librepgp: reject packet tag 0 as required by RFC 4880

[ronaldtse]

Summary

• RFC 4880, section 4.3, states that a packet tag must not have the value 0, which is reserved. The parser previously accepted such a header and skipped the packet as unknown, while GnuPG rejects the same input, and the difference was found through differential fuzzing and reported by @afldl in Reserved packet tag 0 accepted — RFC 4880 §4.3 MUST NOT violation #2397.
• stream_peek_packet_hdr() now returns RNP_ERROR_BAD_FORMAT when the decoded tag is 0, for both the new-format (0xc0) and the old-format (0x80) header spellings, which covers the parsing, dumping and keyring load paths that share this header gate.
• Other reserved tag values, which the RFC does not single out in the same way, continue to be skipped as unknown, and this change does not affect them.
• The new test_issue_2397_tag0 test feeds both header spellings to the packet dump and the key import paths and asserts that each fails.

Test plan

• rnp_tests.test_issue_2397_tag0 passes against both spellings of the reserved tag
• the parser-related tests (load_pgp, load_g10, load_g23, load_kbx, stream_key_load, the dump tests and fuzz_dump) continue to pass
• full CI matrix

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.43%. Comparing base (26482f6) to head (8c4d1f9).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2489      +/-   ##
==========================================
- Coverage   85.45%   85.43%   -0.03%     
==========================================
  Files         125      125              
  Lines       23042    23042              
==========================================
- Hits        19691    19685       -6     
- Misses       3351     3357       +6     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ronaldtse]

@ni4 a small batch of CI and process pull requests has been prepared from the issue backlog, each of which needs only your approval:

• librepgp: reject packet tag 0 as required by RFC 4880 #2489 rejects the reserved packet tag 0 in the parser, closing Reserved packet tag 0 accepted — RFC 4880 §4.3 MUST NOT violation #2397 which @afldl found through differential fuzzing against GnuPG.
• ci: add linelint and normalise whitespace across first-party files #2490 adds linelint to the lint workflow per Add linelint to the lint workflows. #2300, with a configuration which excludes the vendored trees and the parser fixtures whose bytes are part of the tests, and normalises the whitespace this surfaced in first-party files. The change is provably whitespace-only.
• ci: add the OpenSSF Scorecard analysis workflow #2491 adds the OpenSSF Scorecard workflow per Add ossf-scorecard action #2311, publishing its report to code scanning next to the existing CodeQL results.
• ci: run ctest serially as a temporary experiment #2492 runs ctest serially as the temporary experiment you were asked about in Use ctest --parallel 1 to check whether it fixes random test failures #2394, so the next runs show whether the random interop failures persist without concurrent execution. It is intended to be reverted once the diagnostic value is spent.
• ffi: succeed with false for revocation-code predicates on live keys #2488, for completeness, fixes the revocation-code predicates to return false instead of an error for live keys (rnp_key_is_compromised/retired/superseded return RNP_ERROR_BAD_PARAMETERS for keys that are not revoked #2467); as a behaviour change in the FFI it would benefit from your review together with a second approval.

Whenever time allows, a look at any of these would be much appreciated.

[ni4]

LGTM!