Bug: old results picked up as current when VM can't start

[Ludvig Janiuk]

I recently tried using JTREG and passing in VM flags for the Java Hotspot VM. I used a set of flags that would lead to the VM failing to start. I did this in order to ensure that JTREG was actually passing my flags properly. I expected the test to fail. To my surprise and great confusion, I was presented with "PASSED". This is a clear bug. In my limited understanding, the cause seems to be that JTREG erroneously picks up results from earlier runs and presents them when this happens. So I am reporting this to you in the hope that the bug will be fixed.

To reproduce, the following should probably roughly work:

  1.  Run an arbitrary test in JTREG, observe that it is passing
  2.  Run it again, but this time, pass e.g: OPTIONS="-Xms3g -Xmx2g" which is a contradiction and should not let the JVM start
  3.  You should see "pass" instead of an expected "error"

Best regards,
//Ludvig Janiuk
Oracle JPG JVM Sustaining Engineering