Its __rlim_t is intentionally signed (__int64_t) because of legacy code that uses -1 for RLIM_INFINITY. llvm-svn: 337892