RFR 8215729: Enhance makefiles to allow collecting code coverage with JCov

[David Holmes]

Hi Shura,

Sorry not a review just a comment on one thing ...

On 9/01/2019 4:35 am, Alexandre (Shura) Iline wrote:
> Hi,
> 
> Could you please take a look on a change which allows to run tests while collecting code coverage with JCov. This is a continuation of work done in
> JDK-8214309: Enhance makefiles to allow generating JCov instrumented build.
> 
> This adds make targets jcov-test and others like jcov-test-tier1 etc. After running the tests, one is left with a coverage report and also a data file containing coverage, for further analysis.
> 
> Part of the changes are related to increasing maximum and initial heap size, which is needed to be done because instrumented classes are bigger in size. The way it is done for JTReg tests is by setting both _JAVA_OPTIONS and JAVA_TOOL_OPTIONS environment variables.
> 
> As it stands now, not all tests pass while running with code coverage, which pass normally. This particular enhancement is not addressing failing tests. The tests will need to be fixed separately if ever. For example, there are 35 tests which fail in open/test/jdk:jdk_core. A quick look tells that the majority of the failures are in tests which verify error output and discover an unexpected line "Picked up _JAVA_OPTIONS: -Xmx4g”.

We've been bitten in the past with VM warnings causing tests to fail due 
to the unexpected output and made a conscious effort to fix all such 
tests. Unfortunately a few still remain. Do your 35 failures match up 
with the tests in the blocking issues for this bug:

https://bugs.openjdk.java.net/browse/JDK-8196741

? 35 seems a lot. I hope we haven't regressed in this area.

Thanks,
David

> 
> Enhancement: https://bugs.openjdk.java.net/browse/JDK-8215729
> Webrev: http://cr.openjdk.java.net/~shurailine/8215729/webrev.01/
> 
> Shura
>