Currently, the Leyden new workflow works ONLY with G1GC.

The problem is when dumping, G1 must be used if you want to write the archived heap objects. Other GCs can "load" archived heap objects, but cannot "write" them.

Without the ability to write, you can't have pre-resolution for indy, etc.

Also, AOT code is tied to the GC used in training run, it generates code specific to G1's barriers. This means the code cannot be used if you use a different GC during the production run.

Suggested fix:

Support SerialGC and ParallelGC to write the archived heap objects. This will make it possible to use Leyden in containers with a low memory limit.

Other collectors such as ZGC are more difficult to support. To be done in future RFEs or JEPs.