Nickolay Olshevsky

> What I don't like is the name of a new subroutine I've created, `key_fetch_and_try()`. Name doesn't convey the meaning well, and there's already a similarly-named `encrypted_try_key()` which does a...

> But this is the kind of error log message for a condition which is not intended to happen normally, and it may be uneconomical to write a test just...

> Here's some strange CI failure. It says type mismatch (`char const*` vs `const char *`?), but I don't see any in the sources. Yeah, most likely it's some compiler...

> Also the `rnp-ruby` test suite failure seems unrelated, maybe flakiness? > > `1) # returns a higher iterations count for MD5 vs SM3` Yeah, this is due to timing...

@andrey-utkin Thanks, I'll review this later today. Usually we discuss issues here on Github, however if there are some questions which is handier to discuss via chat - feel free...

@andrey-utkin I'd suggest another approach, making key provider responsible to check for zero keyid: - in `key_fetch_and_try()` just call a key provider, without checking for zero keyid. - in `lib/rnp.cpp`,...

`` should be added to the include path of your project.

Tried exporting revoked subkeys from GnuPG and importing/decrypting on RNP side, and wasn't able to reproduce the issue.

Probably some older versions of GnuPG didn't use dummy s2k but just exported empty secret key data.

Probably we may re-ping them once v0.15.2 is released.