Angus Gratton

> If the hardware engine switches out to the software engine when concurrency occurs, how come it is slower than just using the software engine? Does the copy out of...

Stumbled across this problem today. FWIW, I think the matching wine bug is here: https://bugs.winehq.org/show_bug.cgi?id=40528

Hi Matt, Thanks for looking into this as well. I think you've actually identified a different, similar, issue - that the chid cache is never cleaned up unless you explicitly...

Hi Matt, Good catches. I've committed 211d2b7a90f52c4fb2128e07f74b734046f8955d which removes the type() test (I knew that was bad Python, I just wasn't thinking.) I've also refactored the conditions around wrapping/unwrapping cached...

Matt, do you have an example of code that causes problems? I might be able to suggest something.

Interesting... the only code path that creates a weakreference should be if you pass a callback directly to create_channel(). create_subscription() shouldn't create a weak reference at all. So I can't...

@Molorius Thanks for contributing this, sorry for the awkward timing! Regarding what I'm working on, my immediate goals are: * Deciding on a device API, as @dpgeorge mentioned. I'm keen...

@Molorius That all sounds good! > I'm not interested in fully developing USB specs within python so I will continue on my Lua project for now. I'm happy to provide...

@dpgeorge @jimmo PTAL, I think this may be useful to document for new contributors (and I've found it useful for myself!)

OK, that's somewhat amusing that "black" is failing on the commit that adds the pre-commit checks... ~~Obviously it passes here, so I think this is a "black" version difference between...