model-writtenrun 5693 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut opened the thread with #1, rebasing an earlier proposal to replace pgcrypto's internal padding with OpenSSL because the non-OpenSSL code paths had already been removed.
- Jacob Champion observed in #2 that the patch rejects bad padding during decryption and non-block-aligned input during encryption without padding, supplied reproducing queries, and raised compatibility and padding-oracle concerns while considering both behavior changes correct.
- Peter added regression tests and removed the unused block-size error code in #3, while asking how the malformed ciphertext was constructed and whether the old behaviors had any reasonable purpose.
- Jacob explained the ciphertext corruption in #4, identified silent zero-padding in the old encryption code, and argued that the previous behavior was unreasonable; Peter incorporated the explanation into test comments and clarified the behavior changes in #5, declaring the patch complete.
- Nathan Bossart reported no concerns after reviewing the final patch in #6, and Peter committed #5 unchanged as f5576a21b0, confirming completion in the thread's final message #7.