NB: this distro is upstream-CPAN, but there hasn't been a new release in
7 years, so I'm patching this intermittently false positive test
directly in blead.
I reported this issue 4 years ago as
https://rt.cpan.org/Public/Bug/Display.html?id=108382
cpan/Memoize/t/speed.t occasionally fails tests 2 and 5 in bleadperl
smokes.
This is because the time deltas (using "time") have a granularity of 1
sec. The basic test is: run fib() for at least 10 secs, then run again
with memoize and check that it takes less than 1/10th of that time.
However, the first measurement may be exactly 10 secs, while the second
run (no matter how speedy) may be 1 sec if it passes over a tick
boundary. 0.001 is then added to it to avoid division by zero.
The test is ($ELAPSED/$ELAPSED2 > 10). With $ELAPSED = 10 and
$ELAPSED2 = 1.001, the test fails. The easy fix is to run for at least
11 secs rather than 10.
$ELAPSED = 0;
-my $LONG_RUN = 10;
+my $LONG_RUN = 11;
while (1) {
my $start = time;
$start = time;
$RESULT2 = fib($N);
$ELAPSED2 = time - $start + .001; # prevent division by 0 errors
-
print (($RESULT == $RESULT2) ? "ok 1\n" : "not ok 1\n");
# If it's not ten times as fast, something is seriously wrong.
+print "# ELAPSED2=$ELAPSED2 seconds.\n";
print (($ELAPSED/$ELAPSED2 > 10) ? "ok 2\n" : "not ok 2\n");
+
# If it called the function more than $N times, it wasn't memoized properly
print (($COUNT > $N) ? "ok 3\n" : "not ok 3\n");
$start = time;
$RESULT2 = fib($N);
$ELAPSED2 = time - $start + .001; # prevent division by 0 errors
-
print (($RESULT == $RESULT2) ? "ok 4\n" : "not ok 4\n");
+print "# ELAPSED2=$ELAPSED2 seconds.\n";
print (($ELAPSED/$ELAPSED2 > 10) ? "ok 5\n" : "not ok 5\n");
# This time it shouldn't have called the function at all.
print ($COUNT == 0 ? "ok 6\n" : "not ok 6\n");