RFR (S) 8236224: Obsolete the FieldsAllocationStyle and CompactFields options

[coleen.phillimore at oracle.com]

On 1/7/20 12:14 AM, David Holmes wrote:
> Hi Coleen,
>
> On 7/01/2020 1:25 pm, coleen.phillimore at oracle.com wrote:
>> Summary: Remove the options and code for options deprecated in JDK 14
>
> Generally looks good.
>
>> open webrev at 
>> http://cr.openjdk.java.net/~coleenp/2019/8236224.01/webrev
>> bug link https://bugs.openjdk.java.net/browse/JDK-8236224
>
> src/hotspot/share/aot/aotCodeHeap.hpp
>
>   typedef struct {
> !   enum { CONFIG_SIZE = 7 * jintSize + 9 };
>     // 8 int values
>
> Now 7 int values
>
>     // byte[11] array map to boolean values here
>
> Now byte[10]. Or should that be byte[9]? I think the original code may 
> be off by one.

Yes, it was wrong. I fixed the comments.
>
> ---
>
> src/hotspot/share/classfile/classFileParser.cpp
>
> 4133   bool allocate_oops_first = false; // was allocation_style == 0
>
> The comment has no context now that there is no selectable allocation 
> style.
>
Removed.  It was mostly to remind myself.
> I don't understand why you removed a bunch of classes from this check:
>
> 4143       (_class_name == 
> vmSymbols::java_lang_AssertionStatusDirectives() ||
> 4144        _class_name == vmSymbols::java_lang_Class() ||
> 4145        _class_name == vmSymbols::java_lang_ClassLoader() ||
>
> 4147        _class_name == vmSymbols::java_lang_ref_SoftReference() ||
> 4148        _class_name == vmSymbols::java_lang_StackTraceElement() ||
> 4149        _class_name == vmSymbols::java_lang_String() ||
> 4150        _class_name == vmSymbols::java_lang_Throwable() ||
>
> ??

The classes removed no longer have hardcoded offsets so did not need to 
follow the oops-first allocation style.  This was not cleaned up when 
the hardcoded offsets were removed from these classes.   Fred also fixes 
this with his field layout patch in perhaps another place.

Thanks,
Coleen
>
> ---
>
> Thanks,
> David
>
>> Ran tier1 on all oracle platforms, and 2, 3 on 
>> linux/windows-x64-debug and hs-tier4-graal because there were jvmci 
>> changes.
>>
>> thanks,
>> Coleen