RFR 8245289: Clean up offset code in JavaClasses

[Lois Foltan]

On 5/28/2020 8:48 AM, coleen.phillimore at oracle.com wrote:
> Summary: Make offset member names consistent and private, move static 
> initializations near owning classes
>
> This one is better.  I gave up on the X macros because they didn't 
> save typing.  The changes here were to add underscores to offset field 
> names, remove _in_bytes for offset accessors and add asserts that the 
> offset is non-zero, made java_lang_ref_Reference offset fields 
> private, and moved static member definitions closer to the class 
> instead of some random place in javaClasses.cpp.
>
> See discussion for 8243996 Remove hardcoded field offsets for more 
> details. 
> https://mail.openjdk.java.net/pipermail/hotspot-dev/2020-May/041732.html
>
> Tested with tier1-6.
>
> open webrev at http://cr.openjdk.java.net/~coleenp/2020/8245289.01/webrev
> bug link https://bugs.openjdk.java.net/browse/JDK-8245289
>
> Thanks,
> Coleen

Hi Coleen,

Looks good.  Minor nits:

share/classfile/javaClasses.hpp
- line #239, can you add the "private" keyword back into the 
java_lang_Class class definition.  Having the "private" keyword is 
consistent with how every other class is defined in the file.
- line #887, since you are making changes to the value_offset() method 
can you change the type check for T_LONG or T_DOUBLE to call 
is_double_word_type().

I don't have a preference concerning the discussion on the 
initialization with explicit value vs. no value for the static field 
declarations.  I'm fine with pushing your patch as is.

Thanks,
Lois