Conversation
Probably not worth doing a major bump for? Do we have a deprecation mechanism here? Stick with 0.5 until rustls 0.24 comes out? |
a8cc72c to
57b2fd6
Compare
cpu
left a comment
There was a problem hiding this comment.
Probably not worth doing a major bump for?
Yeah, probably not, but the README's ABI stability claims mean consumers that take our .so need to treat every release as a major bump. I'd personally be inclined to wait for more changes to pile up in main to do any kind of release (or revisiting if there's an explicit request from downstream).
Do we have a deprecation mechanism here?
I think cbindgen supports translating #[deprecated] into __attribute__((deprecated)) annotations in the .h if that's what you meant 🤔
Handle fallible verifier creation. Disable miri isolation as the existing tests now touch the file system (to load root certs).
Added a commit to enable this. Then I've added |
This is `rustls_platform_server_cert_verifier_with_provider` but with a more orderly way of reporting errors. Mark `rustls_platform_server_cert_verifier_with_provider` as deprecated.`
Notably this means
rustls_platform_server_cert_verifier_with_providercan now return NULL. Avoiding that would mean using an out param and returning arustls_result, but would be a major version bump. Thoughts?(I checked and curl uses
rustls_platform_server_cert_verifierand is therefore suitably fallible.)