Request for review: 8009595: The UseSplitVerifier option needs to be deprecated

[harold seigel]

Hi,

We have filed a CCC to deprecate -XX:-UseSplitVerifier in JDK 8 and it 
has been approved.

Previous users of -XX:-UseSplitVerify can initially use -noverify until 
they have correct version 51 bytecodes.  Also, due to requests from 
folks who do generate version 51 bytecodes, we have added improvements 
to the new verifier error diagnostics - and specifically targeted them 
to give additional detailed information about stackmap tables.  This 
should help in the transition to generating correct version 51 class files.

Thanks, Harold


On 3/8/2013 1:37 PM, Tao Mao wrote:
> I agree that "the deprecation is not equal to the removal". If you 
> really want to remove the flag, I wonder if you should file a CCC 
> request against pm.
>
> Thanks.
> Tao
>
> On 3/8/13 5:35 AM, Jin Mingjian wrote:
>> Hi, Kris, AspectJ is not a right example for a long time[1].  But the 
>> deprecation is not equal to the removal. It is better to keep such 
>> kind option in sequential release in some way. At least, the WLS 
>> itself's componenets follow this principle. FYI:)
>> regards,
>> Jin
>>
>> [1] https://bugs.eclipse.org/bugs/show_bug.cgi?id=353467
>>
>>
>> On Fri, Mar 8, 2013 at 12:02 PM, Krystal Mo <krystal.mo at oracle.com 
>> <mailto:krystal.mo at oracle.com>> wrote:
>>
>>     There has been a workaround for folks upgrading to JDK7 that
>>     involves -XX:-UseSplitVerifier:
>>     If they were using some older version of bytecode manipulation
>>     libraries (e.g. AspectJ), then those libraries may generate class
>>     files without stackmaps but tagged as version 51 (which is
>>     required to use the new verifier).
>>
>>     That said, as a major release I believe it's okay to deprecate
>>     this flag in JDK8.
>>
>>     Thanks,
>>     Kris
>>
>>
>>     On 2013/3/7 19:04, Daniel D. Daugherty wrote:
>>
>>         > Open webrev at
>>         http://cr.openjdk.java.net/~hseigel/bug_8009595/
>>         <http://cr.openjdk.java.net/%7Ehseigel/bug_8009595/>
>>
>>         Thumbs up.
>>
>>         src/share/vm/classfile/classFileParser.cpp
>>             No comments.
>>
>>         src/share/vm/classfile/verifier.cpp
>>             So we still call the older verifier for pre-stackmap
>>         class files.
>>
>>         src/share/vm/runtime/arguments.cpp
>>             I didn't even know about this table. Cool.
>>
>>         src/share/vm/runtime/globals.hpp
>>             One less option!
>>
>>         The bug says that WLS is no longer using the flag. Is there
>>         anyone
>>         else that is using the flag? Since JDK8 is a major release do we
>>         care if anyone is using it?
>>
>>         So we have any tests that are explicitly using either
>>         -XX:+UseSplitVerifier or -XX:-UseSplitVerifier? My grep didn't
>>         find any in the HotSpot repo, but I didn't check the
>>         'jdk' repo nor the JCK tests.
>>
>>         Dan
>>
>>
>>
>>         On 3/7/13 5:52 PM, harold seigel wrote:
>>
>>             Hi,
>>
>>             Please review this small change to obsolete the
>>             UseSplitVerifier option.
>>
>>             Summary:
>>             The UseSplitVerifier option has been added to the
>>             obsolete_jvm_flags table and its use removed everywhere else.
>>
>>             Open webrev at
>>             http://cr.openjdk.java.net/~hseigel/bug_8009595/
>>             <http://cr.openjdk.java.net/%7Ehseigel/bug_8009595/>
>>             <http://cr.openjdk.java.net/%7Ehseigel/bug_8009595/>
>>
>>             Bug link at http://bugs.sun.com/view_bug.do?bug_id=8009595
>>
>>             The change was tested by hand to ensure that the
>>             deprecate message appears when the option is specified
>>             and to verify that the option had no affect.  Additional
>>             testing was done with the JCK Lang and VM, JTREG runtime,
>>             ute vm.quick.testlist, and JPRT tests.
>>
>>             Thank you!
>>             Harold
>>
>>
>>
>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/attachments/20130322/5241dc5d/attachment.html