RFR(S): 8157984 - [TESTBUG] Several compiler tests fails when are executed with -XX:TieredStopAtLevel=1
vladimir.x.ivanov at oracle.com
Fri Jul 8 16:47:37 UTC 2016
>> General question: why do you fix the tests differently?
> Because TestRangeCheckSmearing can be modified to run on all tiers and
> we'll keep good test coverage.
>> + * @requires vm.opt.TieredStopAtLevel!=1 &
>> vm.opt.TieredStopAtLevel!=2 & vm.opt.TieredStopAtLevel!=3
>> Why isn't it just vm.opt.TieredStopAtLevel=4?
> because unfortunately jtreg only parses explicit vm options for
> "vm.opt.*" parameter i.e. only those which are present in comandline.
> In case vm.opt.TieredStopAtLevel == 4 is specified in requires it won't
> work in case of default vm options (so, 4th level is set by default),
> so, we have to use such construction.
Can you ignore the test if TieredStopAtLevel is overridden on
command-line? I don't see much value in supporting
> I've created v02 with testlibrary changes :
>> Best regards,
>> Vladimir Ivanov
>> On 7/6/16 7:56 PM, Dmitrij Pochepko wrote:
>>> please review fix for 8157984 - [TESTBUG] Several compiler tests fails
>>> when are executed with -XX:TieredStopAtLevel=1
>>> 3 tests failed with TieredStopAtLevl=1 option included (these tests were
>>> quarantined by
>>> 1) compiler/tiered/NonTieredLevelsTest.java modified to have respective
>>> "requires" tag
>>> 2) compiler/rangechecks/TestRangeCheckSmearing.java modified to have
>>> correct checks inside test
>>> 3) separate issue for compiler/ciReplay/TestVM_no_comp_level.sh was
>>> created so, respective @ignore tag changed.
>>> webrev: http://cr.openjdk.java.net/~dpochepk/8157984/webrev.01/
>>> CR: https://bugs.openjdk.java.net/browse/JDK-8157984
>>> I've tested these locally on linux-arm64 providing various vm flags like
>>> TieredCompilation, TieredStopAtLevel.
More information about the hotspot-compiler-dev