Chore: AEAD_Mode can deal with generic containers

[reneme]

This introduces AEAD_Mode::set_associated_data(std::span<>) alongside the ptr-length-overloads. It switches the virtual method to the std::span<> overload and adapts the AEAD implementations accordingly. Also set_associated_data_vec() is marked as deprecated as it is just an alias now.

Caveat

This might be somewhat a show-stopper for the overload-based introduction of std::span<> we had in mind (also affecting #3297, #3294, #3195, #3201). In short: overloads of virtual methods in the base class are not necessarily dispatched to from an object of a derived type. A godbolt says more than a hundred words.

Same code for reference:

class AEAD_Mode {
public:
    virtual void set_associated_data(std::span<const uint8_t> ad) = 0;
    void set_associated_data(const uint8_t ad[], size_t ad_len)
        { set_associated_data(std::span(ad, ad_len)); }
};

class GCM : public AEAD_Mode {
public:
    // Otherwise the compiler does not statically dispatch to the
    // ptr-length overload in the base class
    using AEAD_Mode::set_associated_data;

    void set_associated_data(std::span<const uint8_t> ad) override final
        { std::cout << "Yeah!" << std::endl; }

};

int main()
{
    GCM gcm;

    std::vector<uint8_t> ad{'a', 'b', 'c'};

    // this works
    gcm.set_associated_data(ad);
    
    // this does not (without the `using` declaration in GCM)
    gcm.set_associated_data(ad.data(), ad.size());

    // - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

    std::unique_ptr<AEAD_Mode> gcm_ptr = std::make_unique<GCM>();
    
    // this works regardless of the `using` declaration
    gcm_ptr->set_associated_data(ad);
    gcm_ptr->set_associated_data(ad.data(), ad.size());

    return 0;
}

Frankly, I'm quite surprised that the method dispatch doesn't find the (publicly available) overload in the base class. 😢 One can work around it by explicitly using Base::method in the derived class, but that's easy to miss in large refactorings.
On the other hand, one could simply use other method names for the std::span overloads. But that complicates the public API even more and forces us to come up with new names for established concepts.

What a mess! Or am I actually missing something?

[codecov-commenter]

Codecov Report

Patch coverage: 97.87% and no project coverage change.

Comparison is base (ff51148) 88.14% compared to head (b1b0ad1) 88.15%.

❗ Current head b1b0ad1 differs from pull request most recent head 2c983c5. Consider uploading reports for the commit 2c983c5 to get more accurate results

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3317   +/-   ##
=======================================
  Coverage   88.14%   88.15%           
=======================================
  Files         617      617           
  Lines       70356    70353    -3     
  Branches     6987     6984    -3     
=======================================
+ Hits        62016    62020    +4     
+ Misses       5399     5394    -5     
+ Partials     2941     2939    -2     

Impacted Files	Coverage Δ
src/lib/modes/aead/aead.cpp	92.72% <ø> (+6.28%)	⬆️
src/lib/tls/tls12/tls_cbc/tls_cbc.cpp	83.41% <80.00%> (ø)
src/cli/cipher.cpp	89.65% <100.00%> (ø)
src/cli/pk_crypt.cpp	71.81% <100.00%> (ø)
src/cli/timing_tests.cpp	88.61% <100.00%> (ø)
src/lib/modes/aead/ccm/ccm.cpp	90.37% <100.00%> (ø)
...b/modes/aead/chacha20poly1305/chacha20poly1305.cpp	94.31% <100.00%> (ø)
src/lib/modes/aead/eax/eax.cpp	97.72% <100.00%> (ø)
src/lib/modes/aead/gcm/gcm.cpp	92.77% <100.00%> (ø)
src/lib/modes/aead/ocb/ocb.cpp	98.38% <100.00%> (ø)
... and 7 more

... and 11 files with indirect coverage changes

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[reneme]

Based on a suggestion by @lieser I added an indirection to work around above-mentioned caveat. The protected pure-virtual method set_ad_n() is now derived by all AEAD implementations. The idx parameter is ignored by most (except SIV), while the public API wrapper ensures that only SIV can successfully receive anything but 0 for the idx.

Come to think of it, that's exactly what we ended up implementing in #3195.

[reneme]

Rebased to master after one of yesterday's PRs conflicted with it.