On Mon, 16 Mar 2026 15:58:12 GMT, Anton Artemov <[email protected]> wrote:

>> src/hotspot/cpu/x86/templateInterpreterGenerator_x86.cpp line 1562:
>> 
>>> 1560:     // If member name is not null and is restored, it has to be 
>>> reflected, as if no
>>> 1561:     // MemberName was popped out of the stack when going through the 
>>> methods handle
>>> 1562:     __ decrement(rbcp, wordSize);
>> 
>> I don't understand what this is doing.  I would expect setting BCP to BCP-8 
>> to cause a crash.
>
> `rbcp` happens to be `r13`, which is touched in methodHandles on this 
> platform.

That may be true, but starting at the   __ restore_bcp() at line 1528, and 
required by InterpreterRuntime::member_name_arg_or_null, this register must be 
the actual BCP.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/29744#discussion_r2942894885

Reply via email to