RFR (XS): 8209689: Compiler.isGraalEnabled should not check jvmci.Compiler property

Christian Thalinger cthalinger at twitter.com
Tue Aug 21 15:49:13 UTC 2018


Thanks.  Testing was clean; I’ll push.

> On Aug 20, 2018, at 10:20 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
> 
> Agree.
> 
> Checking UseJVMCICompiler flag and tiered level should be enough.
> 
> Thanks,
> Vladimir
> 
> On 8/20/18 5:50 AM, Christian Thalinger wrote:
>> https://bugs.openjdk.java.net/browse/JDK-8209689
>> I noticed that some tests are run in our Graal nightlies although the @requires explicitly states:
>> !vm.graal.enabled
>> Turns out the code in Compiler.isGraalEnabled still checks for the jvmci.Compiler property but this property is not set and is null by default.
>> (A bit more information in the ticket.)
>> diff --git a/test/lib/sun/hotspot/code/Compiler.java b/test/lib/sun/hotspot/code/Compiler.java
>> --- a/test/lib/sun/hotspot/code/Compiler.java
>> +++ b/test/lib/sun/hotspot/code/Compiler.java
>> @@ -55,11 +55,6 @@
>>          if (useJvmciComp == null || !useJvmciComp) {
>>              return false;
>>          }
>> -        // This check might be redundant but let's keep it for now.
>> -        String jvmciCompiler = System.getProperty("jvmci.Compiler");
>> -        if (jvmciCompiler == null || !jvmciCompiler.equals("graal")) {
>> -            return false;
>> -        }
>>          Boolean tieredCompilation = WB.getBooleanVMFlag("TieredCompilation");
>>          Long compLevel = WB.getIntxVMFlag("TieredStopAtLevel");



More information about the hotspot-compiler-dev mailing list