buildbot

failures in --view, --embedded, an old/non-gcc compilers

config.h.cmake

@@ -43,6 +43,7 @@
 #cmakedefine HAVE_IA64INTRIN_H 1
 #cmakedefine HAVE_IEEEFP_H 1
 #cmakedefine HAVE_INTTYPES_H 1
+#cmakedefine HAVE_IMMINTRIN_H 1
 #cmakedefine HAVE_KQUEUE 1
 #cmakedefine HAVE_LIMITS_H 1
 #cmakedefine HAVE_LINK_H 1

configure.cmake

@@ -187,6 +187,7 @@ CHECK_INCLUDE_FILES (fpu_control.h HAVE_FPU_CONTROL_H)
 CHECK_INCLUDE_FILES (grp.h HAVE_GRP_H)
 CHECK_INCLUDE_FILES (ieeefp.h HAVE_IEEEFP_H)
 CHECK_INCLUDE_FILES (inttypes.h HAVE_INTTYPES_H)
+CHECK_INCLUDE_FILES (immintrin.h HAVE_IMMINTRIN_H)
 CHECK_INCLUDE_FILES (langinfo.h HAVE_LANGINFO_H)
 CHECK_INCLUDE_FILES (link.h HAVE_LINK_H)
 CHECK_INCLUDE_FILES (linux/unistd.h HAVE_LINUX_UNISTD_H)

mysql-test/main/vector,myisam.rdiff

@@ -7,7 +7,7 @@
 +
 +MyISAM file:         datadir/test/t1#i#01
 +Record format:       Packed
-+Character set:       latin1_swedish_ci (8)
++Character set:       ? (0)
 +Data records:                    1  Deleted blocks:                 0
 +Recordlength:                   30
 +
@@ -19,7 +19,7 @@
 +
 +MyISAM file:         datadir/test/t1#i#01
 +Record format:       Packed
-+Character set:       latin1_swedish_ci (8)
++Character set:       ? (0)
 +Data records:                    0  Deleted blocks:                 0
 +Recordlength:                   30
 +

mysql-test/main/vector.result

@@ -12,7 +12,7 @@ t1	CREATE TABLE `t1` (
   `v` blob NOT NULL,
   PRIMARY KEY (`id`),
   VECTOR KEY `v` (`v`)
-) ENGINE=MyISAM DEFAULT CHARSET=latin1 COLLATE=latin1_swedish_ci
+) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci
 show keys from t1;
 Table	Non_unique	Key_name	Seq_in_index	Column_name	Collation	Cardinality	Sub_part	Packed	Null	Index_type	Comment	Index_comment	Ignored
 t1	0	PRIMARY	1	id	A	0	NULL	NULL		BTREE			NO
@@ -260,7 +260,7 @@ t2	CREATE TABLE `t2` (
   `v` blob NOT NULL,
   PRIMARY KEY (`id`),
   VECTOR KEY `v` (`v`)
-) ENGINE=MyISAM DEFAULT CHARSET=latin1 COLLATE=latin1_swedish_ci
+) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci
 drop table t1, t2;
 db.opt
 # Test insert ... select with vector index
@@ -315,7 +315,7 @@ t1	CREATE TABLE `t1` (
   `v` blob NOT NULL,
   PRIMARY KEY (`id`),
   VECTOR KEY `v` (`v`)
-) ENGINE=MyISAM DEFAULT CHARSET=latin1 COLLATE=latin1_swedish_ci
+) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci
 drop table t1;
 db.opt
 # Test RENAME TABLE with vector index

mysql-test/main/vector_innodb.result

@@ -6,7 +6,7 @@ t1	CREATE TABLE `t1` (
   `v` blob NOT NULL,
   PRIMARY KEY (`id`),
   VECTOR KEY `v` (`v`)
-) ENGINE=InnoDB DEFAULT CHARSET=latin1 COLLATE=latin1_swedish_ci
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci
 insert t1 (v) values
 (x'106d263fdf68ba3eb08d533f97d46e3fd1e1ec3edc4c123f984c563f621a233f'),
 (x'd55bee3c56eb9e3e84e3093f838dce3eb7cd653fe32d7d3f12de133c5715d23e'),

mysql-test/suite/funcs_1/r/is_columns_is_embedded.result

@@ -396,7 +396,7 @@ def	information_schema	SQL_FUNCTIONS	FUNCTION	1	NULL	YES	varchar	64	192	NULL	NUL
 def	information_schema	STATISTICS	CARDINALITY	10	NULL	YES	bigint	NULL	NULL	19	0	NULL	NULL	NULL	bigint(21)					NEVER	NULL	NO	NO
 def	information_schema	STATISTICS	COLLATION	9	NULL	YES	varchar	1	3	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(1)					NEVER	NULL	NO	NO
 def	information_schema	STATISTICS	COLUMN_NAME	8	NULL	NO	varchar	64	192	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(64)					NEVER	NULL	NO	NO
-def	information_schema	STATISTICS	COMMENT	15	NULL	YES	varchar	16	48	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(16)					NEVER	NULL	NO	NO
+def	information_schema	STATISTICS	COMMENT	15	NULL	NO	varchar	16	48	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(16)					NEVER	NULL	NO	NO
 def	information_schema	STATISTICS	IGNORED	17	NULL	NO	varchar	3	9	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(3)					NEVER	NULL	NO	NO
 def	information_schema	STATISTICS	INDEX_COMMENT	16	NULL	NO	varchar	1024	3072	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(1024)					NEVER	NULL	NO	NO
 def	information_schema	STATISTICS	INDEX_NAME	6	NULL	NO	varchar	64	192	NULL	NULL	NULL	utf8mb3	utf8mb3_general_ci	varchar(64)					NEVER	NULL	NO	NO

mysql-test/suite/sys_vars/r/sysvars_server_embedded.result

@@ -2162,6 +2162,36 @@ NUMERIC_BLOCK_SIZE	1
 ENUM_VALUE_LIST	NULL
 READ_ONLY	YES
 COMMAND_LINE_ARGUMENT	REQUIRED
+VARIABLE_NAME	MHNSW_CACHE_SIZE
+VARIABLE_SCOPE	GLOBAL
+VARIABLE_TYPE	BIGINT UNSIGNED
+VARIABLE_COMMENT	Size of the cache for the MHNSW vector index
+NUMERIC_MIN_VALUE	1048576
+NUMERIC_MAX_VALUE	18446744073709551615
+NUMERIC_BLOCK_SIZE	1
+ENUM_VALUE_LIST	NULL
+READ_ONLY	NO
+COMMAND_LINE_ARGUMENT	REQUIRED
+VARIABLE_NAME	MHNSW_LIMIT_MULTIPLIER
+VARIABLE_SCOPE	SESSION
+VARIABLE_TYPE	DOUBLE
+VARIABLE_COMMENT	Defines the number of result candidates to look for in the vector index for ORDER BY ... LIMIT N queries. Specified in term of LIMIT (1 means look for exactly N candidates, 2 means look for 2*N, etc). Larger values means the search will be slower, but the result will be closer to perfect
+NUMERIC_MIN_VALUE	1
+NUMERIC_MAX_VALUE	100
+NUMERIC_BLOCK_SIZE	NULL
+ENUM_VALUE_LIST	NULL
+READ_ONLY	NO
+COMMAND_LINE_ARGUMENT	REQUIRED
+VARIABLE_NAME	MHNSW_MAX_EDGES_PER_NODE
+VARIABLE_SCOPE	SESSION
+VARIABLE_TYPE	INT UNSIGNED
+VARIABLE_COMMENT	Larger values means slower INSERT, larger index size and higher memory consumption, but better search results. Not used for SELECT
+NUMERIC_MIN_VALUE	2
+NUMERIC_MAX_VALUE	200
+NUMERIC_BLOCK_SIZE	1
+ENUM_VALUE_LIST	NULL
+READ_ONLY	NO
+COMMAND_LINE_ARGUMENT	REQUIRED
 VARIABLE_NAME	MIN_EXAMINED_ROW_LIMIT
 VARIABLE_SCOPE	SESSION
 VARIABLE_TYPE	BIGINT UNSIGNED

sql/bloom_filters.h

@@ -28,7 +28,9 @@ SOFTWARE.
 #include <cmath>
 #include <vector>
 #include <algorithm>
+#ifdef HAVE_IMMINTRIN_H
 #include <immintrin.h>
+#endif
 
 template <typename T>
 struct PatternedSimdBloomFilter
@@ -36,7 +38,7 @@ struct PatternedSimdBloomFilter
   PatternedSimdBloomFilter(int n, float eps) : n(n), epsilon(eps)
   {
     m = ComputeNumBits();
-    int log_num_blocks = 32 - __builtin_clz(m) - rotate_bits;
+    int log_num_blocks = my_bit_log2_uint32(m) + 1 - rotate_bits;
     num_blocks = (1ULL << log_num_blocks);
     bv.resize(num_blocks);
   }
@@ -47,6 +49,7 @@ struct PatternedSimdBloomFilter
     return std::max<uint64_t>(512, bits_per_val * n + 0.5);
   }
 
+#if __GNUC__ > 7 && defined(HAVE_IMMINTRIN_H)
   __attribute__ ((target ("avx2,avx,fma")))
   __m256i CalcHash(__m256i vecData)
   {
@@ -159,6 +162,7 @@ struct PatternedSimdBloomFilter
     uint8_t res_bits = static_cast<uint8_t>(_mm256_movemask_epi8(_mm256_set1_epi64x(res_bytes)) & 0xff);
     return res_bits;
   }
+#endif
 
   /********************************************************
   ********* non-SIMD fallback version ********************/

sql/item_create.cc

@@ -6238,7 +6238,7 @@ Create_func_year_week::create_native(THD *thd, const LEX_CSTRING *name,
 class Create_func_vec_distance: public Create_func_arg2
 {
 public:
-  Item *create_2_arg(THD *thd, Item *arg1, Item *arg2)
+  Item *create_2_arg(THD *thd, Item *arg1, Item *arg2) override
   { return new (thd->mem_root) Item_func_vec_distance(thd, arg1, arg2); }
 
   static Create_func_vec_distance s_singleton;
@@ -6254,7 +6254,7 @@ Create_func_vec_distance Create_func_vec_distance::s_singleton;
 class Create_func_vec_totext: public Create_func_arg1
 {
 public:
-  Item *create_1_arg(THD *thd, Item *arg1)
+  Item *create_1_arg(THD *thd, Item *arg1) override
   { return new (thd->mem_root) Item_func_vec_totext(thd, arg1); }
 
   static Create_func_vec_totext s_singleton;
@@ -6271,7 +6271,7 @@ Create_func_vec_totext Create_func_vec_totext::s_singleton;
 class Create_func_vec_fromtext: public Create_func_arg1
 {
 public:
-  Item *create_1_arg(THD *thd, Item *arg1)
+  Item *create_1_arg(THD *thd, Item *arg1) override
   { return new (thd->mem_root) Item_func_vec_fromtext(thd, arg1); }
 
   static Create_func_vec_fromtext s_singleton;

sql/item_vectorfunc.cc

@@ -59,11 +59,10 @@ double Item_func_vec_distance::val_real()
 
 bool Item_func_vec_totext::fix_length_and_dec(THD *thd)
 {
-  collation.set(default_charset());
   decimals= 0;
   max_length= ((args[0]->max_length / 4) *
-               (MAX_FLOAT_STR_LENGTH + 1 /* comma */)) + 2 /* braces */ *
-               collation.collation->mbmaxlen;
+               (MAX_FLOAT_STR_LENGTH + 1 /* comma */)) + 2 /* braces */;
+  fix_length_and_charset(max_length, default_charset());
   set_maybe_null();
   return false;
 }
@@ -140,13 +139,12 @@ Item_func_vec_fromtext::Item_func_vec_fromtext(THD *thd, Item *a)
 
 bool Item_func_vec_fromtext::fix_length_and_dec(THD *thd)
 {
-  collation.set(&my_charset_bin);
   decimals= 0;
   /* Worst case scenario, for a valid input we have a string of the form:
      [1,2,3,4,5,...] single digit numbers.
-     This means we can have (max_length - 2) / 2 floats.
-     Each float takes 4 bytes, so we do max_length - 2 * 2. */
-  max_length= (args[0]->max_length - 2) * 2 * my_charset_bin.mbmaxlen;
+     This means we can have (max_length - 1) / 2 floats.
+     Each float takes 4 bytes, so we do (max_length - 1) * 2. */
+  fix_length_and_charset((args[0]->max_length - 1) * 2, &my_charset_bin);
   set_maybe_null();
   return false;
 }

sql/item_vectorfunc.h

@@ -40,6 +40,11 @@ class Item_func_vec_distance: public Item_real_func
 public:
   Item_func_vec_distance(THD *thd, Item *a, Item *b)
    :Item_real_func(thd, a, b) {}
+  bool fix_length_and_dec(THD *thd) override
+  {
+    set_maybe_null();
+    return Item_real_func::fix_length_and_dec(thd);
+  }
   double val_real() override;
   LEX_CSTRING func_name_cstring() const override
   {

sql/sql_base.cc

@@ -9829,17 +9829,17 @@ int TABLE::hlindex_open(uint nr)
     mysql_mutex_lock(&s->LOCK_share);
     if (!s->hlindex)
     {
+      mysql_mutex_unlock(&s->LOCK_share);
       TABLE_SHARE *share;
       char *path= NULL;
-      mysql_mutex_unlock(&s->LOCK_share);
+      size_t path_len= s->normalized_path.length + HLINDEX_BUF_LEN;
 
       share= (TABLE_SHARE*)alloc_root(&s->mem_root, sizeof(*share));
-      path= (char*)alloc_root(&s->mem_root,
-                              s->normalized_path.length + HLINDEX_BUF_LEN);
+      path= (char*)alloc_root(&s->mem_root, path_len);
       if (!share || !path)
         return 1;
 
-      my_snprintf(path, HLINDEX_BUF_LEN, "%s" HLINDEX_TEMPLATE,
+      my_snprintf(path, path_len, "%s" HLINDEX_TEMPLATE,
                   s->normalized_path.str, nr);
       init_tmp_table_share(in_use, share, s->db.str, 0, s->table_name.str,
                            path, false);

sql/sql_type.h

@@ -7228,7 +7228,7 @@ class Type_handler_string: public Type_handler_longstr
   bool Key_part_spec_init_ft(Key_part_spec *part,
                              const Column_definition &def) const override;
   bool Key_part_spec_init_vector(Key_part_spec *part,
-                                 const Column_definition &def) const;
+                                 const Column_definition &def) const override;
   Field *make_table_field(MEM_ROOT *root,
                           const LEX_CSTRING *name,
                           const Record_addr &addr,
@@ -7327,7 +7327,7 @@ class Type_handler_varchar: public Type_handler_longstr
   bool Key_part_spec_init_ft(Key_part_spec *part,
                              const Column_definition &def) const override;
   bool Key_part_spec_init_vector(Key_part_spec *part,
-                                 const Column_definition &def) const;
+                                 const Column_definition &def) const override;
   Field *make_table_field(MEM_ROOT *root,
                           const LEX_CSTRING *name,
                           const Record_addr &addr,
@@ -7431,7 +7431,7 @@ class Type_handler_blob_common: public Type_handler_longstr
   bool Key_part_spec_init_ft(Key_part_spec *part,
                              const Column_definition &def) const override;
   bool Key_part_spec_init_vector(Key_part_spec *part,
-                                 const Column_definition &def) const;
+                                 const Column_definition &def) const override;
   bool Key_part_spec_init_primary(Key_part_spec *part,
                                   const Column_definition &def,
                                   const handler *file) const override;

sql/vector_mhnsw.cc

@@ -89,7 +89,7 @@ struct FVector
     return vec;
   }
 
-#if __GNUC__ > 7
+#if __GNUC__ > 7 && defined(HAVE_IMMINTRIN_H)
   __attribute__ ((target ("avx2,avx")))
   static float dot_product(const int16_t *v1, const int16_t *v2, size_t len)
   {
@@ -473,18 +473,24 @@ int MHNSW_Trx::MHNSW_hton::do_savepoint_rollback(handlerton *, THD *thd, void *)
 
 int MHNSW_Trx::MHNSW_hton::do_rollback(handlerton *, THD *thd, bool)
 {
+  MHNSW_Trx *trx_next;
   for (auto trx= static_cast<MHNSW_Trx*>(thd_get_ha_data(thd, &hton));
-       trx; trx= trx->next)
+       trx; trx= trx_next)
+  {
+    trx_next= trx->next;
     trx->~MHNSW_Trx();
+  }
   thd_set_ha_data(current_thd, &hton, nullptr);
   return 0;
 }
 
 int MHNSW_Trx::MHNSW_hton::do_commit(handlerton *, THD *thd, bool)
 {
+  MHNSW_Trx *trx_next;
   for (auto trx= static_cast<MHNSW_Trx*>(thd_get_ha_data(thd, &hton));
-       trx; trx= trx->next)
+       trx; trx= trx_next)
   {
+    trx_next= trx->next;
     auto ctx= MHNSW_Context::get_from_share(trx->table_share, nullptr);
     if (ctx)
     {