RFR: 8360555: Archive all unnamed modules in CDS full module graph

[Ioi Lam]

On Thu, 17 Jul 2025 12:35:54 GMT, Coleen Phillimore <coleenp at openjdk.org> wrote:

>> src/hotspot/share/classfile/modules.cpp line 777:
>> 
>>> 775:   if (CDSConfig::is_using_full_module_graph()) {
>>> 776:     precond(unnamed_module == ClassLoaderDataShared::archived_boot_unnamed_module());
>>> 777:     unnamed_module->restore_archived_oops(boot_loader_data);
>> 
>> If you're restoring the module oop from the archive, what is the module oop passed into this that the rest of the code is using?
>
> If you're storing the unnamed module oop in the archive should this method not be called?  If it is, what are you saving by archiving the unnamed module?

The callstack is:


jdk.internal.loader.BootLoader.setBootLoaderUnnamedModule0(java.base at 26-internal/Native Method)
jdk.internal.loader.BootLoader.<clinit>(java.base at 26-internal/BootLoader.java:71)
jdk.internal.module.ModuleBootstrap.boot(java.base at 26-internal/ModuleBootstrap.java:162)
java.lang.System.initPhase2(java.base at 26-internal/System.java:1932)


Both the Java code and the native code have a handle to this unnamed module oop. The `precond` checks that they indeed are pointing the same oop.

Also, even though the oop is archived, we still need to set up some native states inside the `unnamed_module->restore_archived_oops(boot_loader_data)` call. E.g., set up the `OopHandle` that binds the oop to the `ModuleEntry`.

---------------
> what are you saving by archiving the unnamed module?

It's for [JDK-8350550](https://bugs.openjdk.org/browse/JDK-8350550)) -- I want to be able to reference the unnamed module before executing any Java code, so that archived classes can be loaded at the very beginning of `vmClasses::resolve_all()`. See my draft PR: https://github.com/openjdk/jdk/pull/26375

---------------
Currently, we still execute a lot of Java code when setting up the archived module graph (inside `ModuleBootstrap.boot()`. I am working on a way to enable the archived module graph without executing any Java code (which will be a few REFs after [JDK-8350550](https://bugs.openjdk.org/browse/JDK-8350550)), so this call will eventually be gone.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/26082#discussion_r2214437363