amcheck: Support for different header sizes of short varlena datum

In the heap, tuples may contain short varlena datum with both 1B header and 4B
headers.  But the corresponding index tuple should always have such varlena's
with 1B headers.  So, for fingerprinting, we need to convert.

Backpatch to all supported versions.

Discussion: https://postgr.es/m/flat/7bdbe559-d61a-4ae4-a6e1-48abdf3024cc%40postgrespro.ru
Author: Michael Zhilin
Reviewed-by: Alexander Lakhin, Andrey Borodin, Jian He, Alexander Korotkov
Backpatch-through: 12

diff --git a/contrib/amcheck/expected/check_btree.out b/contrib/amcheck/expected/check_btree.out
index 86b38d93f41e5de7a7796593180fc1ad272fa71e..d87e71788660ff4057a930fdf1997cdac995a9d7 100644 (file)
--- a/contrib/amcheck/expected/check_btree.out
+++ b/contrib/amcheck/expected/check_btree.out
@@ -240,6 +240,18 @@ SELECT bt_index_check('bttest_unique_nulls_b_c_idx', heapallindexed => true, che
  
 (1 row)
 
+-- Check support of both 1B and 4B header sizes of short varlena datum
+CREATE TABLE varlena_bug (v text);
+ALTER TABLE varlena_bug ALTER column v SET storage plain;
+INSERT INTO varlena_bug VALUES ('x');
+COPY varlena_bug from stdin;
+CREATE INDEX varlena_bug_idx on varlena_bug(v);
+SELECT bt_index_check('varlena_bug_idx', true);
+ bt_index_check 
+----------------
+ 
+(1 row)
+
 -- cleanup
 DROP TABLE bttest_a;
 DROP TABLE bttest_b;
@@ -250,3 +262,4 @@ DROP FUNCTION ifun(int8);
 DROP TABLE bttest_unique_nulls;
 DROP OWNED BY regress_bttest_role; -- permissions
 DROP ROLE regress_bttest_role;
+DROP TABLE varlena_bug;

diff --git a/contrib/amcheck/sql/check_btree.sql b/contrib/amcheck/sql/check_btree.sql
index aa461f7fb976b9c74c2d7a3e9da87a9568aabd99..b37fff0507876572311ac31f1c2d5df3e8f680fd 100644 (file)
--- a/contrib/amcheck/sql/check_btree.sql
+++ b/contrib/amcheck/sql/check_btree.sql
@@ -148,6 +148,16 @@ SELECT bt_index_check('bttest_unique_nulls_c_key', heapallindexed => true, check
 CREATE INDEX on bttest_unique_nulls (b,c);
 SELECT bt_index_check('bttest_unique_nulls_b_c_idx', heapallindexed => true, checkunique => true);
 
+-- Check support of both 1B and 4B header sizes of short varlena datum
+CREATE TABLE varlena_bug (v text);
+ALTER TABLE varlena_bug ALTER column v SET storage plain;
+INSERT INTO varlena_bug VALUES ('x');
+COPY varlena_bug from stdin;
+x
+\.
+CREATE INDEX varlena_bug_idx on varlena_bug(v);
+SELECT bt_index_check('varlena_bug_idx', true);
+
 -- cleanup
 DROP TABLE bttest_a;
 DROP TABLE bttest_b;
@@ -158,3 +168,4 @@ DROP FUNCTION ifun(int8);
 DROP TABLE bttest_unique_nulls;
 DROP OWNED BY regress_bttest_role; -- permissions
 DROP ROLE regress_bttest_role;
+DROP TABLE varlena_bug;

diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index 1ef4cff88e87bda711579fb614352eccfa550c21..e0dffd9bcca72281f07f9ceecd31a4517d352605 100644 (file)
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -2943,7 +2943,7 @@ bt_normalize_tuple(BtreeCheckState *state, IndexTuple itup)
    TupleDesc   tupleDescriptor = RelationGetDescr(state->rel);
    Datum       normalized[INDEX_MAX_KEYS];
    bool        isnull[INDEX_MAX_KEYS];
-   bool        toast_free[INDEX_MAX_KEYS];
+   bool        need_free[INDEX_MAX_KEYS];
    bool        formnewtup = false;
    IndexTuple  reformed;
    int         i;
@@ -2962,7 +2962,7 @@ bt_normalize_tuple(BtreeCheckState *state, IndexTuple itup)
        att = TupleDescAttr(tupleDescriptor, i);
 
        /* Assume untoasted/already normalized datum initially */
-       toast_free[i] = false;
+       need_free[i] = false;
        normalized[i] = index_getattr(itup, att->attnum,
                                      tupleDescriptor,
                                      &isnull[i]);
@@ -2985,11 +2985,32 @@ bt_normalize_tuple(BtreeCheckState *state, IndexTuple itup)
        {
            formnewtup = true;
            normalized[i] = PointerGetDatum(PG_DETOAST_DATUM(normalized[i]));
-           toast_free[i] = true;
+           need_free[i] = true;
+       }
+
+       /*
+        * Short tuples may have 1B or 4B header. Convert 4B header of short
+        * tuples to 1B
+        */
+       else if (VARATT_CAN_MAKE_SHORT(DatumGetPointer(normalized[i])))
+       {
+           /* convert to short varlena */
+           Size        len = VARATT_CONVERTED_SHORT_SIZE(DatumGetPointer(normalized[i]));
+           char       *data = palloc(len);
+
+           SET_VARSIZE_SHORT(data, len);
+           memcpy(data + 1, VARDATA(DatumGetPointer(normalized[i])), len - 1);
+
+           formnewtup = true;
+           normalized[i] = PointerGetDatum(data);
+           need_free[i] = true;
        }
    }
 
-   /* Easier case: Tuple has varlena datums, none of which are compressed */
+   /*
+    * Easier case: Tuple has varlena datums, none of which are compressed or
+    * short with 4B header
+    */
    if (!formnewtup)
        return itup;
 
@@ -2999,6 +3020,11 @@ bt_normalize_tuple(BtreeCheckState *state, IndexTuple itup)
     * (normalized input datums).  This is rather naive, but shouldn't be
     * necessary too often.
     *
+    * In the heap, tuples may contain short varlena datums with both 1B
+    * header and 4B headers.  But the corresponding index tuple should always
+    * have such varlena's with 1B headers.  So, if there is a short varlena
+    * with 4B header, we need to convert it for for fingerprinting.
+    *
     * Note that we rely on deterministic index_form_tuple() TOAST compression
     * of normalized input.
     */
@@ -3007,7 +3033,7 @@ bt_normalize_tuple(BtreeCheckState *state, IndexTuple itup)
 
    /* Cannot leak memory here */
    for (i = 0; i < tupleDescriptor->natts; i++)
-       if (toast_free[i])
+       if (need_free[i])
            pfree(DatumGetPointer(normalized[i]));
 
    return reformed;