Request for Review JDK-8006948: Update javac for MethodParameters format change
Kumar Srinivasan
kumar.x.srinivasan at oracle.com
Fri Jan 25 11:55:34 PST 2013
On 1/25/2013 11:31 AM, Eric McCorkle wrote:
> On 01/25/13 13:54, Kumar Srinivasan wrote:
>>> On 01/25/13 13:32, Kumar Srinivasan wrote:
>>>> Eric,
>>>>
>>>>> Hello,
>>>>>
>>>>> Please review this change, which updates javac and the classfile
>>>>> library
>>>>> to reflect changes in the class file format of MethodParameters
>>>>> attributes.
>>>> What was the reason to make this change ?
>>> It was decided that the SYNTHESIZED flag would be 0x8000, not 0x10000 as
>>> it was in a previous version of the spec. Therefore, there is no reason
>>> for the flags field to be u4.
>> Right I read the discussions on the rationale now.
>>> It bears mention that the u4 causes alignment trap issues if not handled
>>> carefully.
>>>
>>>> Not true, this will break pack200. and there are tests in place for
>>>> this.
>>>> http://cr.openjdk.java.net/~ksrini/8005252/webrev.1/
>>>>
>>> What would it take to update these tests accordingly? I assume they are
>>> in TL, and thus can be addressed with a follow-up?
>> Yes.
>>
> Honestly, the pack200 tests for this should be disabled for now. The
> *unit* tests aren't enabled at this point, because the implementation is
> under active testing and modification.
>
> Should I include something in this patch to disable the test, or should
> it be done in a separate changeset?
I think what needs to happen is this:
1. run all the pack200 tests in jdk/test/tools/pack200
2. compare the failures to what has already been disabled
3. if there are new ones file a bug against pack200:
core-libs/java.util.jar
(assign it to me or you if you want to handle this, the changes are
trivial but busy work)
4. add the new test lines in #3 to jdk/test/ProblemList.txt, after a
line describing the bug
which will ultimately address it, you can see the existing pattern.
5. get this update to ProblemList.txt reviewed on core-libs using JBS
id: JDK-8006948,
I will approve it.
Thanks
Kumar
More information about the compiler-dev
mailing list