Is there any coverage trade off on setting the clover.pertest.coverage=diff ?

[Pedro Garcia Cortina]

Hi, I followed the recommenations for CPU intensive tests mentioned in the http://confluence.atlassian.com/display/CLOVER/Clover+Performance+Tuning#CloverPerformanceTuning-RuntimePerformanceSettings. I would like to know if there's any trade off on using that setting in terms of coverage?

Thanks in advance

Pedro

[Marek Parfianowicz · Accepted solution]

Hi Pedro,

There's no trade-off in terms of functionality.

There are very subtle differences related with JVM implementations (IBM Java vs Oracle Java vs OpenJDK) regarding threads, locks, volatile variables and data race conditions. You can read "JSR 133: Java Memory Model and Thread Specification Revision" if you like such stuff.

For this reason different strategies have been implemented in Clover, especially for running multi-threaded tests (i.e. when multiple test methods are executed at once; not to be mislead with sequential tests of multi-threaded applications). For example:

• single-threaded vs multi-threaded per-test recorders
• using synchronized blocks vs volatile variables

Regards
Marek