[lworld] RFR: 8376135: [lworld] Add JIT support for NULLABLE_NON_ATOMIC_FLAT layout [v4]

[Quan Anh Mai]

On Fri, 30 Jan 2026 07:51:27 GMT, Tobias Hartmann <thartmann at openjdk.org> wrote:

>> src/hotspot/share/ci/ciInlineKlass.cpp line 161:
>> 
>>> 159:   VM_ENTRY_MARK
>>> 160:   InlineKlass* vk = get_InlineKlass();
>>> 161:   oop null_reset_value = vk->null_reset_value();
>> 
>> The null_reset_value instance is allocated at class initialization time. This means that  there's a time window, between the moment the class is loaded and the moment where it is initialized, where the null_reset_value is not available yet. During this time interval, null_reset_value() returns null.
>> Do we have the guarantee that ciInlineKlass::get_null_reset_value() is always called on an initialized class, or should the case of the missing null_reset_value be handled here?
>
> I think it's checked in c1_GraphBuilder.cpp line 2119. This limitation is tracked by [JDK-8376254](https://bugs.openjdk.org/browse/JDK-8376254) but an assert/guarantee wouldn't hurt.

That's a good idea, I have added an `assert` in that function.

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

PR Review Comment: https://git.openjdk.org/valhalla/pull/1970#discussion_r2745325492