On Thu, 3 Sep 2026 14:18:51 GMT, Oli Gillespie <[email protected]> wrote:

> It solves most of the confusion and the data leak. It doesn't solve 
> performance overhead. It regresses dump size.

I'd favor this direction. The quirks are workable downstream -- the class name 
is a stable marker, so analyzers can special-case it (derive true footprint 
from length, discount fillers, aggregate waste), and the all-null elements 
compress nearly for free. Omitting fillers in the JVM instead removes the 
choice entirely. (Ofc, this is rather subjective; up to you.)

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

PR Comment: https://git.openjdk.org/jdk/pull/32542#issuecomment-5539872797

Reply via email to