model-writtenrun 7858 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Dean Rasheed reported that numeric_power() lost precision for exponent INT_MIN, returning 1.0000000000000000 for a calculation whose result should be approximately 0.7678656556403084 #1.
- Dean traced the failure to taking the integer absolute value of INT_MIN and proposed using fabs() so conversion to double happens before taking the absolute value #1.
- Dean attached the one-line fix and a regression test, proposing to back-patch through PostgreSQL 9.6, where the faulty calculation was introduced #1.
- Tom Lane agreed that the fix seemed appropriate, and the thread ended without objections or requests for changes #2.
- Dean subsequently committed the patch unchanged as bc43b7c2c0, with backports recorded for the affected stable branches and no remaining work identified.