RFR(S) 8205078: [Graal] org.graalvm.compiler.core.test.VerifyDebugUsageTest fails with "Expected exception: VerificationError"

[Vladimir Kozlov]

Okay. Changes are good then.

Vladimir

On 8/6/18 11:40 PM, Doug Simon wrote:
> 
> 
>> On 7 Aug 2018, at 08:07, Ekaterina Pavlova <ekaterina.pavlova at oracle.com> wrote:
>>
>> Vladimir, thanks for prompt review.
>>
>> I can revert the changes and add -XDstringConcat=inline to JAVAC flags for all tests.
>> I did it only for org/graalvm/compiler/core/test/VerifyDebugUsageTest.java because
>> Doug was not sure if this should be applied to all tests. Here what he said: "I can't say for sure since
>> we've only ever compiled the tests with JDK 8 (or --release 8). I would only apply it to this test for now.
>>
>> Doug, what is your opinion on this?
> 
> Still the same. Tests should fail fast if they cannot handle Indy strings so I think dealing with it on a case-by-case basis is the right approach.
> 
> -Doug
> 
>>
>> On 8/6/18 2:17 PM, Vladimir Kozlov wrote:
>>> Can we switch off Indy strings concat for all Graal unit tests? Based on what Doug said: "We don't support the indy string pattern in these tests".
>>> Note, Graal itself (and JVMCI) are compiled without Indy strings concat:
>>> http://hg.openjdk.java.net/jdk/jdk/file/c23f1e4910f3/make/CompileJavaModules.gmk#l425
>>> Thanks,
>>> Vladimir
>>> On 8/6/18 1:42 PM, Ekaterina Pavlova wrote:
>>>> Hi All,
>>>>
>>>> please review the change which introduces additional make target in JtregGraalUnit.gmk
>>>> to compile org/graalvm/compiler/core/test/VerifyDebugUsageTest.java with -XDstringConcat=inline.
>>>> Otherwise the test fails. Also moved building of jdk.vm.compiler.replacements.verifier.jar into separate target.
>>>>
>>>>       JBS: https://bugs.openjdk.java.net/browse/JDK-8205078
>>>>    webrev: http://cr.openjdk.java.net/~epavlova//8205078/webrev.00/index.html
>>>> testing: tested by building on all platforms and running graal unit tests on Graal supported platforms.
>>>>
>>>>
>>>> thanks,
>>>> -katya
>>>>
>>
>