Re: BUG #5673: Optimizer creates strange execution plan leading to wrong results

David Schmitt <david(at)dasz(dot)at> writes:
> Executing the attached example.sql on a fresh database demonstrates the
> problem I'm seeing:

Found it. If you need a patch right away, it's a one-liner:

diff --git a/src/backend/optimizer/path/allpaths.c b/src/backend/optimizer/path/allpaths.c
index 2d86da3..b7cf0b8 100644
--- a/src/backend/optimizer/path/allpaths.c
+++ b/src/backend/optimizer/path/allpaths.c
@@ -425,7 +425,7 @@ set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
childpath = childrel->cheapest_total_path;
if (IsA(childpath, AppendPath))
subpaths = list_concat(subpaths,
- ((AppendPath *) childpath)->subpaths);
+ list_copy(((AppendPath *) childpath)->subpaths));
else
subpaths = lappend(subpaths, childpath);

regards, tom lane