Michel Hermier

> > The problem lies in the fact that tracking WrenHandle and the associated WrenVM > > Isn't there just a single loop and single WrenVM? I'm talking about making...

> > you can't subclass a uv_handle_t at user level. ... you can't create handles without > > Create handles from Wren? That sounds like pretty tight coupling to me......

> > with possibly multiple loops > > To what end? At first glance that sounds... unnecessary? One loop (plus a stronger scheduler on the Wren side) should be more...

Digged into this and there is no solutions for this for now. Redirections like these results in UV_FILE type. Current implementation tried to mitigate the issue by making it behave...

`readLine` should technically returns a buffer no matter what encoding errors happens in there. Strictly speaking UTF-8 encoding is supported but not required to be in a `String`. If we...

To solve the original problem, the proper solution would be to add range support to `StringByteSequence.[_]` or do some proper buffering for efficiency.

@ruby0x1 The doc is correct, but the non documented range implementation is not byte indexed....

@ruby0x1 test it with unicode characters, results should be broken, and be based on UTF-8 indexes: - https://github.com/wren-lang/wren/blob/main/src/vm/wren_value.c#L735 lenght of the new buffer is based on utf8 not on the...

EDIT: Previous message contained a broken test. The following patch should be more conform to the byte indexing specification. I can make a branch if required. ```patch commit 5baaffacb4909c83ea9302c8feed582d101e2600 Author:...

It is completely different, since some produce invalid UTF-8 strings that makes the test suite explodes.