model-writtenrun 7557 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Dave Cramer reported that a PL/R test produced ERROR on Linux but WARNING on Windows; Tom Lane suspected recently renumbered error levels, while Dave confirmed that the failure occurred in a fresh CI build #1 #4 #5.
- Tomas Vondra identified PL/R's hard-coded ERROR value of 20, which PostgreSQL now assigned to WARNING_CLIENT_ONLY, with ERROR moved to 21 #6.
- Andrew Dunstan recommended preserving PostgreSQL's values across inclusion of R headers instead of hard-coding them; Tom Lane pointed to the existing PGERROR alias, and Joe Conway explained that R also defines WARNING and requested a corresponding PGWARNING alias and backpatching #8 #9 #10.
- Tom Lane proposed adding PGWARNING only to HEAD, using a conditional fallback in PL/R for older headers, and removing the Windows-only restriction on PGERROR; he then attached the implementation with explanatory comments #11 #12.
- Dave Cramer welcomed the clearer comments and Joe Conway expressed approval, but neither reported testing, identified a code problem, or proposed an adopted change to the posted patch #13 #14.
- Tom Lane committed #12 unchanged as d7cff12c4c and closed the thread by confirming that it was pushed to HEAD only; the PostgreSQL change landed without backpatching, and no PL/R-side implementation was shown #15.