On Mon Mar 24 23:18:57 2025 +0000, Alfred Agrell wrote:
One could argue the ENOSYS check should be right beside the dlsym call, not beside the regular invocation. That change would remove a branch from the loop, and make it a little more clear what the code is doing; but it will add a syscall to the setup step, and make it more code in total. I have no opinion on which is better.
Okay, so it aims for forward compatibility while compiling with older arches.
I'm not sure it's convincing still, the way we usually do it is to just use syscall(2) and account for arch differences. (No, we don't support x32 ABI.)