30 Aug
2008
30 Aug
'08
10:18 a.m.
Vincent Massol wrote:
Hi Sergiu,
This test that you wrote has failed on Hudson:
public void testTimeComplexity()
{
ArrayList<String> tests = new ArrayList<String>();
ArrayList<String> expects = new ArrayList<String>();
// Something like this should be (negatively) matched in
linear time, thus it should take no
// time. If the build takes a lot, then the regular
expression is not in linear time, thus
// wrong.
String text = "*";
for (int i = 0; i < 1000; ++i) {
text += "abc *";
}
tests.add(text);
expects.add(text);
long startTime = System.currentTimeMillis();
test(tests, expects);
// Even on very slow systems this should not take more than
one second. Putting 10 seconds,
// just to be sure we don't get false test errors.
assertTrue(System.currentTimeMillis() - startTime < 10000);
}
Result here:
http://hudson.xwiki.org/job/xwiki-platform-core-trunk/com.xpn.xwiki.platfor…
So it seems it took more than 10 seconds.
Does it mean the code is not working?
No, since locally it takes less than a second. It either means that the
build machine is really really slow, or that some environment components
behave differently (like a different JVM version, or a regexp
implementation that overrides the JDK one). Also, it seems that the test
passes most times with less than 1 second (see
http://hudson.xwiki.org/job/xwiki-platform-core-trunk/com.xpn.xwiki.platfor…
), and fails sometimes with a huge running time (see
http://hudson.xwiki.org/job/xwiki-platform-core-trunk/com.xpn.xwiki.platfor…
).
What should we do?
No idea... We could increase the time limit to 60 seconds to let the
test pass, but we should find out why is this happening.
--
Sergiu Dumitriu
http://purl.org/net/sergiu/