On 10/09/2015 01:57 PM, Roland McGrath wrote: > Hmm. Is it the case that loff_t or off_t that we want to use will always > be 64 bits? The way we use it in the libelf interfaces always wants a > positive offset (or size). I don't think change of signedness really needs > to count as an ABI change here. Could we just use Elf64_Off and be done > with it?
It was explicitly off64_t until your commit 1a2e8f4f5bd2. And yes, as written I expect either loff_t or off_t to always be 64 bits too. (I think technically loff_t is "at least" 64 bits.) However, I'm afraid to change signedness, if only because -1 is used in error conditions, e.g. returned from elf_update(). Someone checking <0 will be in for a bad time. How about just int64_t? We already pull stdint.h for elf.h, so it seems fine to include and use that in libelf.h too.
