Prevent pgstats from getting confused when relkind of a relation changes

anarazel · anarazel · commit af3517c15c84 · 2022-12-02T18:17:54.000-08:00
When the relkind of a relache entry changes, because a table is converted into a view, pgstats can get confused in 15+, leading to crashes or assertion failures. For HEAD, Tom fixed this in b23cd18, by removing support for converting a table to a view, removing the source of the inconsistency. This commit just adds an assertion that a relcache entry's relkind does not change, just in case we end up with another case of that in the future. As there's no cases of changing relkind anymore, we can't add a test that that's handled correctly. For 15, fix the problem by not maintaining the association with the old pgstat entry when the relkind changes during a relcache invalidation processing. In that case the pgstat entry needs to be unlinked first, to avoid PgStat_TableStatus->relation getting out of sync. Also add a test reproducing the issues. No known problem exists in 11-14, so just add the test there. Reported-by: vignesh C <vignesh21@gmail.com> Author: Andres Freund <andres@anarazel.de> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Discussion: https://postgr.es/m/CALDaNm2yXz+zOtv7y5zBd5WKT8O0Ld3YxikuU3dcyCvxF7gypA@mail.gmail.com Discussion: https://postgr.es/m/CALDaNm3oZA-8Wbps2Jd1g5_Gjrr-x3YWrJPek-mF5Asrrvz2Dg@mail.gmail.com Backpatch: 15-
diff --git a/src/test/regress/expected/create_view.out b/src/test/regress/expected/create_view.out
@@ -1776,9 +1776,32 @@ select pg_get_ruledef(oid, true) from pg_rewrite
      43 AS col_b;
 (1 row)
 
+-- Test that changing the relkind of a relcache entry doesn't cause
+-- trouble. Prior instances of where it did:
+-- CALDaNm2yXz+zOtv7y5zBd5WKT8O0Ld3YxikuU3dcyCvxF7gypA@mail.gmail.com
+-- CALDaNm3oZA-8Wbps2Jd1g5_Gjrr-x3YWrJPek-mF5Asrrvz2Dg@mail.gmail.com
+CREATE TABLE tt26(c int);
+BEGIN;
+CREATE TABLE tt27(c int);
+SAVEPOINT q;
+CREATE RULE "_RETURN" AS ON SELECT TO tt27 DO INSTEAD SELECT * FROM tt26;
+SELECT * FROM tt27;
+ c 
+---
+(0 rows)
+
+ROLLBACK TO q;
+CREATE RULE "_RETURN" AS ON SELECT TO tt27 DO INSTEAD SELECT * FROM tt26;
+ROLLBACK;
+BEGIN;
+CREATE TABLE tt28(c int);
+CREATE RULE "_RETURN" AS ON SELECT TO tt28 DO INSTEAD SELECT * FROM tt26;
+CREATE RULE "_RETURN" AS ON SELECT TO tt28 DO INSTEAD SELECT * FROM tt26;
+ERROR:  "tt28" is already a view
+ROLLBACK;
 -- clean up all the random objects we made above
 \set VERBOSITY terse \\ -- suppress cascade details
 DROP SCHEMA temp_view_test CASCADE;
 NOTICE:  drop cascades to 27 other objects
 DROP SCHEMA testviewschm2 CASCADE;
-NOTICE:  drop cascades to 64 other objects
+NOTICE:  drop cascades to 65 other objects
diff --git a/src/test/regress/sql/create_view.sql b/src/test/regress/sql/create_view.sql
@@ -608,6 +608,29 @@ select pg_get_viewdef('tt23v', true);
 select pg_get_ruledef(oid, true) from pg_rewrite
   where ev_class = 'tt23v'::regclass and ev_type = '1';
 
+
+-- Test that changing the relkind of a relcache entry doesn't cause
+-- trouble. Prior instances of where it did:
+-- CALDaNm2yXz+zOtv7y5zBd5WKT8O0Ld3YxikuU3dcyCvxF7gypA@mail.gmail.com
+-- CALDaNm3oZA-8Wbps2Jd1g5_Gjrr-x3YWrJPek-mF5Asrrvz2Dg@mail.gmail.com
+CREATE TABLE tt26(c int);
+
+BEGIN;
+CREATE TABLE tt27(c int);
+SAVEPOINT q;
+CREATE RULE "_RETURN" AS ON SELECT TO tt27 DO INSTEAD SELECT * FROM tt26;
+SELECT * FROM tt27;
+ROLLBACK TO q;
+CREATE RULE "_RETURN" AS ON SELECT TO tt27 DO INSTEAD SELECT * FROM tt26;
+ROLLBACK;
+
+BEGIN;
+CREATE TABLE tt28(c int);
+CREATE RULE "_RETURN" AS ON SELECT TO tt28 DO INSTEAD SELECT * FROM tt26;
+CREATE RULE "_RETURN" AS ON SELECT TO tt28 DO INSTEAD SELECT * FROM tt26;
+ROLLBACK;
+
+
 -- clean up all the random objects we made above
 \set VERBOSITY terse \\ -- suppress cascade details
 DROP SCHEMA temp_view_test CASCADE;
