[TLS 1.3] Preparations for addition of TLS 1.3

[reneme]

Depends on: #2941, #2947, #2948 (requires rebase after merge)

This includes refactorings and preparatory steps for the integration of TLS 1.3 into the code base. The intention is similar to the previous TLS 1.2 refactoring pull request by @pist-eb and colleagues.

The central preparation is to add a Pimpl-Layer behind the public interfaces of TLS::Client and TLS::Server. Furthermore, central TLS handshake messages (such as Client Hello and Server Hello) are renamed into Client_Hello_12 with a Client_Hello base class. Generally, such base classes encapsulate common behaviour (usually mostly wire format marshalling) while _12 provides protocol version specific functionality. The addition of TLS 1.3 will fill in the gaps with Client_Hello_13 and friends.

This PR was created by squashing the changes in the pending TLS 1.3 PR and then removing (most) of the TLS 1.3 specific code. Hence, some interfaces or functionality might be unused or dangling right now. I tried to keep API and functionality adaptions to an absolute minimum to ease grafting the TLS 1.3 changes onto this PR later.

[reneme]

Thanks for the quick turnaround with the smaller PRs. This is now rebased to latest master and hence doesn't include the changes in tls_reader.h, tls_transitions.h, tests.h, ... any longer.

[codecov-commenter]

Codecov Report

Merging #2946 (4c0c50f) into master (b040969) will decrease coverage by 0.02%.
The diff coverage is 89.47%.

@@            Coverage Diff             @@
##           master    #2946      +/-   ##
==========================================
- Coverage   92.60%   92.58%   -0.03%     
==========================================
  Files         573      577       +4     
  Lines       66485    66710     +225     
  Branches     6365     6375      +10     
==========================================
+ Hits        61570    61762     +192     
- Misses       4882     4915      +33     
  Partials       33       33              

Impacted Files	Coverage Δ
src/bogo_shim/bogo_shim.cpp	88.72% <ø> (ø)
src/lib/tls/tls12/msg_cert_status.cpp	76.66% <ø> (ø)
src/lib/tls/tls12/msg_hello_verify.cpp	100.00% <ø> (ø)
src/lib/tls/tls12/tls_cbc/tls_cbc.cpp	90.38% <ø> (ø)
src/lib/tls/tls12/tls_handshake_hash.cpp	100.00% <ø> (ø)
src/lib/tls/tls12/tls_record.cpp	95.81% <ø> (ø)
src/lib/tls/tls12/tls_session_key.cpp	100.00% <ø> (ø)
src/lib/tls/tls_algos.cpp	83.93% <ø> (ø)
src/tests/test_tls.cpp	92.46% <ø> (ø)
src/tests/test_tls_utils.cpp	0.00% <0.00%> (ø)
... and 42 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b040969...4c0c50f. Read the comment docs.

[randombit]

Thanks for splitting this out. This is the "easy" part for review (nothing really new, unlike the actual 1.3 logic) but is already a big patch. Working on the review now, may take a day or two.

[randombit]

OK that went faster than I thought. Overall looks very good! Left some comments, and I do want to do a final pass review before merge, but we're almost there on this one.

[reneme]

I acted on all your comments.

Regarding coding style comments: Let's do style fixes on the final PR to avoid unnecessary rebase frustration. I'm very much in favour of clang-format, by the way! We're currently using the astyle formatter somewhat consistently with varying satisfaction.

Actually clang-format seems to support Whitesmith style as of recently, though I didn't figure out a configuration to match the current code base's style reasonably well anyway.

[randombit]

Let's do style fixes on the final PR to avoid unnecessary rebase frustration.

Sounds good.

Actually clang-format seems to support Whitesmith style as of recently, though I didn't figure out a configuration to match the current code base's style reasonably well anyway.

I have also so far been unable to get anything that is close to the current codebase. I think when it happens I'll just do a reformat into some other style that is convenient to express using clang-format configuration and that I don't find too ugly. For this purpose I'll probably wait until just before we do a final 3.0 release when things are otherwise very quiet in terms of new development. And then we'll just have to suffer that any backports to 2.x are going to be a little painful, but hopefully not too many will be required.

[randombit]

Let's go!

[reneme]

Great! I'll get going on the rebase of #2922. Probably that'll mean squashing its commit history, at least partially.