Charles Lane wrote:
!I wanted it simpler, but just couldn't seem to make it come out that way!
!
!Peter, your changes are just fine...with the possible exception of the
!delays. While much shorter delays are appropriate for modern fast
!machines, when running on an old uVAX3 one might need delays in ~5 sec
!increments to make sure that the tests run in the expected sequence.
OK thanks. The scenario you describe is one that I was wary of,
but unable to test. On a moderately powered laptop running Win2k
and cygwin I could change the various sleep() calls down to 1 second
and still see a 100% pass rate. If I ran the test with sleep(5) under
the cygwin time utility it took 3 min 27 sec to run. Running the same
test with s/sleep\(5\)/sleep(1)/ it took about 3 min 9 seconds to
run. Hence the savings in real time were just not worth the effort
of altering the sleep(5) duration.
!BTW, I don't expect this to go into the current release-candidate that
!Jarkko is gearing up for, there's just too large a potential for
!adding test failures in a wide variety of platforms.
I thought so too, but who knows? A lot of other tests have
recently gone in.
One other thing to consider: add the usual
BEGIN { use Config; } type trickery and skip the test
C<if $Config{'d_waitpid'} ne 'define'>. Would this
be worthwhile? Would it inadvertantly skip platforms that
have written their own version of Perl's waitpid() that
does not necessarily call a C RTL waitpid()?
Peter Prymmer
