RFR: 8345678: compute_modifiers should not be in create_mirror

[Coleen Phillimore]

On Mon, 9 Dec 2024 04:22:54 GMT, David Holmes <dholmes at openjdk.org> wrote:

>> This moves the modifier_flag computation to when InstanceKlass, ObjArrayKlass and TypeArrayKlass are created.
>> 
>> Tested with jck:vm and tier1-4.
>
> src/hotspot/share/classfile/javaClasses.cpp line 1119:
> 
>> 1117:   // to support Class.getModifiers().  Instance classes recalculate
>> 1118:   // the cached flags after the class file is parsed, but before the
>> 1119:   // class is put into the system dictionary.
> 
> Is this comment already out-of-date? I'm trying to see where the flags are recomputed after parsing.

I thought the comment was referring to this code below that recalculates the modifier flags.  Don't know why system dictionary was called out though.

> src/hotspot/share/oops/typeArrayKlass.hpp line 76:
> 
>> 74:   void  copy_array(arrayOop s, int src_pos, arrayOop d, int dst_pos, int length, TRAPS);
>> 75: 
>> 76:   // jvm support
> 
> "jvm support" for what? I'm not even sure why we need `compute_modifier_flags` for array types when they have a fixed value. Can't we just hardwire them and only call `compute_modifier_flags` for instanceKlasses?

ObjArrayKlass computes the modifier flags based on the bottom_klass, and there is a place that calls compute_modifier_flags in an assert that requires the virtual function, albeit for an assert, otherwise I would have made it not a virtual function.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/22618#discussion_r1875946187
PR Review Comment: https://git.openjdk.org/jdk/pull/22618#discussion_r1875948835