[quinnj-bot]

cipher_update sized its output buffer as in_length + block_size - 1. OpenSSL documents that decryption with padding can write up to in_length + block_size bytes (EVP_DecryptUpdate), so on OpenSSL 1.1 an update whose input is a whole number of blocks, after an earlier partial update, wrote one byte past the buffer. OpenSSL 3 stays within the shorter bound. The buffer is now in_length + block_size, which also covers encryption (in_length + block_size - 1).

Found by the sweep's OpenSSL memory-safety audit. No regression test: the write lands in allocator slack and is only observable with a memory checker. Checked locally: AES-256-CBC encrypt/decrypt round trips with split decrypt updates for lengths 0, 1, 15, 16, 17, 32, 100 and 4096 on Julia 1.12.

The old expression also used float division (/); with power-of-two block sizes it was exact, so the change in size is exactly one byte.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.41%. Comparing base (c7f1272) to head (a95c3d8).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #113      +/-   ##
==========================================
- Coverage   83.43%   83.41%   -0.02%     
==========================================
  Files           2        2              
  Lines        1352     1351       -1     
==========================================
- Hits         1128     1127       -1     
  Misses        224      224              

☔ 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.