Fix another bug in parent page splitting during GiST index build.

Yet another bug in the ilk of commits a7ee7c851 and 741b88435. In
741b88435, we took care to clear the memorized location of the
downlink when we split the parent page, because splitting the parent
page can move the downlink. But we missed that even *updating* a tuple
on the parent can move it, because updating a tuple on a gist page is
implemented as a delete+insert, so the updated tuple gets moved to the
end of the page.

This commit fixes the bug in two different ways (belt and suspenders):

1. Clear the downlink when we update a tuple on the parent page, even
   if it's not split. This the same approach as in commits a7ee7c851
   and 741b88435.

   I also noticed that gistFindCorrectParent did not clear the
   'downlinkoffnum' when it stepped to the right sibling. Fix that
   too, as it seems like a clear bug even though I haven't been able
   to find a test case to hit that.

2. Change gistFindCorrectParent so that it treats 'downlinkoffnum'
   merely as a hint. It now always first checks if the downlink is
   still at that location, and if not, it scans the page like before.
   That's more robust if there are still more cases where we fail to
   clear 'downlinkoffnum' that we haven't yet uncovered. With this,
   it's no longer necessary to meticulously clear 'downlinkoffnum',
   so this makes the previous fixes unnecessary, but I didn't revert
   them because it still seems nice to clear it when we know that the
   downlink has moved.

Also add the test case using the same test data that Alexander
posted. I tried to reduce it to a smaller test, and I also tried to
reproduce this with different test data, but I was not able to, so
let's just include what we have.

Backpatch to v12, like the previous fixes.

Reported-by: Alexander Lakhin
Discussion: https://www.postgresql.org/message-id/18129-caca016eaf0c3702@postgresql.org

1 files changed, 35 insertions, 0 deletions

diff --git a/contrib/intarray/sql/_int.sql b/contrib/intarray/sql/_int.sql
index a947c4f8fdd..dd5c668a0f4 100644
--- a/contrib/intarray/sql/_int.sql
+++ b/contrib/intarray/sql/_int.sql
@@ -195,4 +195,39 @@ SELECT count(*) from test__int WHERE a @@ '(20&23)|(50&68)';
 SELECT count(*) from test__int WHERE a @@ '20 | !21';
 SELECT count(*) from test__int WHERE a @@ '!20 & !21';
 
+DROP INDEX text_idx;
+
+-- Repeat the same queries with an extended data set. The data set is the
+-- same that we used before, except that each element in the array is
+-- repeated three times, offset by 1000 and 2000. For example, {1, 5}
+-- becomes {1, 1001, 2001, 5, 1005, 2005}.
+--
+-- That has proven to be unreasonably effective at exercising codepaths in
+-- core GiST code related to splitting parent pages, which is not covered by
+-- other tests. This is a bit out-of-place as the point is to test core GiST
+-- code rather than this extension, but there is no suitable GiST opclass in
+-- core that would reach the same codepaths.
+CREATE TABLE more__int AS SELECT
+   -- Leave alone NULLs, empty arrays and the one row that we use to test
+   -- equality
+   CASE WHEN a IS NULL OR a = '{}' OR a = '{73,23,20}' THEN a ELSE
+     (select array_agg(u) || array_agg(u + 1000) || array_agg(u + 2000) from (select unnest(a) u) x)
+   END AS a, a as b
+   FROM test__int;
+CREATE INDEX ON more__int using gist (a gist__int_ops(numranges = 252));
+
+SELECT count(*) from more__int WHERE a && '{23,50}';
+SELECT count(*) from more__int WHERE a @@ '23|50';
+SELECT count(*) from more__int WHERE a @> '{23,50}';
+SELECT count(*) from more__int WHERE a @@ '23&50';
+SELECT count(*) from more__int WHERE a @> '{20,23}';
+SELECT count(*) from more__int WHERE a <@ '{73,23,20}';
+SELECT count(*) from more__int WHERE a = '{73,23,20}';
+SELECT count(*) from more__int WHERE a @@ '50&68';
+SELECT count(*) from more__int WHERE a @> '{20,23}' or a @> '{50,68}';
+SELECT count(*) from more__int WHERE a @@ '(20&23)|(50&68)';
+SELECT count(*) from more__int WHERE a @@ '20 | !21';
+SELECT count(*) from more__int WHERE a @@ '!20 & !21';
+
+
 RESET enable_seqscan;