model-writtenrun 2171 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Feb - Apr 2024 · 12 entries
- Maiquel Grassi proposed \conninfo+ to give administrators richer connection details without deploying helper functions across databases; the initial implementation queried the server and displayed Attribute/Value rows #1.
- Nathan Bossart and Erik Wienhold advocated a single result row with one column per attribute and expanded display controlled by \x; Maiquel adopted Nathan's concrete SELECT after an initial misunderstanding #2 #3 #5 #7.
- Erik corrected literal NULL strings and unreliable localhost inference, while Pavel Luzanov distinguished authenticated, current, session, and system users; successive revisions used libpq for the authenticated identity and host and guarded system_user() for older servers #11 #12 #13 #14 #16.
- Jim Jones found that querying unix_socket_directories denied access to ordinary users, leading Maiquel to derive the socket directory from libpq; documentation followed, and Erik's review removed unused pattern parsing #18 #20 #23 #22 #24.
- Nathan questioned Server Version and separately printed encryption details, while Pavel reported a prototype warning, unfamiliar PID naming, and inconsistent behavior after connection termination; Maiquel removed Server Version, corrected prototypes, and adopted Backend PID #27 #29 #30 #31.
- Álvaro Herrera proposed incorporating encryption information into the table, and Maiquel implemented that before adding a query to plain \conninfo to detect terminated connections #33 #35 #36.
- Jim's SSL and host-address testing exposed differences between client and server address information and prompted fuller documentation; Maiquel temporarily changed plain \conninfo to display inet_server_addr() and expanded the field descriptions #38 #39 #41 #43.
- Nathan requested stable output columns and server statistics views for reusable SQL; Maiquel first standardized missing values, then replaced libpq encryption lookups with version-aware queries of pg_stat_ssl and pg_stat_gssapi #42 #44 #46 #50.
- Sami Imseih requested fuller GSSAPI information and application_name, then reviewed field ordering and documentation; those additions entered the SQL-backed patch while the original prose format of plain \conninfo was retained #52 #53 #54 #55.
- Álvaro and Sami guided translation support, simpler version checks, documentation cross-references, and valid SGML, producing a sequence of implementation and documentation cleanups #56 #61 #65 #67 #68.
- Peter Eisentraut challenged the shift from client connection information to server queries, and Tom Lane explained that queries fail in aborted transactions and obscure client/server discrepancies; Maiquel restored plain \conninfo and removed application_name, but \conninfo+ still queried the server #70 #73 #75.
- Sami demonstrated that changing roles could make the statistics-based output empty, and Peter maintained that a plus variant should retain \conninfo's libpq purpose; a separate session-information command or system view was discussed, but no such implementation was submitted here #76 #77 #78 #80.
May - Jun 2024 · 1 entry
- After Maiquel stepped back because of family commitments, Hunaid Sohail took over and rewrote the enhancement around libpq APIs in August, eliminating the server query; a later rebase also replaced the unusable UTF-16 patch attachment #81 #82 #83 #84.
Aug 2024 - Feb 2025 · 10 entries
- Jim and Álvaro helped restore tabular output to Hunaid's implementation, and Jim clarified the difference between a two-column table and expanded display of one record #85 #86 #87 #90 #91.
- Álvaro and Tom objected to mixing prose with tabular information, so Hunaid moved all connection and SSL details into the table; subsequent review fixed misleading user labels, boolean presentation, initialization warnings, and code organization #94 #95 #96 #97 #98 #99 #101.
- David G. Johnston proposed organizing the libpq information into categories, and Hunaid implemented four selectable tables for connection information, status, parameters, and encryption #109 #112.
- Review resumed in January with Sami questioning the categories and requesting complete reported parameters, role information, and a public enumeration API; Álvaro opposed the multiple-table interface and its selectors #115 #117 #119 #120.
- Dean Rasheed favored a single Parameter/Value table that would be readable without expanded mode; an experimental description column was dropped after Dean and Álvaro judged it mostly redundant #127 #128 #129 #130.
- Hunaid submitted a three-patch series combining the two-column display with PQparameterNames() and GUC_REPORT for role, enabling dynamic display of all server-reported parameters #131.
- Álvaro questioned duplicating \dconfig and the relevance of every reported parameter, then proposed replacing plain \conninfo itself; Maiquel supported a smaller essential field set, and Hunaid agreed to pursue that direction #132 #135 #136 #137.
- Álvaro supplied a simplified patch replacing plain \conninfo with the table and retaining only is_superuser and in_hot_standby among reported parameters; it omitted parameter enumeration and role reporting because their complexity and protocol implications did not need to block the feature #138.
- Tom endorsed the simplified design and requested removal of question marks, while Sami explicitly approved the final patch; Álvaro committed it with those label changes, capitalization fixes, a Server Port label, and documentation indentation adjustments as bba2fbc623, crediting Maiquel and Hunaid as co-authors #139 #141 #144.
- After landing, Álvaro left role and session-authorization additions open for discussion and noted inconsistent boolean spellings, older-server fallback wording, and missing richer GSS attributes; his proposed const cleanup was reconsidered, and Dagfinn Ilmari Mannsåker pointed out that cell validation can modify strings #144 #145 #146.
Jun 2025 · 2 entries
- In June, Peter proposed replacing the new TLS terminology with SSL for consistency with the rest of PostgreSQL, and Álvaro agreed while suggesting a possible project-wide TLS rename in a future release #147 #148.
- Peter committed the terminology patch unchanged as 6d6480066c and confirmed completion in the final message; the tabular command and terminology fix had landed, while parameter enumeration, role reporting, and the other suggested extensions remained outside these commits #149.