[PATCH] wcscasecmp does exist on OS X Lion

[historic_bruno]

Contrary to #414, wcscasecmp is now available on OS X Lion, causing many redundant redeclaration warnings during the build process. I haven't tried Apple's wcscasecmp to know if it will work for us, but at least we can get rid of these warnings. The same goes for strnlen, wcsnlen, and wcsdup.

[Markus]

patch from #2005 / RedFox

[historic_bruno]

I think it's not the correct solution even if it solves the build warning on 10.7+. The problem is the game can be built on 10.7 but run on earlier versions, which could result in a crash when it tries to call those functions, if they're not implemented. So the proper fix involves two things:

• Compile time fix: if we build against 10.6 SDK or earlier, it should use our implementation or the build would fail, the patch solves that in a hacky way, it should use OS X version macros. This way it also runs on 10.7+ with no problem.
• Run-time fix: if we build against 10.7+ SDK, it should detect if the functions are valid before calling them, if not, fall back to our own implementation. This would be the ugly part.

Honestly it seems easier to always use our own implementation on OS X and just fix the build warnings for 10.7+.

[Markus]

If we want the build to run with 10.6 we should set the deployment target accordingly. Linking to 10.7 and then check at runtime if the function exists looks like the wrong way.

(I think we should always build with the newest version, the packaged downloads can still be linked to 10.6.)

#if OS_MACOSX && !defined(AVAILABLE_MAC_OS_X_VERSION_10_7_AND_LATER) ?

[historic_bruno]

Replying to Markus:

If we want the build to run with 10.6 we should set the deployment target accordingly.

In fact, that's exactly what we do with the OS X bundle build script :)

Linking to 10.7 and then check at runtime if the function exists looks like the wrong way.

According to Apple, it's one aspect of OS X ​SDK-based development. But as I said it's probably overkill in this case, we could squash the build warning on 10.7+ and use our own implementation. I actually don't know what the compiler's doing now, is it undefined behavior?

[wraitii]

OS 10.7 is now super old, might be time to reconsider.

[elexis]

Backlogging due to lack of progress.

[historic_bruno]

We no longer support 10.7 (SM38 requires 10.9), so older builds aren't relevant. Will revisit this as such.

[wraitii]

I fixed strnlen in ​Phab:rP24227, not the others though, gonna push this to A25.

[Stan]

Unlikely to get done for A26