RFR(S): JDK-8145317 ReservedStackTest fails with ReentrantLock looks corrupted
David Holmes
david.holmes at oracle.com
Fri Dec 18 00:36:51 UTC 2015
On 18/12/2015 2:44 AM, Daniel D. Daugherty wrote:
> On 12/17/15 1:59 AM, Frederic Parain wrote:
>> Dan,
>>
>> thank you for the review, comment are in-lined below.
>>
>> On 12/16/2015 04:35 PM, Daniel D. Daugherty wrote:
>>> On 12/16/15 5:11 AM, Frederic Parain wrote:
>>>> Please review this small fix in ReservedStackTest test.
>>>> The test didn't exclude correctly some platforms where
>>>> the feature is not supported. The logic has been changed
>>>> to explicitly list the supported platforms.
>>>>
>>>> CR: https://bugs.openjdk.java.net/browse/JDK-8145317
>>>>
>>>> Webrev: http://cr.openjdk.java.net/~fparain/8145317/webrev/
>>>
>>> test/runtime/ReservedStack/ReservedStackTest.java
>>> L27: * @build ReservedStackTest
>>> I'm reasonably certain that the @run line takes care of
>>> building ReservedStackTest. What I meant by my comment
>>> in the bug is that you need:
>>>
>>> @build jdk/test/lib/Platform
>>>
>>> or something like it. David H is the guy to ask for
>>> what you need to build in the library stuff.
>>
>> I need some guidance here. Looking at other tests in
>> hotspot/test/runtime, none of the tests using jdk/test/lib/Platform
>> class have a build instruction, they only have the
>> "@library /testlibrary" instruction.
>
> I'm no help here. I've seen a number of different comments about
> whether the @build lines are needed and I've seen a number of
> different styles of the @build line itself.
>
> We need definitive guidance from VM/SQE and Jon G about:
>
> - the necessity of @build lines with @library lines are present
We have already had that several times. This issue has been flagged
numerous times now. Many of the tests are incorrect and may fail
unexpectedly - as we have also seen. As Ioi most recently re-reposted on
another thread:
"... I can testify that what David said is true. We have seen mysterious
nightly test failures because of the missing @build.
Also see http://openjdk.java.net/jtreg/tag-spec.html
"@library <path|jar>+
Classes in library directories are not automatically compiled. A test
that relies upon library classes should contain appropriate @build
directives to ensure that the classes will be compiled."
---
Ioi recently fixed a whole bunch of tests in this regard, and Rachel is
doing the same for the logging tests. The common form seems to be:
@build jdk.test.lib.*
David
-----
> - the preferred format of @build lines
>
> Dan
>
>
>>
>>>
>>> L202: System.exit(-1);
>>> I missed this in my original review of the test. Why
>>> use '-1' for the test exit code. For JTREG tests, I
>>> typically see '1' or some other small positive integer.
>>> Feel free to ignore this comment since any non-zero
>>> should work.
>>
>> Old habit of someone who learned programming on Unix, fixed by
>> returning '1'.
>>
>> Regards,
>>
>> Fred
>>
>>>>
>>>> Tested with JPRT, testset hotspot with aarch64 platform manually
>>>> added to the target platforms list.
>>>>
>>>> Thanks,
>>>>
>>>> Fred
>>>>
>>>
>>
>
More information about the hotspot-runtime-dev
mailing list