RFR 8224137: Analyze and port invocation tests to jtreg and co-locate to jdk repo

David Holmes david.holmes at oracle.com
Sat Jun 22 01:18:11 UTC 2019


Hi Harold,

Not a review ...

On 21/06/2019 10:55 am, Harold Seigel wrote:
> Hi Coleen,
> 
> Can you take one more look?
> 
>     http://cr.openjdk.java.net/~hseigel/bug_8224137.3/webrev/
> 
> I changed three files, invokespecialTests.java, 
> invokeinterfaceTests.java, and invokevirtualTests.java to explicitly 
> specify -Xint and -Xcomp to the sub-tests.

If you want to check all the compilation environments you need to be 
more elaborate than just using -Xcomp. You may need to test C1 only and 
C2 only. And then you need to consider Graal as well.

IIUC from a brief look at these tests they will never pass through the 
VM flags given to jtreg - is that correct?

Thanks,
David

> Thanks again!
> 
> Harold
> 
> On 6/20/2019 6:37 PM, coleen.phillimore at oracle.com wrote:
>> Great, thanks. Looks good!
>> Coleen
>>
>> On 6/20/19 1:00 PM, Harold Seigel wrote:
>>> Hi Coleen,
>>>
>>> Thanks for reviewing this change!
>>>
>>> Please review this updated webrev:
>>>
>>> http://cr.openjdk.java.net/~hseigel/bug_8224137.2/webrev/index.html
>>>
>>> The -XDignore.symbol.file has been removed and TEST.groups has been 
>>> changed.  Otherwise, it is the same as the previous webrev.
>>>
>>> Mgmt said to leave the copyright years as 2009, 2019, so I did not 
>>> change those.
>>> Thanks! Harold
>>>
>>> On 6/19/2019 5:41 PM, coleen.phillimore at oracle.com wrote:
>>>>
>>>> http://cr.openjdk.java.net/~hseigel/bug_8224137/webrev/test/hotspot/jtreg/runtime/InvocationTests/invokeinterfaceTests.java.html 
>>>>
>>>>
>>>>   35  * @compile -XDignore.symbol.file invokeinterface/Checker.java 
>>>> invokeinterface/ClassGenerator.java
>>>>
>>>>
>>>> Do these still need -XDignore.symbol.file ?
>>>>
>>>> http://cr.openjdk.java.net/~hseigel/bug_8224137/webrev/test/hotspot/jtreg/TEST.groups.udiff.html 
>>>>
>>>>
>>>> Can you specify all the tests in the directory by directory? like:
>>>>
>>>> + -runtime/InvocationTests \
>>>>
>>>>
>>>> http://cr.openjdk.java.net/~hseigel/bug_8224137/webrev/test/hotspot/jtreg/runtime/InvocationTests/shared/Caller.java.html 
>>>>
>>>>
>>>>    2  * Copyright (c) 2009, 2019, Oracle and/or its affiliates. All 
>>>> rights reserved.
>>>>
>>>>
>>>>
>>>> Some of the copyrights say 2009, which is when the tests were 
>>>> written but I think we're supposed to have the original date when 
>>>> they're added to the repository.
>>>>
>>>> Well done getting these tests into jtreg and the repository!
>>>> Thanks,
>>>> Coleen
>>>>
>>>> On 6/17/19 2:57 PM, Harold Seigel wrote:
>>>>> Hi,
>>>>>
>>>>> Please review this JDK-14 change to move the invocation tests 
>>>>> written by Vladimir Ivanov into the JDK repo and make them runnable 
>>>>> using JTReg.
>>>>>
>>>>> This webrev adds three tests, invokeinterfaceTests.java, 
>>>>> invokespecialTests.java, and invokevirtualTests.java.  Each tests 
>>>>> run its set of sub-tests twice, once using class file version 51 
>>>>> and once using the current class file version.
>>>>>
>>>>> Open Webrev: http://cr.openjdk.java.net/~hseigel/bug_8224137/webrev/
>>>>>
>>>>> JBS Bug: https://bugs.openjdk.java.net/browse/JDK-8224137
>>>>>
>>>>> The tests were tested on Linux-x64, Solaris, Windows, and Mac OS X.
>>>>>
>>>>> The original tests can be found attached to JDK-8163974 
>>>>> <https://bugs.openjdk.java.net/browse/JDK-8163974>. Besides the 
>>>>> changes needed for JTReg and adding copyrights, I made the 
>>>>> following additional changes.
>>>>>
>>>>> 1. The tests now use the JDK's asm support instead of providing its 
>>>>> own
>>>>>    asm libraries.
>>>>> 2. Only sub-test failures are written to the .jtr files. Writing all
>>>>>    sub-test results caused JTReg to truncate the output.
>>>>> 3. Changed src/invokeinterface/Checker.java to skip private methods
>>>>>    when looking for an overloading method.
>>>>>
>>>>> The tests contain "TODO" comments and other thing needing 
>>>>> clean-up.  These will be addressed in a future RFE.
>>>>>
>>>>> I put the tests into hs-tier3 because the invokeInterfaceTests.java 
>>>>> test can run for up to 10 minutes (on Mac).  The other two tests 
>>>>> take only 1-2 minutes.  Is there a better tier for these tests?
>>>>>
>>>>> Thanks, Harold
>>>>>
>>>>
>>


More information about the hotspot-runtime-dev mailing list