Change "7902090: Update JDK_Version to be more future-proof" breaks some jtreg tests
Jonathan Gibbons
jonathan.gibbons at oracle.com
Fri Feb 23 00:10:05 UTC 2018
Volker,
Yes, I'd forgotten that specific use case of JDK_Version.forName, and
regrettably agree that the method should agree with javac. I see javac
has (finally) started restricting the version name in 11.
-- Jon
On 02/20/2018 08:07 AM, Jonathan Gibbons wrote:
> Volker,
>
> Thanks for the report.
>
> 1.9 is not an official version number, but neither is NPE the intended
> effect. I will check out the code to see the possibilities.
>
> -- Jon
>
>
> On 2/20/18 8:01 AM, Volker Simonis wrote:
>> Hi Johnathan,
>>
>> it seems that your change "7902090: Update JDK_Version to be more
>> future-proof" breaks some jtreg tests. Specifically, all tests which
>> set "-source 1.9 -target 1.9" for the compile step (e.g.
>> test/jdk/java/lang/String/concat/ImplicitStringConcatArgCount.java
>> ) will fail with a null pointer exception:
>>
>> ----------Stack trace:(9/649)----------
>> java.lang.NullPointerException
>> at
>> com.sun.javatest.regtest.exec.CompileAction.isModuleOptionsAllowed(CompileAction.java:525)
>> at
>> com.sun.javatest.regtest.exec.CompileAction.getJavacCommandArgs(CompileAction.java:469)
>> at
>> com.sun.javatest.regtest.exec.CompileAction.run(CompileAction.java:357)
>> at
>> com.sun.javatest.regtest.exec.RegressionScript.run(RegressionScript.java:247)
>> at com.sun.javatest.Script.run(Script.java:241)
>> at
>> com.sun.javatest.DefaultTestRunner.runTest(DefaultTestRunner.java:174)
>> at
>> com.sun.javatest.DefaultTestRunner.access$100(DefaultTestRunner.java:43)
>> at
>> com.sun.javatest.DefaultTestRunner$1.run(DefaultTestRunner.java:66)
>> result: Not run. Test running...
>>
>> So the new behavior doesn't really seem to be more "future-proof" :)
>>
>> Do you plan to fix this in jtreg or do you want to change all the
>> jtreg tests which use "1.9" to use "9" instead? I personally think
>> that at least as long as "javac" supports both notations, jtreg should
>> support them as well.
>>
>> Thanks,
>> Volker
>
More information about the code-tools-dev
mailing list