Request for review: 8015385 Remove RelaxAccessControlCheck for JDK 8 bytecodes

[David Holmes]

Thanks Harold - yes I misread the parentheses, sorry.

David

On 31/05/2013 11:14 PM, harold seigel wrote:
> Hi David,
>
> Thanks for the review.  Please see inline comments.
>
> Thanks, Harold
>
> On 5/31/2013 3:31 AM, David Holmes wrote:
>> Hi Harold,
>>
>> I'm not sure what the actual rules for this should be in terms of the
>> accessor and accessee having different versions. But isn't the main
>> issue the version of the accessee?
> I'm also not sure what the actual rules should be.  I chose the stricter
> path of requiring the access check if either class is >= 52 in part
> because the existing checks for pre 49 class files require both accessor
> and accessee to be < 49.  If people feel strongly I can change it.
>>
>> Further your change will now require RelaxAccessControlCheck  be true
>> for pre 49 class files, but I thought this old code was supposed to
>> continue working?
> Here's the code change using numbers instead of enums to make it easier
> to see the parentheses.  I think it shows that RelaxAccessControlCheck
> is not needed for pre 49 class files.
>
>    if (*(*RelaxAccessControlCheck && accessor_ik->major_version() < 52
> && accessee_ik->major_version() < 52*)* *||*
> ***(*accessor_ik->major_version() < 49 && accessee_ik->major_version() <
> 49*)*) {
>      return classloader_only &&
>        Verifier::relax_verify_for(accessor_ik->class_loader()) &&
>        accessor_ik->protection_domain() ==
> accessee_ik->protection_domain() &&
>        accessor_ik->class_loader() == accessee_ik->class_loader();
>    } else {
>      return false;
>    }
> }
>
>>
>> Thanks,
>> David
>>
>> On 29/05/2013 11:25 PM, harold seigel wrote:
>>> Hi,
>>>
>>> Please review this fix for bug 8015385.
>>>
>>> This fix removes the use of the RelaxAccessControlCheck option for
>>> bytecode versions >= 52 and so requires the stricter access checking
>>> that previously could be avoided by using this option. Note that the
>>> original purpose of this option was for use with pre-Java_1_5 bytecodes
>>> that did not contain the necessary accessors needed to properly access
>>> private fields.
>>>
>>> This change was tested with JCK lang and vm tests, UTE vm.quick.testlist
>>> and vm.mlvm.testlist tests, JPRT, and jtreg tests. Tests were run on
>>> Linux and Solaris.
>>>
>>> Open webrev at http://cr.openjdk.java.net/~hseigel/bug_8015385/
>>> <http://cr.openjdk.java.net/%7Ehseigel/bug_8015385/>
>>>
>>> Bug link at http://bugs.sun.com/view_bug.do?bug_id=8015385
>>>
>>> Thanks, Harold
>