VideoBackends/Vulkan: Drop BUG_BROKEN_DISCARD_WITH_EARLY_Z workaround

[OatmealDome]

Supersedes #14271. I will split the changes to the build system into a separate PR.

I also removed the nil check in GetMacOSVersion() because [NSProcessInfo processInfo] is guaranteed to never be nil.

[TellowKrinkle]

Seems reasonable. Would it be worth adding a Common/CommonFuncsObjC.mm for implementing ObjC stuff so we don't need to manually msgsend?

[OatmealDome]

That seems reasonable. Not sure how I'd implement it off the top of my head though.

[TellowKrinkle]

Not sure how I'd implement it off the top of my head though.

I think it's just adding the .mm file to CMake, and then putting the function implementation in there instead of in a .cpp file. The compiler will pick between c++ and obj-c++ by the file extension.

[OatmealDome]

Ah, that's what you mean lol

For some reason I somehow read your suggestion as creating some sort of function to make this easier:

id processInfo = reinterpret_cast<id (*)(Class, SEL)>(objc_msgSend)(
      objc_getClass("NSProcessInfo"), sel_getUid("processInfo"));

[TellowKrinkle]

Ahh no, just that if you're in a .mm you can do

const NSOperatingSystemVersion ver = [[NSProcessInfo processInfo] operatingSystemVersion];
return { ver.majorVersion, ver.minorVersion, ver.patchVersion };

which looks a bit nicer

[TellowKrinkle]

BTW since Metal and Vulkan are using the same calculation for driver version on macOS, maybe put that in some shared place and call from both?

[OatmealDome]

OK, I think this is in a good place now. I've moved the version calculation to be in DriverDetails and created CommonFuncsObjC.mm to hold the Objective-C code.

[TellowKrinkle]

Can you add the .h to CMake too? Otherwise it won't be included in cmake-generated Xcode project files.

[OatmealDome]

I didn't add any new .h files - the new func was added to CommonFuncs.h.

[TellowKrinkle]

All good, I was apparently seeing things that don't exist.