Francesco

Yes, there are a **lot** of missing things. My implementation was actually like a "seed" to spark the interest and start working on that. In addition I only barely know...

I suspect that we could have a lot of regressions in the new directory monitoring because we completely eliminated the old code that I had tested in a very comprehensive...

I tested on Windows using my MinGW64 build toolchain. The function CancelIoEx is not declared apparently so I had to add to declare explicitly: ```c++ extern BOOL WINAPI CancelIoEx(HANDLE hFile,...

As a side note: with the previous release any change in the filesystem was immediately visible in the TreeView. Now with the new implementation this is no longer the case....

Oh I forgot to say everything is all right on linux :-)

> What if we added core.add_background_thread, and had certain lua coroutines which just always ran, regardless of focus? Well, that would be a regression because it means the application will...

> Can you confirm that it's building with the appropriate back-end, and the back-end is engaged? I confirm, meson prints a message that for dirmonitor it will use the win32...

> This is odd. I am also on the MinGW 64-bit toolchain, and that declaration wasn't necessary. What version are you using? Ok, I updated mingw64 to the latest version...

> Well, on 2.0.5, we do still use CPU when we're not focused That's not supposed to be the case. I am going to check on Linux but normally I...

This problem was fixed in SDL 2.0.16 with a modification that I contributed but it works only if SDL is compiled without joystick and sensor support or if no joystick...