Removing -XX:+AOTClassLinking support for dynamic CDS archive

[Ashutosh Mehra]

+1 for removing AOTClassLinking support for dynamic archives. One less
scenario to think of.
Since we are adopting and advocating AOTCache based workflow, perhaps we
should remove support for AOTClassLinking with any of the older CDS based
workflows - static or dynamic archive.
I am not sure if it would have any impact on the code base, but it would
keep the two workflows completely separate, making it easier to reason
about the future changes.

Thanks,
- Ashutosh Mehra


On Wed, Sep 10, 2025 at 3:14 PM <ioi.lam at oracle.com> wrote:

> I've filed https://bugs.openjdk.org/browse/JDK-8367366 to remove
> -XX:+AOTClassLinking support for dynamic CDS archive. Any objections?
>
> Background: we needed this for the old (and removed) "5 step workflow"
> in early Leyden development. And it somehow leaked into the mainline
> when we upstreamed JEP 483.
>
> However, most (if not all) current and future Leyden optimizations will
> depend on BOTH -XX:+AOTClassLinking and archive heap objects. Because we
> don't support archived heap objects in the dynamic archive, so it won't
> benefit from any of those optimization.
>
> My main concerns are (1) code bloat, and (2) the chance of introducing
> bugs when we add new optimizations without sufficient testing for the
> dynamic archive. I think I actually have one such bug in
> `fixup_module_field_list()` in https://github.com/openjdk/jdk/pull/26375
> (preload classes from AOT cache).
>
> Thanks
>
> - Ioi
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/leyden-dev/attachments/20250910/daa9aa7b/attachment.htm>