Re: [Xmlunit-general] Diff.similar() seems to cause thread to exit 50%+ of the time

[Eric S. <pub...@da...>]

On 10/23/2013 03:31 PM, Bill Ross wrote:
 > It looks like this is an artifact of println rather than XMLUnit: 
when I add this to the end of my test, it triggers ok:
 > System.out.println("@@@");
 > //napMilliSeconds(2000);
 > Assert.fail("Should fail here.");
 >
 > With the ‘nap’, all the prints appear, but in any case the assert is 
reached and fires.

So, are you saying that:
- with the nap, you get both the System.out output and the "Should fail 
here" test failure, as expected?
- without the nap, you get the (still-expected) test failure, but 
nothing printed on System.out?

If so, it sounds as though, when the test finishes (whether failing, as 
here, or passing, as in your original post), System.out's buffer is 
discarded instead of being flushed. I'm not sure why that would be 
timing-dependent.

Try replacing the nap with:
System.out.flush();
and see what happens.

Also try removing XmlUnit from the equation entirely.

I'd do both of those at once by making a set of tests that do:
{XmlUnit code | nothing}
println // The other 12 combinations, without the println, are of no 
interest
{nap | flush | nothing}
{fail | nothing}

And see which ones do or do not produce the println output.

If, as I expect:
- nap and flush both cause the println output to appear, but with 
neither of them, it's suppressed
- presence or absence of the preceding XmlUnit stuff makes no difference
- presence or absence of the final Assert.fail call also makes no difference
then you have a solid bug report for the IntelliJ folks :-)

(Of course, it being timing-dependent, the results may not be entirely 
identical from run to run. I'd expect nap (and flush, if my theory is 
correct) to *reliably* produce the output, but having neither of them to 
only sometimes suppress it.)

- Eric