On Friday, October 23rd, 2026 at 5:41 PM, Robert Haas <[email protected]> wrote:
> #19 might be a real bug, but needs validation, and can't be easily > validated with in-core code. Hi Robert, I am still getting used to how discussions are organized on this list. I posted a reproducer and patch in a separate thread [1]. Since the related discussion is continuing here, I thought it might be helpful to add a link here as well. I confirmed #19 with a small test-only CustomScan provider. When a CustomScan replaces a join within a larger join, EXPLAIN (PLAN_ADVICE) can fail with "plan node has no RTIs". The patch makes pgpa_relids() return CustomScan.custom_relids. When you have a chance, I would appreciate your thoughts on the patch. I would also welcome your advice on adding a join case to test_extensible for regression coverage. [1] https://www.postgresql.org/message-id/sSMpyoJ12CV5Y3uS-T_8Z3M1xsJBELBd-Vh5DmlC9_jy9TOqMuWzr8-TuXrBpL8AlPlyEFkyQ4wX5guJCyVizY-R2UET0DpIEeA1aSsUjk4%3D%40pm.me Best regards, Dongpo Liu
