Posting a new version of the patches, with the following updates:
- Moved the update of glob->relationOIDs (i.e. addition of partition
OIDs that plan depends on, resulting from parallel-safety checks) from
within max_parallel_hazard() to set_plan_references().
- Added an extra test for partition plan-cache invalidation.
- Simplified query_has_modifying_cte() temporary bug-fix.
- Added a comment explaining why parallel-safety of partition column
defaults is not checked.
- Minor simplification: hasSubQuery return to is_parallel_possible_for_modify().
Regards,
Greg Nancarrow
Fujitsu Australia