This is a very large diff. You may want to split it up in more logical chunks, such as generating the different variation of the png. This would make review easier, and ensure that parts of it that are good do not end up blocked by arts that are not.

I trust this is correct.

Remove PublicOnlySigningProvider

In D4322#104831, @Fabien wrote:

The code looks good, but the PublicOnlySigningProvider is unrelated to the title and summary of the diff.
Can you split it apart, or add some content into the summary/title ?

You are missing the usecli version of the test.

Address various comments

Move outputs_is_obj business instead of copying it.

rebase, ping?

nits, but lgtm

Ok, so I don't think this API is very good. add_library_with_type is kind of a tell, and the fact it has to duplicate a bunch of facility such as specifying linking and all. So here is the API that you want, IMO:

It would be beneficial to build another executable for the seeder tests.

Actually do the rebase an the backport, not just run the git rebase command.

Rebase, ping?

I trust this works, because I have no way èl check.

Is it a rebase script, or an arc patch script?

Please rebase, otherwise LGTM.

The test plan is inadequate. If you actually checked for thread safety annotation warning from the compiler, you'd know that there are some since Oct, 3rd (!)

Addres comments

The changes made in D3371 do not affect the constructor, so the rationale for not backporting makes no sense whatsoever.

Where are the other rename done, then ? I don't see a patch in the stack doing that. I don't understand why the double write doesn't come first, then the scripted diff? The way this is done is exceedingly difficult to review.

Actually run your test plan.

It's not clear you go the dependency in the right order.in any case, it seems that you have https://github.com/bitcoin/bitcoin/pull/11062 to go through first.

I think you got the dependency reversed.