Setting a curve type (which shouldn't be necessary) like "Curve-Type: ed25519" doesn't help either. While this makes the check in gpg pass, the gpg-agent process re-checks the parameter set and rejects it with the same error message.

My concern is not a disloyal administrator, so I disagree with that priority.

Looks good to me: "make && make check" passes.

Admin here. I'm sorry your replies did not make it to this site but somewhere got stuck. So copying them for completeness:

[2021-02-24] gnupg2 2.2.27-1 MIGRATED to testing (Debian testing watch)

Okay. Now since configure.ac is already touching CFLAGS, it seemed like a good place to add that additional option here. All this is guarded by a test for GCC, and since clang mimics that behaviour, it works for them as well.

After some more checking: LLVM-11 introduced the same behaviour in that regard, but appearently not a pragma/attribute to override this: https://releases.llvm.org/11.0.0/tools/clang/docs/ReleaseNotes.html

According to list of attributes in the clang 12 documention, there is no such attribute in clang. However, the clang-11 compiler (as seen in Debian) does not define __GNUC__, so the proposed patch does not affect the status when building with clang.