Chris Speciale

> I'm almost certain that's more than the number of cores your machine has. Did 50 really work better than 8, 16, or however many cores you have? Its generally...

> /* Note for anyone following along: [arrays outperform linked lists in most cases](https://youthdev.net/en/performance-of-array-vs-linked-list-on-modern-computers/), because cache misses are way more important than theoretical time complexity. Always use arrays by default,...

Oooh Ok, I see the check now: https://github.com/openfl/lime/blob/69bbcaea2708567f501eaf89d29cb2591de18b79/src/lime/system/ThreadPool.hx#L341 but isn't this kind of an arbitrary limitation? If I wanted to add a job from another thread, I would need to...

> would it really be that bad if we just kind of...waited for the job to start before canceling it? Passively canceling this way might even be preferable.

> > but isn't this kind of an arbitrary limitation? > > A little bit, but I always figured it wasn't my responsibility. The old version didn't support it either...

There's no reason to hash the password on the client over TLS. ~More importantly - migrate to bcrypt. Md5 shouldn't even be used anymore.~ (Edit: Nevermind! I see you're already...

Iirc, this caused random crashes on older versions of SDL, and required recreating the window. I do believe that it is fixed in current versions though, so let me give...

No objections here. Just make the PR. :D

From closing the window? I wonder if this is related to the atExit changes.

https://github.com/openfl/lime/blob/c6cd26f699587704e1d1f9bac1487013863c94db/src/lime/_internal/backend/native/NativeApplication.hx#L181 Points to something perhaps in the native code. Or handle is already deallocated or something. This doesn't occur in 8.2? Possibly linked to any dep changes.