RFR: 8329858: G1: Make G1VerifyLiveAndRemSetClosure stateless
Thomas Schatzl
tschatzl at openjdk.org
Mon Apr 8 14:06:10 UTC 2024
On Mon, 8 Apr 2024 12:36:06 GMT, Stefan Karlsson <stefank at openjdk.org> wrote:
> In [JDK-8329570](https://bugs.openjdk.org/browse/JDK-8329570): 'G1: Excessive is_obj_dead_cond calls in verification' we moved the verification of the containing object so that it was done only once. We left the repeated null-check because of concerns that moving it would introduce the risk that someone in the future would miss calling `set_containing_obj` or calling it with a null object.
>
> I propose that we restructure the code to make the closure's state const and remove the set_containing_obj function so that we can feel confident that we don't hit such a bug in the future.
>
> The patch also combines the failures from verification of the oops with failures from checking the containing objects. I had a version that kept these two separate, but I don't see why we would want that. Tell me if you'd want me to reintroduce this separation.
Marked as reviewed by tschatzl (Reviewer).
-------------
PR Review: https://git.openjdk.org/jdk/pull/18677#pullrequestreview-1986549735
More information about the hotspot-gc-dev
mailing list