On Thu, 3 Sep 2026 07:41:10 GMT, Axel Boldt-Christmas <[email protected]> 
wrote:

> Also given the `JDK-8391174` I would probably have changed all the markWord 
> hash zero checks and asserts to be `[!]obj->has_identity_hash()` checks / 
> asserts as it reflects the object invariants better.

And as I wrote earlier we should probably remove this invariant on the table 
interface, use `oopDesc::identity_hash` inside the table when doing a put, and 
remove the requirement for lookups and simply short-circuit. 
(`obj->has_identity_hash()` is invariant after something has been added to the 
table, but we should not require it for lookups)

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

PR Comment: https://git.openjdk.org/jdk/pull/32573#issuecomment-5522651280

Reply via email to