NAME

sleep — do nothing, slowly

SYNOPSIS

DESCRIPTION

Sleep for the time(s) specified. Each delay is a floating-point amount of seconds, optionally suffixed with one of the following cumulative multipliers:

EXAMPLES

Wait a half-second:

Wait a year:

Wait infinitely, or until the heat death of the universe:

SEE ALSO

nanosleep(2)

STANDARDS

Conforms to IEEE Std 1003.1-2008 (“POSIX.1”), which only allows a single non-negative decimal integer delay. Multiple, floating-point delays and smhd suffixes are extensions, originating from the GNU system. wy suffixes are extensions.

HISTORY

sleep(II) ("delay execution") appeared in Version 2 AT&T UNIX:

Version 3 AT&T UNIX sleep(II) takes whole seconds, but the precision limitations remain.

Version 4 AT&T UNIX sees full-range sleep(II), and sleep(I) — "suspend execution for an interval", limited to 65536 (16 bits, unsigned, but it's more likely actually 32767, int) seconds (this is noted in the BUGS, despite being entirely expected with a 16-bit r0 on a 16-bit machine).

Version 7 AT&T UNIX replaces sleep(II) with ftime(2) and implements sleep(3), based on alarm(2), pause(2), and longjmp(3), in Standard C Library (libc, -lc), but leaves sleep unchanged.

Version 7 AT&T UNIX/32V just blindly updates the BUGS to say that the max is 2147483647 seconds. Considering this is now 68 years, rather than the 18ish hours (or, indeed, 9ish, as the case may be), this raises question about the fore-seen use-cases.

This sleep implementation persists into AT&T UNIX and the BSD (but Version 3 AT&T UNIX and 4.2BSD cool it on the longjmp(3)s in sleep(3)), and is standardised verbatim in X/Open Portability Guide Issue 2 (“XPG2”).