What are the prospects for this PR? Do we want to land it in some form? I rebased it to mainline and with some minor fixes the test passes. Now that release/18.x has branched, we'd have some time to revisit it and stabilize it. What do you think?

One thing I'd like to discuss is how we can separate management of symbols from stubs. We have an IndirectStubsManager that we use for lazy JITing and it's already wired up with EPC right?

Can we integrate it in the redirection manager and avoid more stub management?

What are the prospects for this PR? Do we want to land it in some form? I rebased it to mainline and with some minor fixes the test passes. Now that release/18.x has branched, we'd have some time to revisit it and stabilize it. What do you think?

We definitely want to land it, I've just been short on time to review.

One thing I'd like to discuss is how we can separate management of symbols from stubs. We have an IndirectStubsManager that we use for lazy JITing and it's already wired up with EPC right?

Can we integrate it in the redirection manager and avoid more stub management?

The aim is for this work to replace the existing IndirectStubsManager (we may reuse the name, but want to update the interface).

The aim is for this work to replace the existing IndirectStubsManager

Ok, we might want to consider this a draft then as it seems to require more work. I think this patch is a good basis for experimentation with the design of such a new interface. Maybe it makes sense though to discuss and land the new stubs manager interface in isolation? I may find the time to think about it in the next weeks.

Superseded by #67050. Sorry for the confusion!