[jdk11u-dev] RFR: 8078725: method adjustments can be done just once for all classes involved into redefinition

[Matthias Baesken]

On Wed, 26 Jun 2024 12:51:43 GMT, Martin Doerr <mdoerr at openjdk.org> wrote:

> Backport of [JDK-8078725](https://bugs.openjdk.org/browse/JDK-8078725) which is a prerequisite of [JDK-8222005](https://bugs.openjdk.org/browse/JDK-8222005). Applies almost cleanly. Manual changes:
> 
> 1.  Only one hunk required manual integration (1st commit):
> 
> 
> --- src/hotspot/share/prims/resolvedMethodTable.cpp
> +++ src/hotspot/share/prims/resolvedMethodTable.cpp
> @@ -232,12 +232,7 @@ void ResolvedMethodTable::adjust_method_entries(bool * trace_name_printed) {
>            continue;
>          }
>  
> -        InstanceKlass* holder = old_method->method_holder();
> -        Method* new_method = holder->method_with_idnum(old_method->orig_method_idnum());
> -        assert(holder == new_method->method_holder(), "call after swapping redefined guts");
> -        assert(new_method != NULL, "method_with_idnum() should not be NULL");
> -        assert(old_method != new_method, "sanity check");
> -
> +        Method* new_method = old_method->get_new_method();
>          java_lang_invoke_ResolvedMethodName::set_vmtarget(mem_name, new_method);
>  
>          ResourceMark rm;
> 
> JDK 11 code has `new_method` already declared and only needs to assign it.
> 
> 2. `Method* get_new_method() const` already exists in 11u. Removed duplicate (2nd commit).

Marked as reviewed by mbaesken (Reviewer).

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

PR Review: https://git.openjdk.org/jdk11u-dev/pull/2816#pullrequestreview-2141828506