Different implementation of atomic operation abstractions for gcc and clang?

[meyerj]

Today I stumbled across 94e3053, which added a preprocessor directive and switches to different implementations for atomic operations that are essential for lock-free buffer implementations.

I am not aware of the pros and cons of using the builtin intrinsics of gcc/clang compared to the custom platform-dependent implementations. But it would be worth to verify if Clang 3 and newer versions are still affected by the problem which triggered that patch back in 2011. I could not find any background information in the commit history, the Orocos forums or the old bug tracker.

[snrkiwi]

I don't think that this is a problem with modern Clang, at least on OS/X or macOS.