model-writtenrun 2604 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Dave Page reported that PostgreSQL 16 failed to compile with MSVC when both OpenSSL and MIT Kerberos GSSAPI were enabled #1; Imran Zaheer reproduced the conflict on PostgreSQL 16 and 17 and proposed moving OpenSSL includes after GSSAPI includes #2.
- Andrew Dunstan and Tom Lane preferred explicitly undefining the conflicting X509_NAME macro because relying on header order was fragile and contradicted project conventions #3 #4, prompting Imran to submit two alternative locations for the undef #5.
- Dave confirmed successful compilation and tests for both undef alternatives #6 #8, while Andrew preferred placing the undef after the GSSAPI includes so sslinfo.c could share the fix #7; Imran incorporated that cleanup and proposed undoing existing OpenSSL include-order workarounds #9.
- Andres Freund observed that other GSSAPI include sites remained exposed and proposed a shared pg-gssapi.h wrapper #10; Imran accepted that design and agreed to leave the existing OpenSSL include ordering intact for protection against other conflicting headers #11.
- After a six-month gap, Daniel Gustafsson revived the unresolved issue #12, Dave requested backpatching #13, and Andrew supported the wrapper approach #14; Daniel then supplied a replacement patch centralizing the GSSAPI headers and macro cleanup, reporting a successful build and test with GSSAPI enabled #17.
- Andres endorsed Daniel's patch and suggested removing a redundant ENABLE_GSS guard #18; Vignesh supported that cleanup, reported trailing whitespace, and questioned the retained copyright dates #20, while Andrew also said the patch looked good despite lacking a suitable test environment #21.
- Daniel committed e92c0632c1 from #17 with redundant guards and trailing whitespace removed, retaining the copyright dates because the content was moved, and reported another CI run before pushing #22.
- Daniel closed the commitfest entry after allowing time for buildfarm feedback #24, and Dave thanked the participants in the final message #25; the development-branch fix had landed, while backpatching remained a stated follow-up in e92c0632c1 with no completion recorded in this thread.