I'm not sure if this is flak or not, but I'm not sure that I understand what this script is trying to do.

Just using POSIX-defined features in your shell script does not mean that your shell script will give you the same results on all POSIX-conforming systems.

If the idea is to sleep for a particular number of seconds and milliseconds on OS X version 10.7.5 with the particular amount and type of memory present on your system with the same CPU you're using and with the same disk drive you're using under similar load conditions, the code above might come close to doing what you want. But, for small sub-second sleeps, the amount of time spent forking and execing awk and sleep may overwhelm your other calculations. And, if you invoke this script several times, the first time it runs is likely to be slower than subsequent runs since awk, sleep, and this script will need to be loaded from disk the first time it runs, while subsequent runs in the not too distant future can be loaded from the cache.

So, although the way you are performing the calculations in this script is portable, the calculations you are performing are not portable to any other hardware and software configuration because what you are really calculating is how long it will take to fork(), exec sleep, run sleep 0, and waitpid() for the exit status from the sleep 0. All of this is very dependent on your hardware and is not portable to any different configuration. Note also that the wait in your loop is a relatively quick no-op. The wait would only take any significant amount of time if your script had started background jobs that had not previously been waited for. (Your script never starts any background jobs, so all the wait built-in does is loop through the zero background jobs running in the current shell waiting for each one to complete before it returns a zero exit status.)

Although:

Code:

sleep sss.mmm

where sss is a given number of seconds and mmm is a given number of milliseconds to sleep is not portable between various systems, it is portable and will give you much more consistent sleep times between various versions of hardware running various versions of OS X (at least for releases of OS X greater than or equal to 10.0). It will also work on other BSD or Linux based systems and, if you use a 1993 or later ksh instead of bash, on any system using that shell. (ksh93 has a built-in sleep that accepts fractional seconds; bash version 3.2.57 does not have a built-in sleep on OS X.)

If you want a portable way to sleep for a given number of seconds and nano-seconds in a shell script that will work on any POSIX-conforming system, you'll need to write your own nanosleep utility in C that calls the nanosleep() function with the seconds and nanoseconds specified by operands given when it is invoked and then call that utility from your script. Or you can configure your script to invoke the system's sleep utility if it supports sub-second sleeps, or invoke your nanosleep utility otherwise.