > I think option 2 sounds great. But I think it would be good to have for a...

I agree, those are easily splitable and will simplify the review. I'll do that.

Yes!

Another related case: Save this as `bug.test.ts`, then run `deno test --unstable-webgpu` ```typescript const adapter = await navigator.gpu.requestAdapter(); if (!adapter) { throw new Error("WebGPU is not supported on this device");...