Hector Martin

It's telling you you don't have `setuptools` installed. `setuptools` is required to run `setup.py` for many, if not the majority, of `setup.py` scripts out there. This is all completely standard...

The reverse problem also happens: keys set by 1.58 as strings are now returned as bytes on 1.59, breaking code that expects them to be strings.

Actually aplay doesn't work on my system with pulse, but that's because of the `-D plug:default`. Pulseaudio already does sample rate conversion, and it seems using the `plug:` prefix in...

There's nothing special about my system as far as I know, it's a pretty bog standard Arch Linux install on a ThinkPad with Intel HDA audio. My guess is this...

Sorry for the long delay, [here's](https://hub.marcan.st/t/sweep-noise.opus) the .opus file.

And you're right - opusdec decodes it properly. So it looks like what I'm hitting is a bug in the ffmpeg opus *decoder*! I guess that means we can close...

I'm using `opusdec opus-tools 0.2 (using libopus 1.3.1)`

If I had to guess that might actually be a divide by zero or corner case vertex alignment issue. It seems to happen when the notes are dead vertical/horizontal, right...

Looking at the same shader as in #1865, I see `v_BlendRange.x` is checked for zero in several places, but not in [this](https://github.com/ppy/osu-framework/blob/f9cb6c89a328f179c5121bc77cd0f711576ef17f/osu.Framework/Resources/Shaders/sh_TextureRounded.fs#L73) line. That's a varying, so if the pixels...

Potential bug [here](https://github.com/ppy/osu-framework/blob/f9cb6c89a328f179c5121bc77cd0f711576ef17f/osu.Framework/Graphics/Lines/PathDrawNode.cs#L139). `Texture.Width` is an integer, so the integer division truncates. The texture is set to clamp mode, so it should still render pixels there, but that might be...