RFR: 8181124 Get rid of compiler.testlibrary.rtm.predicate

Ekaterina Pavlova ekaterina.pavlova at oracle.com
Wed May 31 20:15:16 UTC 2017

Vladimir, thanks for prompt review.


On 5/31/17 9:39 AM, Vladimir Kozlov wrote:
> Nice. Thank you for fixing it.
> Vladimir
> On 5/30/17 2:28 PM, Ekaterina Pavlova wrote:
>> Hi all,
>> Please revive these changes which refactor compiler/rtm tests.
>> Many compiler/rtm tests use compiler.testlibrary.rtm.predicate.* predicates
>> to check if test should be executed or not. If test is not considered to be run it will
>> not do real testing and be marked as passed. It will be better to don't run such tests at all.
>> It will be more efficient from performance point of view and more accurate from reporting point of view.
>> The fix removes using of SupportedCPU, SupportedOS and SupportedVM predicates and use proper
>> "@requires" instead.
>> New vm.rtm.cpu and vm.rtm.os 'requires' properties have been implemented.
>> These change are in second webrev.
>> Local Platform.fileAsString() function was added instead of using of Utils.fileAsString()
>> to avoid full dependency on test/lib/jdk/test/lib library. It is not good but otherwise
>> we need refactor test/lib/jdk/test/lib. We agreed with Igor Ig. to postpone this refactoring.
>> Recent JDK-8180612 fix added range checks for RTMAbortRatio and RTMTotalCountIncrRate.
>> Fixed compiler/rtm/cli/TestRTMAbortRatioOptionOnUnsupportedConfig.java and
>> compiler/rtm/cli/TestRTMTotalCountIncrRateOptionOnUnsupportedConfig.java to don't pass invalid value options.
>> Otherwise these tests will fail.
>>      bug: https://bugs.openjdk.java.net/browse/JDK-8181124
>>   webrev[1]: http://cr.openjdk.java.net/~epavlova//8181124_hs/webrev.00/
>>         [2]: http://cr.openjdk.java.net/~epavlova//8181124_test/webrev.00/
>> Tested by running jprt and running compiler/rtm tests on all supported platforms.
>> thanks,
>> -katya
>> p.s.
>>   Igor Ignatyev volunteered to sponsor this change.

More information about the hotspot-compiler-dev mailing list