[crac] RFR: Merge jdk:jdk-27+5 [v2]

[Timofei Pushkin]

On Mon, 2 Feb 2026 10:40:45 GMT, Radim Vansa <rvansa at openjdk.org> wrote:

> The changes due to GC look like there is some chance for a regression. I wonder if we have something to run to validate that there's none.

I ran a helidon-based benchmark I used for the related changes in G1 and checked that removing `Universe::heap()->set_cleanup_unused(true)` in `crac.cpp` leads to CRIU image size increasing significantly for the 3 affected GCs. For G1 I also checked that the excess GCs I was fixing by those changes do not come back (with `Universe::heap()->set_cleanup_unused(true)` in place).

> In the `CracBuilder` changes, I am not really sure if changing the type used by tests is a good thing. I like isolating the container-related code to a separate class, but encapsulation would be my first choice rather than inheritance in this case.

I think encapsulation would result in more complicated code, if possible at all: `CracBuilder` cannot reference `CracContainerBuilder` because that would be a transitive reference to `DockerTestUtils` — the thing we want to avoid.

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

PR Comment: https://git.openjdk.org/crac/pull/288#issuecomment-3834479471