model-writtenrun 6179 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Phil Krylov proposed dynamically allocating pg_ctl command strings with psprintf() after observing that truncation at MAXPGPATH could produce shell syntax errors, and included a regression test plus explicit memory cleanup #1.
- Ranier Vilela raised Windows command-length limits and supported freeing allocations; Phil argued that platform limits varied and that dynamically allocating the command avoided oversized stack buffers without requiring another hard-limit check #2 #3.
- Tom Lane endorsed psprintf(), rejected small-allocation cleanup as unnecessary for short-lived pg_ctl processes, and considered the proposed regression test an unhelpful expenditure of test cycles #4.
- Phil submitted a revision removing the test and error-path frees while retaining the dynamic command construction and three normal-path frees #5.
- Tom committed the revised patch as 87ad491472, removing the remaining added frees and adjusting declarations, with backports to all supported branches; his final message confirmed the patch was pushed #6.