On 07/09/2013 07:11 AM, Ian Abbott wrote:
However, my version is faster (for non-floating-point time_t) if you leave out the optimization options
True, but for system builders who care about performance, it's safe to assume nowadays that optimization will be enabled and the compiler will inline the conversion function. And those who don't care about performance won't care a whit about the difference, except maybe to prefer the slower version because it makes things clearer during debugging. Come to think of it, several macros in the tz source code could be changed to functions: is_digit, _, isleap, isleap_sum, isascii, end, emalloc, erealloc, ecpyalloc, ecatalloc. Functions tend to be easier to debug than macros, and there's no real performance reason to prefer macros nowadays.