Ryan Ofsky

> This is a regression bugfix and a blocker for sqlite wallets. 0.21 should not be released without it. Can you give an example of something that would be broken,...

> Adding a UUID later would mean backups are missing it, and restoring from a backup could break features relying on it. This isn't true. Unrecognized rows are not stripped...

> The UUID isn't simply unrecognised, it's ENTIRELY MISSING in sqlite wallets right now. I know this. You know this. What I don't know is what specific problem is fixed...

re: https://github.com/bitcoin/bitcoin/pull/20205#issuecomment-722593240 > Because of that, I think it makes conceptual sense to have an ID, and use that to warn (not reject) when opening multiple copies. Without knowing who...

> @ryanofsky There already is, why are you ignoring it? #19463 I'm aware and not ignoring it. If #19463 is close to being merged and it depends on this PR,...

Sjors could you clarify what the broken case is: What steps does the user perform, and what is the unexpected behavior in this case and the expected behavior? It seems...

> > What steps does the user perform, and what is the unexpected behavior in this case and the expected behavior? > > 1. User configures a multisig wallet in...

> I agree there's something to be said for identifying wallets based on the descriptors (descriptor id) rather than the wallet id. From the description it seems like #23417 could...

> However, what we probably want here is a `configure` check Debian seems to have patched this in: https://sources.debian.org/patches/capnproto/0.9.1-4/07_libatomic.patch/

> @ryanofsky Hmm, that patch looks mildly wrong. It looks like it'll force linking against libatomic on all platforms, even those that don't need it. Most platforms have built-in instructions...