On Tue, 21 Apr 2009, Hannah Schroeter wrote:
On Tue, Apr 21, 2009 at 07:07:13AM -0400, RD Thrush wrote:
"a" == Ariane van der Steldt <[email protected]> writes:
a> On Mon, Apr 20, 2009 at 01:57:55PM -0400, RD Thrush wrote:
[...]
The problem can be reproduced by busying each core w/ a 'nice'd
process. Then, 'make clean;time make fake' in $PORTSDIR/devel/libtool
illustrates the problem.
a> Using a make and recording the time used is useless: the most important
a> numbers (user and sys) are only recorded for the initial 'make' program,
a> not the programs it starts.
I didn't know that time(1) didn't accumulate user and sys. Thanks
for that.
It does. Ariane is wrong IMO. Only the resource usage of
children/grandchildren that weren't wait()ed for is lost.
But IIRC make usually *does* wait() for its children.
Also note this: time, the built-in command in ksh(1) reports different
numbers than time(1) (i.e. /usr/bin/time) when measuring pipelined
commands. Once, I was very surprised by this. Look at the man.
Regards,
David
