On Thu, 29 Jun 2023 19:51:43 GMT, Coleen Phillimore <cole...@openjdk.org> wrote:
>> src/hotspot/share/prims/jvmtiRedefineClasses.cpp line 2195:
>>
>>> 2193: case Bytecodes::_ldc:
>>> 2194: {
>>> 2195: u1 cp_index = *(bcp + 1);
>>
>> Constant pool indices are usually u2, why does this need to be a u1?
>
> This could be a u2 to avoid confusion. Since it's ldc, the cp_index in the
> ldc bytecode is only a u1 but this didn't get a Wconversion error so I should
> probably keep it as int.
> Edit: the bcp offset fetched is a u1 (byte) size, but we assign cp_index into
> new_index below so cp_index needs to be smaller than new_index. That's why I
> changed it. Making it u1 is more precise and doesn't have warnings.
I agree - using u1 to match the spec is a good thing here.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/14710#discussion_r1247322728
