It might also be worth dropping synstructure. zeroize_derive isn't a particularly complicated proc macro.

Never mind, synstructure has been updated.

I missed the edit and wrote a commit removing the synstructure dep. Do you want me to drop that commit or leave it in?

Oh, if you’ve already done it one fewer dependency is great, thank you!

Hmm, seems like the zeroize tests are failing on fields that use #[zeroize(skip)] because they aren't being accessed.

I think you could go ahead and add #[allow(unused_variables)] to those fields?

Added the #[allow(unused_variables)], and put a #![deny(warnings)] into the zeroize derive test that was breaking - it had scrolled past clean locally because the warnings got pushed off the screen, but you must be overriding that in CI somewhere. That should make it fail locally for others.

We deny warnings in CI but allow them locally to make refactoring easier. I'd suggest removing the #![deny(warnings)].

Hmm, seems this will entail an MSRV bump for the derive feature. That's fine.

I opened #869 which does a 2021 edition upgrade for zeroize and zeroize_derive

@maurer if you rebase it should take care of the MSRV-related issues

Awesome, thanks!