model-writtenrun 1654 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Sergey Tatarintsev reported a crash when passing null hints to pg_getaddrinfo_all() and proposed a null check so the wrapper could accept them like getaddrinfo() #1.
- Daniel Gustafsson questioned the semantics because the wrapper supports Unix sockets and pg_freeaddrinfo_all() expects the original hints' address family #2; Sergey revised the patch to document passing AF_UNSPEC to the freeing routine when hints were null #3.
- Daniel remained unconvinced that the wrapper should mimic getaddrinfo() and suggested returning EAI_FAIL for null hints #4, while Tom Lane favored retaining the longstanding behavior and documenting the requirement for valid hints #5.
- Daniel supplied an inline documentation-only patch using wording consistent with pg_getnameinfo_all() #6, and Tom approved it #7.
- Daniel committed the inline patch unchanged as d22cc7326c and confirmed completion #8, closing the thread with the non-null requirement documented and the proposed null-hints support left out.