RFR 8029567: Clean up linkResolver code
Coleen Phillimore
coleen.phillimore at oracle.com
Fri May 29 21:34:03 UTC 2015
#/bin/csh -f
hg status | grep "^M" | sed -e "s/M//" >file.list
foreach f (`cat file.list`)
echo $f
sed -e "s/Copyright (c) \([0-9][0-9][0-9][0-9]\),
[0-9][0-9][0-9][0-9], Oracle/Copyright (c) \1, 2015, Oracle/" \
-e "s/Copyright (c) \([0-9][0-9][0-9][012346789]\),
Oracle/Copyright (c) \1, 2015, Oracle/" <$f >$f.new
diff $f $f.new
mv $f.new $f
end
hg commit
On 5/29/15 3:21 PM, Yumin Qi wrote:
> Thanks. I would think if JPRT integrated with your commit script so we
> don't care about the file header updating.
>
> On 5/29/2015 11:57 AM, Coleen Phillimore wrote:
>>
>> Hi Yumin,
>> My 'commit' script automatically updates copyright years. I should
>> have mentioned that.
>> Thanks,
>> Coleen
>>
>> On 5/29/15 2:25 PM, Yumin Qi wrote:
>>> Coleen,
>>>
>>> *src/share/vm/ci/ciMethod.cpp*
>>> *src/share/vm/classfile/loaderConstraints.cpp
>>> **src/share/vm/classfile/loaderConstraints.hpp
>>> **src/share/vm/interpreter/bytecode.cpp*
>>> *src/share/vm/interpreter/linkResolver.hpp*
>>> *src/share/vm/runtime/javaCalls.cpp
>>>
>>> *Those files need copyright 'year' to be updated.
>>>
>>> Thanks
>>> Yumkin
>>> On 5/28/2015 3:34 PM, Coleen Phillimore wrote:
>>>>
>>>> The change (jck testing in progress) with the loader constraint
>>>> changes backed out.
>>>>
>>>> open webrev at http://cr.openjdk.java.net/~coleenp/8029567.03/
>>>>
>>>> Thanks,
>>>> Coleen
>>>>
>>>> On 5/28/15 5:59 PM, Coleen Phillimore wrote:
>>>>>
>>>>> Hi Karen,
>>>>>
>>>>> Yes, I'll back out the loader constraint changes. I was trying to
>>>>> merge the error message construction code for failing loader
>>>>> constraint checks since it's similar code in 5 places but I gave
>>>>> up, and made the CLD change as a side effect.
>>>>> We store ClassLoaderData in the loader constraint table now but
>>>>> this is a good find, because I think there is a behavior change
>>>>> here in the comparison.
>>>>>
>>>>> Thanks!
>>>>> Coleen
>>>>>
>>>>> On 5/28/15 5:43 PM, Karen Kinnear wrote:
>>>>>> Coleen,
>>>>>>
>>>>>> 1. Thank you for highlighting where we have check_access as false
>>>>>> - so we can make sure we
>>>>>> really really meant to do that going forward.
>>>>>>
>>>>>> 2. I am concerned about the loader constraint changes to use
>>>>>> class loader data rather than class loaders.
>>>>>> Since these are not 1:1 - what about use cases in which the class
>>>>>> loaders are the same but the loader datas are not?
>>>>>>
>>>>>> In the Valhalla exploration of ClassDynamic, we are exploring
>>>>>> having dynamic classes, which would be defined
>>>>>> via DefineAnonymousClass - which can be used in signatures, so we
>>>>>> would need to be able to compare types which may
>>>>>> have different loader datas but might need to compare as
>>>>>> equivalent since they have the same class loader.
>>>>>>
>>>>>> I believe the intention of this set of changes is cleanup - not
>>>>>> meant to change behavior.
>>>>>>
>>>>>> But that change actually is a meaningful behavior change - and I
>>>>>> think it is premature to make it.
>>>>>>
>>>>>> Could you possibly back out the loader constraint related changes?
>>>>>>
>>>>>> Those will take more thought, more future designs, and detailed
>>>>>> targeted tests.
>>>>>>
>>>>>> thanks,
>>>>>> Karen
>>>>>>
>>>>>> On May 14, 2015, at 7:16 PM, Coleen Phillimore wrote:
>>>>>>
>>>>>>> Summary: Moved non-const reference return values to actual
>>>>>>> return values, refactored error handling code, pass CLD rather
>>>>>>> than class_loader oop, remove oop from Method* variable names.
>>>>>>>
>>>>>>> Also modified the long parameter lists into a class LinkInfo to
>>>>>>> hold information from the constant pool that is always passed
>>>>>>> through several layers of functions together. Also reformatted
>>>>>>> to split parameter list lines to some reasonable width.
>>>>>>>
>>>>>>> The type methodHandle should be passed as const reference types
>>>>>>> to avoid copy construction, because it has a non-trivial
>>>>>>> destructor. This sort of change could be made in more places in
>>>>>>> the JVM, but I stopped with linkResolver.
>>>>>>>
>>>>>>> Ran all hotspot jtreg tests, jck lang/vm/api/java_lang tests,
>>>>>>> internal testbase tests: vm.quick.testlist, vm.defmeth.testlist,
>>>>>>> proposed new selection-resolution tests, jdk/java/lang/invoke
>>>>>>> tests (see RFR).
>>>>>>>
>>>>>>> open webrev at http://cr.openjdk.java.net/~coleenp/8029567/
>>>>>>> bug link https://bugs.openjdk.java.net/browse/JDK-8029567
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Coleen
>>>>>
>>>>
>>>
>>
>
More information about the hotspot-runtime-dev
mailing list