diff options
author | Andrew Gierth <rhodiumtoad@postgresql.org> | 2018-03-21 10:42:04 +0000 |
---|---|---|
committer | Andrew Gierth <rhodiumtoad@postgresql.org> | 2018-03-21 11:39:28 +0000 |
commit | d2d79887eadff72c339a072ef693bb6016651d30 (patch) | |
tree | dc777dcbe00114aef09beeb6bdc2eea6524202b7 /src/backend/optimizer/plan/planner.c | |
parent | aea7c17e86e99a7ed4da489b3df2b5493b5e5e95 (diff) | |
download | postgresql-d2d79887eadff72c339a072ef693bb6016651d30.tar.gz postgresql-d2d79887eadff72c339a072ef693bb6016651d30.zip |
Repair crash with unsortable grouping sets.
If there were multiple grouping sets, none of them empty, all of which
were unsortable, then an oversight in consider_groupingsets_paths led
to a null pointer dereference. Fix, and add a regression test for this
case.
Per report from Dang Minh Huong, though I didn't use their patch.
Backpatch to 10.x where hashed grouping sets were added.
Diffstat (limited to 'src/backend/optimizer/plan/planner.c')
-rw-r--r-- | src/backend/optimizer/plan/planner.c | 23 |
1 files changed, 22 insertions, 1 deletions
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c index b452da02042..85805ff5c70 100644 --- a/src/backend/optimizer/plan/planner.c +++ b/src/backend/optimizer/plan/planner.c @@ -3997,7 +3997,28 @@ consider_groupingsets_paths(PlannerInfo *root, Assert(can_hash); - if (pathkeys_contained_in(root->group_pathkeys, path->pathkeys)) + /* + * If the input is coincidentally sorted usefully (which can happen + * even if is_sorted is false, since that only means that our caller + * has set up the sorting for us), then save some hashtable space by + * making use of that. But we need to watch out for degenerate cases: + * + * 1) If there are any empty grouping sets, then group_pathkeys might + * be NIL if all non-empty grouping sets are unsortable. In this case, + * there will be a rollup containing only empty groups, and the + * pathkeys_contained_in test is vacuously true; this is ok. + * + * XXX: the above relies on the fact that group_pathkeys is generated + * from the first rollup. If we add the ability to consider multiple + * sort orders for grouping input, this assumption might fail. + * + * 2) If there are no empty sets and only unsortable sets, then the + * rollups list will be empty (and thus l_start == NULL), and + * group_pathkeys will be NIL; we must ensure that the vacuously-true + * pathkeys_contain_in test doesn't cause us to crash. + */ + if (l_start != NULL && + pathkeys_contained_in(root->group_pathkeys, path->pathkeys)) { unhashed_rollup = lfirst_node(RollupData, l_start); exclude_groups = unhashed_rollup->numGroups; |