!1772 修复在部分场景下，非sysadmin用户执行gs_dump失败的问题

Merge pull request !1772 from pengjiong/drop_ext

build/script/utils/common.sh

@@ -72,7 +72,7 @@ select_package_command
 export PLAT_FORM_STR=$(sh "${ROOT_DIR}/src/get_PlatForm_str.sh")
 if [ "${PLAT_FORM_STR}"x == "Failed"x -o "${PLAT_FORM_STR}"x == ""x ]
 then
-    echo "We only support openEuler(aarch64), EulerOS(aarch64), CentOS, Kylin(aarch64) platform."
+    echo "We only support openEuler(aarch64), EulerOS(aarch64), CentOS, Kylin(aarch64), Asianux platform."
     exit 1;
 fi
 
@@ -96,16 +96,21 @@ elif [[ "$PLAT_FORM_STR" =~ "kylin" ]]; then
     if [ "$PLATFORM_ARCH"X == "aarch64"X ];then 
         GAUSSDB_EXTRA_FLAGS=" -D__USE_NUMA"
     fi
+elif [[ "$PLAT_FORM_STR" =~ "asianux" ]]; then
+    dist_version="Asianux"
+    if [ "$PLATFORM_ARCH"X == "aarch64"X ];then
+        GAUSSDB_EXTRA_FLAGS=" -D__USE_NUMA"
+    fi
 else
-    echo "We only support openEuler(aarch64), EulerOS(aarch64), CentOS, Kylin(aarch64) platform."
+    echo "We only support openEuler(aarch64), EulerOS(aarch64), CentOS, Kylin(aarch64), Asianux platform."
     echo "Kernel is $kernel"
     exit 1
 fi
 
 ##add platform architecture information
 if [ "$PLATFORM_ARCH"X == "aarch64"X ] ; then
-    if [ "$dist_version" != "openEuler" ] && [ "$dist_version" != "EulerOS" ] && [ "$dist_version" != "Kylin" ] ; then
-        echo "We only support NUMA on openEuler(aarch64), EulerOS(aarch64), Kylin(aarch64) platform."
+    if [ "$dist_version" != "openEuler" ] && [ "$dist_version" != "EulerOS" ] && [ "$dist_version" != "Kylin" ] && [ "$dist_version" != "Asianux" ]; then
+        echo "We only support NUMA on openEuler(aarch64), EulerOS(aarch64), Kylin(aarch64), Asianux platform."
         exit 1
     fi
 fi

doc/src/sgml/ref/allfiles.sgmlin

@@ -26,6 +26,7 @@ Complete list of usable sgml source files in this directory.
 <!ENTITY alterOperator      SYSTEM "alter_operator.sgml">
 <!ENTITY alterOperatorClass SYSTEM "alter_opclass.sgml">
 <!ENTITY alterOperatorFamily SYSTEM "alter_opfamily.sgml">
+<!ENTITY alterProcedure     SYSTEM "alter_procedure.sgml">
 <!ENTITY alterRole          SYSTEM "alter_role.sgml">
 <!ENTITY alterSchema        SYSTEM "alter_schema.sgml">
 <!ENTITY alterServer        SYSTEM "alter_server.sgml">

doc/src/sgml/ref/alter_procedure.sgml

@@ -0,0 +1,37 @@
+<refentry id="sql-alterprocedure">
+ <indexterm zone="sql-alterprocedure">
+  <primary>ALTER PROCEDURE</primary>
+ </indexterm>
+
+ <refmeta>
+  <refentrytitle>ALTER PROCEDURE</refentrytitle>
+  <manvolnum>7</manvolnum>
+  <refmiscinfo>SQL - Language Statements</refmiscinfo>
+ </refmeta>
+
+ <refnamediv>
+  <refname>ALTER PROCEDURE</refname>
+  <refpurpose>change the definition of a procedure</refpurpose>
+ </refnamediv>
+
+ <refsynopsisdiv>
+<synopsis>
+ALTER PROCEDURE <replaceable>name</replaceable> [ ( [ [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">argtype</replaceable> [, ...] ] ) ]
+    <replaceable class="parameter">action</replaceable> [ ... ] [ RESTRICT ]
+ALTER PROCEDURE <replaceable>name</replaceable> [ ( [ [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">argtype</replaceable> [, ...] ] ) ]
+    RENAME TO <replaceable>new_name</replaceable>
+ALTER PROCEDURE <replaceable>name</replaceable> [ ( [ [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">argtype</replaceable> [, ...] ] ) ]
+    OWNER TO { <replaceable>new_owner</replaceable> | CURRENT_ROLE | CURRENT_USER | SESSION_USER }
+ALTER PROCEDURE <replaceable>name</replaceable> [ ( [ [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">argtype</replaceable> [, ...] ] ) ]
+    SET SCHEMA <replaceable>new_schema</replaceable>
+    
+<phrase>where <replaceable class="parameter">action</replaceable> is one of:</phrase>
+
+    [ EXTERNAL ] SECURITY INVOKER | [ EXTERNAL ] SECURITY DEFINER
+    SET <replaceable class="parameter">configuration_parameter</replaceable> { TO | = } { <replaceable class="parameter">value</replaceable> | DEFAULT }
+    SET <replaceable class="parameter">configuration_parameter</replaceable> FROM CURRENT
+    RESET <replaceable class="parameter">configuration_parameter</replaceable>
+    RESET ALL
+</synopsis>
+ </refsynopsisdiv>
+ </refentry>

doc/src/sgml/ref/drop_table.sgmlin

@@ -11,7 +11,7 @@
 <refsynopsisdiv>
 <synopsis>
 DROP TABLE [ IF EXISTS ] 
-{[schema.]table_name} [, ...] [ CASCADE | RESTRICT ] [ PURGE ]};
+{[schema.]table_name} [, ...] [ CASCADE | RESTRICT ] [ PURGE ];
 </synopsis>
 </refsynopsisdiv>
 </refentry>

src/bin/pg_ctl/pg_ctl.cpp

@@ -5923,7 +5923,7 @@ int main(int argc, char** argv)
             &option_index)) != -1)
 #endif
 #else
-        while ((c = getopt_long(argc, argv, "b:cD:e:fi:G:l:m:M:N:o:O:p:P:r:R:v:x:sS:t:u:U:wWZ:dqL:T:Q:", long_options,
+        while ((c = getopt_long(argc, argv, "b:cD:e:fi:G:l:m:M:N:o:O:p:P:r:R:v:x:sS:t:u:U:wWZ:C:dqL:T:Q:", long_options,
             &option_index)) != -1)
 #endif
 #endif

src/bin/pg_dump/pg_dump.cpp

@@ -233,6 +233,7 @@ char* all_data_nodename_list = NULL;
 const uint32 USTORE_UPGRADE_VERSION = 92368;
 const uint32 PACKAGE_ENHANCEMENT = 92444;
 const uint32 SUBSCRIPTION_VERSION = 92580;
+const uint32 SUBSCRIPTION_BINARY_VERSION_NUM = 92606;
 
 #ifdef DUMPSYSLOG
 char* syslogpath = NULL;
@@ -4444,23 +4445,16 @@ void getSubscriptions(Archive *fout)
     int i_subslotname;
     int i_subsynccommit;
     int i_subpublications;
-    int i, ntups;
+    int i_subbinary;
+    int i;
+    int ntups;
 
     if (no_subscriptions || GetVersionNum(fout) < SUBSCRIPTION_VERSION) {
         return;
     }
 
     if (!isExecUserSuperRole(fout)) {
-        res = ExecuteSqlQuery(fout,
-            "SELECT count(*) FROM pg_subscription "
-            "WHERE subdbid = (SELECT oid FROM pg_catalog.pg_database"
-            "                 WHERE datname = current_database())",
-            PGRES_TUPLES_OK);
-        uint64 n = (res != NULL) ? strtoul(PQgetvalue(res, 0, 0), NULL, 10) : 0;
-        if (n > 0) {
-            write_msg(NULL, "WARNING: subscriptions not dumped because current user is not a superuser\n");
-        }
-        PQclear(res);
+        write_msg(NULL, "WARNING: subscriptions not dumped because current user is not a superuser\n");
         return;
     }
 
@@ -4469,14 +4463,20 @@ void getSubscriptions(Archive *fout)
     resetPQExpBuffer(query);
 
     /* Get the subscriptions in current database. */
-    appendPQExpBuffer(query,
-        "SELECT s.tableoid, s.oid, s.subname,"
-        "(%s s.subowner) AS rolname, "
-        " s.subconninfo, s.subslotname, s.subsynccommit, s.subpublications "
-        "FROM pg_catalog.pg_subscription s "
+    appendPQExpBuffer(query, "SELECT s.tableoid, s.oid, s.subname,"
+        "(%s s.subowner) AS rolname, s.subconninfo, s.subslotname, "
+        "s.subsynccommit, s.subpublications, \n", username_subquery);
+
+    if (GetVersionNum(fout) >= SUBSCRIPTION_BINARY_VERSION_NUM) {
+        appendPQExpBuffer(query, " s.subbinary\n");
+    } else {
+        appendPQExpBuffer(query, " false AS subbinary\n");
+    }
+
+    appendPQExpBuffer(query, "FROM pg_catalog.pg_subscription s "
         "WHERE s.subdbid = (SELECT oid FROM pg_catalog.pg_database"
-        "                   WHERE datname = current_database())",
-        username_subquery);
+        "                   WHERE datname = current_database())");
+
     res = ExecuteSqlQuery(fout, query->data, PGRES_TUPLES_OK);
 
     ntups = PQntuples(res);
@@ -4494,6 +4494,7 @@ void getSubscriptions(Archive *fout)
     i_subslotname = PQfnumber(res, "subslotname");
     i_subsynccommit = PQfnumber(res, "subsynccommit");
     i_subpublications = PQfnumber(res, "subpublications");
+    i_subbinary = PQfnumber(res, "subbinary");
 
     subinfo = (SubscriptionInfo *)pg_malloc(ntups * sizeof(SubscriptionInfo));
 
@@ -4512,6 +4513,7 @@ void getSubscriptions(Archive *fout)
         }
         subinfo[i].subsynccommit = gs_strdup(PQgetvalue(res, i, i_subsynccommit));
         subinfo[i].subpublications = gs_strdup(PQgetvalue(res, i, i_subpublications));
+        subinfo[i].subbinary = gs_strdup(PQgetvalue(res, i, i_subbinary));
 
         if (strlen(subinfo[i].rolname) == 0) {
             write_msg(NULL, "WARNING: owner of subscription \"%s\" appears to be invalid\n", subinfo[i].dobj.name);
@@ -4578,6 +4580,10 @@ static void dumpSubscription(Archive *fout, const SubscriptionInfo *subinfo)
         appendPQExpBufferStr(query, "NONE");
     }
 
+    if (strcmp(subinfo->subbinary, "t") == 0) {
+        appendPQExpBuffer(query, ", binary = true");
+    }
+
     if (strcmp(subinfo->subsynccommit, "off") != 0) {
         appendPQExpBuffer(query, ", synchronous_commit = %s", fmtId(subinfo->subsynccommit));
     }
@@ -10780,6 +10786,11 @@ static void dumpDirectory(Archive* fout)
     char* dirpath = NULL;
     char* diracl = NULL;
 
+    if (!isExecUserSuperRole(fout)) {
+        write_msg(NULL, "WARNING: directory not dumped because current user is not a superuser\n");
+        return;
+    }
+
     /* Make sure we are in proper schema */
     selectSourceSchema(fout, "pg_catalog");
 
@@ -21389,6 +21400,11 @@ static void dumpSynonym(Archive* fout)
     PQExpBuffer q;
     PQExpBuffer delq;
 
+    if (!isExecUserSuperRole(fout)) {
+        write_msg(NULL, "WARNING: synonym not dumped because current user is not a superuser\n");
+        return;
+    }
+
     selectSourceSchema(fout, "pg_catalog");
     query = createPQExpBuffer();
     printfPQExpBuffer(query,

src/bin/pg_dump/pg_dump.h

@@ -498,6 +498,7 @@ typedef struct _SubscriptionInfo {
     char *subslotname;
     char *subsynccommit;
     char *subpublications;
+    char *subbinary;
 } SubscriptionInfo;
 
 /* global decls */

src/bin/pg_probackup/backup.cpp

@@ -31,6 +31,9 @@
    it will be backuped up in external dirs  */
 parray *pgdata_nobackup_dir = NULL;
 
+/* list of logical replication slots */
+parray *logical_replslot = NULL;
+
 static int standby_message_timeout_local = 10 ;	/* 10 sec = default */
 static XLogRecPtr stop_backup_lsn = InvalidXLogRecPtr;
 static XLogRecPtr stop_stream_lsn = InvalidXLogRecPtr;
@@ -89,10 +92,11 @@ static void backup_cleanup(bool fatal, void *userdata);
 
 static void *backup_files(void *arg);
 
-static void do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool backup_logs);
+static void do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool backup_logs,
+    bool backup_replslots);
 
 static void pg_start_backup(const char *label, bool smooth, pgBackup *backup,
-                                PGNodeInfo *nodeInfo, PGconn *conn);
+                                PGNodeInfo *nodeInfo, PGconn *conn, bool backup_replslots);
 static void pg_stop_backup(pgBackup *backup, PGconn *pg_startbackup_conn, PGNodeInfo *nodeInfo);
 static int checkpoint_timeout(PGconn *backup_conn);
 
@@ -558,7 +562,7 @@ static void sync_files(parray *database_map, const char *database_path, parray *
  * Move files from 'pgdata' to a subdirectory in 'backup_path'.
  */
 static void
-do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool backup_logs)
+do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool backup_logs, bool backup_replslots)
 {
     int i;
     char    database_path[MAXPGPATH];
@@ -591,7 +595,7 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool
     securec_check_c(rc, "\0", "\0");
 
     /* Call pg_start_backup function in openGauss connect */
-    pg_start_backup(label, smooth_checkpoint, &current, nodeInfo, backup_conn);
+    pg_start_backup(label, smooth_checkpoint, &current, nodeInfo, backup_conn, backup_replslots);
 
     /* Obtain current timeline */
 #if PG_VERSION_NUM >= 90600
@@ -624,10 +628,10 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool
     /* list files with the logical path. omit $PGDATA */
     if (fio_is_remote(FIO_DB_HOST))
         fio_list_dir(backup_files_list, instance_config.pgdata,
-                            true, true, false, backup_logs, true, 0);
+                            true, true, false, backup_logs, true, 0, backup_replslots);
     else
         dir_list_file(backup_files_list, instance_config.pgdata,
-                             true, true, false, backup_logs, true, 0, FIO_LOCAL_HOST);
+                             true, true, false, backup_logs, true, 0, FIO_LOCAL_HOST, backup_replslots);
 
     /*
      * Get database_map (name to oid) for use in partial restore feature.
@@ -749,6 +753,11 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool
     }
     pgdata_nobackup_dir = NULL;
 
+    if (logical_replslot) {
+        free_dir_list(logical_replslot);
+    }
+    logical_replslot = NULL;
+
     /* Cleanup */
     if (backup_list)
     {
@@ -849,7 +858,7 @@ static void do_after_backup()
  */
 int
 do_backup(time_t start_time, pgSetBackupParams *set_backup_params,
-        bool no_validate, bool no_sync, bool backup_logs)
+        bool no_validate, bool no_sync, bool backup_logs, bool backup_replslots)
 {
     PGconn  *backup_conn = NULL;
     PGNodeInfo  nodeInfo;
@@ -925,7 +934,7 @@ do_backup(time_t start_time, pgSetBackupParams *set_backup_params,
         add_note(&current, set_backup_params->note);
 
     /* backup data */
-    do_backup_instance(backup_conn, &nodeInfo, no_sync, backup_logs);
+    do_backup_instance(backup_conn, &nodeInfo, no_sync, backup_logs, backup_replslots);
     pgut_atexit_pop(backup_cleanup, NULL);
 
     /* compute size of wal files of this backup stored in the archive */
@@ -1034,13 +1043,15 @@ confirm_block_size(PGconn *conn, const char *name, int blcksz)
  */
 static void
 pg_start_backup(const char *label, bool smooth, pgBackup *backup,
-                PGNodeInfo *nodeInfo, PGconn *conn)
+                PGNodeInfo *nodeInfo, PGconn *conn, bool backup_replslots)
 {
     PGresult   *res;
     const char *params[2];
     uint32  lsn_hi;
     uint32  lsn_lo;
     int     ret;
+    int i;
+    XLogRecPtr startLsn;
 
     params[0] = label;
 
@@ -1068,7 +1079,33 @@ pg_start_backup(const char *label, bool smooth, pgBackup *backup,
     XLogDataFromLSN(ret, PQgetvalue(res, 0, 0), &lsn_hi, &lsn_lo);
     securec_check_for_sscanf_s(ret, 2, "\0", "\0");
     /* Calculate LSN */
-    backup->start_lsn = ((uint64) lsn_hi )<< 32 | lsn_lo;
+    startLsn = ((uint64) lsn_hi )<< 32 | lsn_lo;
+
+    if (backup_replslots) {
+        logical_replslot = parray_new();
+        /* query for logical replication slots of subscriptions */
+        res = pgut_execute(conn,
+                           "SELECT slot_name, restart_lsn FROM pg_catalog.pg_get_replication_slots()"
+                           "WHERE slot_type = 'logical' AND plugin = 'pgoutput'", 0, NULL);
+        if (PQntuples(res) == 0) {
+            elog(LOG, "logical replication slots for subscriptions not found");
+        } else {
+            XLogRecPtr repslotLsn;
+
+            for (i = 0; i < PQntuples(res); i++) {
+                XLogDataFromLSN(ret, PQgetvalue(res, i, 1), &lsn_hi, &lsn_lo);
+                securec_check_for_sscanf_s(ret, 2, "\0", "\0");
+                repslotLsn = ((uint64) lsn_hi )<< 32 | lsn_lo;
+                startLsn = Min(startLsn, repslotLsn);
+
+                char* slotname = pg_strdup(PQgetvalue(res, i, 0));
+                parray_append(logical_replslot, slotname);
+            }
+            elog(WARNING, "logical replication slots for subscriptions will be backed up. "
+                 "If don't use them after restoring, please drop them to avoid affecting xlog recycling.");
+        }
+    }
+    backup->start_lsn = startLsn;
 
     PQclear(res);
 }

src/bin/pg_probackup/dir.cpp

@@ -42,13 +42,6 @@ const char *pgdata_exclude_dir[] =
     (const char *)"pg_stat_tmp",
     (const char *)"pgsql_tmp",
 
-    /*
-    * It is generally not useful to backup the contents of this directory even
-    * if the intention is to restore to another master. See backup.sgml for a
-    * more detailed description.
-    */
-    (const char *)"pg_replslot",
-
     /* Contents removed on startup, see dsm_cleanup_for_mmap(). */
     (const char *)"pg_dynshmem",
 
@@ -68,7 +61,7 @@ const char *pgdata_exclude_dir[] =
     (const char *)"pg_subtrans",
 
     /* end of list */
-    NULL,   /* pg_log will be set later */
+    NULL,   /* pg_log and pg_replslot will be set later */
     NULL
 };
 
@@ -128,17 +121,20 @@ may be removed int the future */
 
 static int pgCompareString(const void *str1, const void *str2);
 
-static char dir_check_file(pgFile *file, bool backup_logs);
+static char dir_check_file(pgFile *file, bool backup_logs, bool backup_replslots);
 static char check_in_tablespace(pgFile *file, bool in_tablespace);
 static char check_db_dir(pgFile *file);
 static char check_digit_file(pgFile *file);
 static char check_nobackup_dir(pgFile *file);
 static void dir_list_file_internal(parray *files, pgFile *parent, const char *parent_dir,
                                                     bool exclude, bool follow_symlink, bool backup_logs,
-                                                    bool skip_hidden, int external_dir_num, fio_location location);
+                                                    bool skip_hidden, int external_dir_num, fio_location location,
+                                                    bool backup_replslots);
 static void opt_path_map(ConfigOption *opt, const char *arg,
                                              TablespaceList *list, const char *type);
 
+char check_logical_replslot_dir(const char *rel_path);
+
 /* Tablespace mapping */
 static TablespaceList tablespace_dirs = {NULL, NULL};
 /* Extra directories mapping */
@@ -538,7 +534,7 @@ db_map_entry_free(void *entry)
 void
 dir_list_file(parray *files, const char *root, bool exclude, bool follow_symlink,
                         bool add_root, bool backup_logs, bool skip_hidden, int external_dir_num,
-                        fio_location location)
+                        fio_location location, bool backup_replslots)
 {
     pgFile  *file;
 
@@ -565,7 +561,7 @@ dir_list_file(parray *files, const char *root, bool exclude, bool follow_symlink
         parray_append(files, file);
 
     dir_list_file_internal(files, file, root, exclude, follow_symlink,
-                                        backup_logs, skip_hidden, external_dir_num, location);
+                                        backup_logs, skip_hidden, external_dir_num, location, backup_replslots);
 
     if (!add_root)
         pgFileFree(file);
@@ -589,7 +585,7 @@ dir_list_file(parray *files, const char *root, bool exclude, bool follow_symlink
  * - datafiles
  */
 static char
-dir_check_file(pgFile *file, bool backup_logs)
+dir_check_file(pgFile *file, bool backup_logs, bool backup_replslots)
 {
     int     i;
     int     sscanf_res;
@@ -652,6 +648,29 @@ dir_check_file(pgFile *file, bool backup_logs)
             }
         }
 
+        /*
+         * Backup pg_replslot if it is specified.
+         * It is generally not useful to backup the contents of this directory even
+         * if the intention is to restore to another master. See backup.sgml for a
+         * more detailed description.
+         */
+        if (!backup_replslots) {
+            if (strcmp(file->rel_path, PG_REPLSLOT_DIR) == 0) {
+                /* Skip */
+                elog(VERBOSE, "Excluding directory content: %s", file->rel_path);
+                return CHECK_EXCLUDE_FALSE;
+            }
+        } else {
+            /* 
+             * Check file that under pg_replslot and judge whether it
+             * belonged to logical replication slots for subscriptions.
+             */
+            if (strcmp(file->rel_path, PG_REPLSLOT_DIR) != 0 &&
+                path_is_prefix_of_path(PG_REPLSLOT_DIR, file->rel_path)) {
+                return check_logical_replslot_dir(file->rel_path);
+            }
+        }
+
         ret = check_nobackup_dir(file);
         if (ret != -1) {  /* -1 means need backup */
             return ret;
@@ -749,6 +768,35 @@ static char check_nobackup_dir(pgFile *file)
     return ret;
 }
 
+char check_logical_replslot_dir(const char *rel_path)
+{
+    char ret = CHECK_FALSE;
+    int i = 0;
+    char *tmp = pg_strdup(rel_path);
+    char *p;
+#define DIRECTORY_DELIMITER "/"
+
+    if (logical_replslot) {
+        /* extract slot name from rel_path, such as sub1 from pg_replslot/sub1/snap */
+        p = strtok(tmp, DIRECTORY_DELIMITER);
+        if (p != NULL) {
+            p = strtok(NULL, DIRECTORY_DELIMITER);
+        }
+
+        for (i = 0; p != NULL && i < (int)parray_num(logical_replslot); i++) {
+            char *slotName = (char *)parray_get(logical_replslot, i);
+            if (strcmp(p, slotName) == 0) {
+                pfree(tmp);
+                return CHECK_TRUE;
+            }
+        }
+    } else {
+        ret = CHECK_TRUE;
+    }
+    pfree(tmp);
+    return ret;
+}
+
 static char check_db_dir(pgFile *file)
 {
     char ret = -1;
@@ -889,7 +937,8 @@ bool SkipSomeDirFile(pgFile *file, struct dirent *dent, bool skipHidden)
 static void
 dir_list_file_internal(parray *files, pgFile *parent, const char *parent_dir,
                                         bool exclude, bool follow_symlink, bool backup_logs,
-                                        bool skip_hidden, int external_dir_num, fio_location location)
+                                        bool skip_hidden, int external_dir_num, fio_location location,
+                                        bool backup_replslots)
 {
     DIR     *dir;
     struct dirent *dent;
@@ -937,7 +986,7 @@ dir_list_file_internal(parray *files, pgFile *parent, const char *parent_dir,
 
         if (exclude)
         {
-            check_res = dir_check_file(file, backup_logs);
+            check_res = dir_check_file(file, backup_logs, backup_replslots);
             if (check_res == CHECK_FALSE)
             {
                 /* Skip */
@@ -963,7 +1012,7 @@ dir_list_file_internal(parray *files, pgFile *parent, const char *parent_dir,
         */
         if (S_ISDIR(file->mode))
             dir_list_file_internal(files, file, child, exclude, follow_symlink,
-                backup_logs, skip_hidden, external_dir_num, location);
+                backup_logs, skip_hidden, external_dir_num, location, backup_replslots);
     }
 
     if (errno && errno != ENOENT)

src/bin/pg_probackup/file.cpp

@@ -51,6 +51,7 @@ typedef struct
     bool exclusive_backup;
     bool skip_hidden;
     int  external_dir_num;
+    bool backup_replslots;
 } fio_list_dir_request;
 
 typedef struct
@@ -1794,7 +1795,7 @@ cleanup:
 /* Compile the array of files located on remote machine in directory root */
 void fio_list_dir(parray *files, const char *root, bool exclude,
                             bool follow_symlink, bool add_root, bool backup_logs,
-                            bool skip_hidden, int external_dir_num)
+                            bool skip_hidden, int external_dir_num, bool backup_replslots)
 {
     fio_header hdr;
     fio_list_dir_request req;
@@ -1811,6 +1812,7 @@ void fio_list_dir(parray *files, const char *root, bool exclude,
     req.exclusive_backup = exclusive_backup;
     req.skip_hidden = skip_hidden;
     req.external_dir_num = external_dir_num;
+    req.backup_replslots = backup_replslots;
 
     hdr.cop = FIO_LIST_DIR;
     hdr.size = sizeof(req);
@@ -1870,7 +1872,14 @@ void fio_list_dir(parray *files, const char *root, bool exclude,
                 securec_check_ss_c(nRet, "\0", "\0");
             }
 
-           
+            /* 
+             * Check file that under pg_replslot and judge whether it
+             * belonged to logical replication slots for subscriptions.
+             */
+            if (backup_replslots && strcmp(buf, PG_REPLSLOT_DIR) != 0 &&
+                path_is_prefix_of_path(PG_REPLSLOT_DIR, buf) && check_logical_replslot_dir(file->rel_path) != 1) {
+                continue;
+            }
 
             parray_append(files, file);
         }
@@ -1914,7 +1923,7 @@ static void fio_list_dir_impl(int out, char* buf)
 
     dir_list_file(file_files, req->path, req->exclude, req->follow_symlink,
                         req->add_root, req->backup_logs, req->skip_hidden,
-                        req->external_dir_num, FIO_LOCAL_HOST);
+                        req->external_dir_num, FIO_LOCAL_HOST, req->backup_replslots);
 
     /* send information about files to the main process */
     for (i = 0; i < (int)parray_num(file_files); i++)

src/bin/pg_probackup/file.h

@@ -163,5 +163,7 @@ extern z_off_t fio_gzseek(gzFile f, z_off_t offset, int whence);
 extern const char* fio_gzerror(gzFile file, int *errnum);
 #endif
 
+extern char check_logical_replslot_dir(const char *rel_path);
+
 #endif
 

src/bin/pg_probackup/help.cpp

@@ -154,6 +154,7 @@ void help_pg_probackup(void)
     printf(_("                 [--remote-port=port] [--ssh-options=ssh_options]\n"));
     printf(_("                 [--remote-libpath=libpath]\n"));
     printf(_("                 [--ttl=interval] [--expire-time=time]\n"));
+    printf(_("                 [--backup-pg-replslot]\n"));
     printf(_("                 [--help]\n"));
 
     printf(_("\n  %s restore -B backup-path --instance=instance_name\n"), PROGRAM_NAME);
@@ -420,6 +421,7 @@ static void help_backup(void)
     printf(_("                 [--remote-port=port] [--ssh-options=ssh_options]\n"));
     printf(_("                 [--remote-libpath=libpath]\n"));
     printf(_("                 [--ttl=interval] [--expire-time=time]\n\n"));
+    printf(_("                 [--backup-pg-replslot]\n"));
 
     printf(_("  -B, --backup-path=backup-path    location of the backup storage area\n"));
     printf(_("      --instance=instance_name     name of the instance\n"));
@@ -441,6 +443,7 @@ static void help_backup(void)
     printf(_("      --note=text                  add note to backup\n"));
     printf(_("                                   (example: --note='backup before app update to v13.1')\n"));
     printf(_("      --archive-timeout=timeout    wait timeout for WAL segment archiving (default: 5min)\n"));
+    printf(_("      --backup-pg-replslot]        backup of '%s' directory\n"), PG_REPLSLOT_DIR);
 
     printf(_("\n  Logging options:\n"));
     printf(_("      --log-level-console=log-level-console\n"));

src/bin/pg_probackup/pg_probackup.cpp

@@ -77,6 +77,7 @@ int        rw_timeout = 0;
 
 /* backup options */
 bool         backup_logs = false;
+bool         backup_replslots = false;
 bool         smooth_checkpoint;
 char        *remote_agent;
 static char *backup_note = NULL;
@@ -186,6 +187,7 @@ static ConfigOption cmd_options[] =
     { 'b', 145, "wal",                &delete_wal,        SOURCE_CMD_STRICT },
     { 'b', 146, "expired",            &delete_expired,    SOURCE_CMD_STRICT },
     { 's', 172, "status",            &delete_status,        SOURCE_CMD_STRICT },
+    { 'b', 186, "backup-pg-replslot",   &backup_replslots,    SOURCE_CMD_STRICT},
 
     { 'b', 147, "force",            &force,                SOURCE_CMD_STRICT },
     { 'b', 148, "compress",            &compress_shortcut,    SOURCE_CMD_STRICT },
@@ -550,7 +552,7 @@ static int do_actual_operate()
                     elog(ERROR, "required parameter not specified: BACKUP_MODE "
                          "(-b, --backup-mode)");
 
-                return do_backup(start_time, set_backup_params, no_validate, no_sync, backup_logs);
+                return do_backup(start_time, set_backup_params, no_validate, no_sync, backup_logs, backup_replslots);
             }
         case RESTORE_CMD:
             return do_restore_or_validate(current.backup_id,

src/bin/pg_probackup/pg_probackupa.h

@@ -69,6 +69,7 @@ extern const char  *PROGRAM_FULL_PATH;
 #define HEADER_MAP                  "page_header_map"
 #define HEADER_MAP_TMP              "page_header_map_tmp"
 #define PG_RELATIVE_TBLSPC_DIR       "pg_location"
+#define PG_REPLSLOT_DIR              "pg_replslot"
 
 /* Timeout defaults */
 #define ARCHIVE_TIMEOUT_DEFAULT        300

src/bin/pg_probackup/pg_probackupc.h

@@ -54,6 +54,9 @@ extern bool        smooth_checkpoint;
    it will be backuped up in external dirs  */
 extern parray *pgdata_nobackup_dir;
 
+/* list of logical replication slots */
+extern parray *logical_replslot;
+
 /* remote probackup options */
 extern char* remote_agent;
 
@@ -89,7 +92,7 @@ extern const char *pgdata_exclude_dir[];
 
 /* in backup.c */
 extern int do_backup(time_t start_time, pgSetBackupParams *set_backup_params,
-                     bool no_validate, bool no_sync, bool backup_logs);
+                     bool no_validate, bool no_sync, bool backup_logs, bool backup_replslots);
 extern BackupMode parse_backup_mode(const char *value);
 extern const char *deparse_backup_mode(BackupMode mode);
 extern void process_block_change(ForkNumber forknum, const RelFileNode rnode,
@@ -239,7 +242,8 @@ extern const char* deparse_compress_alg(int alg);
 /* in dir.c */
 extern void dir_list_file(parray *files, const char *root, bool exclude,
                           bool follow_symlink, bool add_root, bool backup_logs,
-                          bool skip_hidden, int external_dir_num, fio_location location);
+                          bool skip_hidden, int external_dir_num, fio_location location,
+                          bool backup_replslots = false);
 
 extern void create_data_directories(parray *dest_files,
                                         const char *data_dir,
@@ -432,7 +436,8 @@ extern int fio_send_file(const char *from_fullpath, const char *to_fullpath, FIL
                                                         pgFile *file, char **errormsg);
 
 extern void fio_list_dir(parray *files, const char *root, bool exclude, bool follow_symlink,
-                         bool add_root, bool backup_logs, bool skip_hidden, int external_dir_num);
+                         bool add_root, bool backup_logs, bool skip_hidden, int external_dir_num,
+                         bool backup_replslots = false);
 
 extern bool pgut_rmtree(const char *path, bool rmtopdir, bool strict);
 

src/common/backend/catalog/heap.cpp

@@ -6230,7 +6230,7 @@ Datum GetPartBoundaryByTuple(Relation rel, HeapTuple tuple)
     return Timestamp2Boundarys(rel, Align2UpBoundary(value, partMap->intervalValue, boundaryTs));
 }
 
-Oid AddNewIntervalPartition(Relation rel, void* insertTuple)
+Oid AddNewIntervalPartition(Relation rel, void* insertTuple, bool isDDL)
 {
     Relation pgPartRel = NULL;
     Oid newPartOid = InvalidOid;
@@ -6327,7 +6327,13 @@ Oid AddNewIntervalPartition(Relation rel, void* insertTuple)
      */
     CommandCounterIncrement();
 
-    UpdatePgObjectChangecsn(RelationGetRelid(rel), rel->rd_rel->relkind);
+    /*
+     * If add interval partition in the DDL, do not need to change the csn
+     * because the scn has been changed in the DDL.
+     */
+    if (!isDDL) {
+        UpdatePgObjectChangecsn(RelationGetRelid(rel), rel->rd_rel->relkind);
+    }
 
     return newPartOid;
 }
@@ -7113,7 +7119,7 @@ int lookupHBucketid(oidvector *buckets, int low, int2 bktId)
  * Description	:
  * Notes		:
  */
-Oid heapTupleGetPartitionId(Relation rel, void *tuple)
+Oid heapTupleGetPartitionId(Relation rel, void *tuple, bool isDDL)
 {
     Oid partitionid = InvalidOid;
 
@@ -7140,7 +7146,7 @@ Oid heapTupleGetPartitionId(Relation rel, void *tuple)
                 (errcode(ERRCODE_NO_DATA_FOUND), errmsg("inserted partition key does not map to any table partition")));
         } break;
         case PART_AREA_INTERVAL: {
-            return AddNewIntervalPartition(rel, tuple);
+            return AddNewIntervalPartition(rel, tuple, isDDL);
         } break;
         case PART_AREA_LIST: {
             ereport(ERROR,

src/common/backend/catalog/pg_builtin_proc.cpp

@@ -52,7 +52,7 @@ static_assert(sizeof(false) == sizeof(char), "illegal bool size");
 static struct HTAB* nameHash = NULL;
 static struct HTAB* oidHash = NULL;
 
-/* for b_sql_plugin */
+/* for dolphin */
 struct HTAB* b_nameHash = NULL;
 struct HTAB* b_oidHash = NULL;
 
@@ -118,7 +118,7 @@ static const FuncGroup* NameHashTableAccess(HASHACTION action, const char* name,
 
     Assert(name != NULL);
 
-    if (DB_IS_CMPT(B_FORMAT) && b_nameHash != NULL && u_sess->attr.attr_sql.b_sql_plugin) {
+    if (DB_IS_CMPT(B_FORMAT) && b_nameHash != NULL && u_sess->attr.attr_sql.dolphin) {
         result = (HashEntryNameToFuncGroup *)hash_search(b_nameHash, &temp_name, action, &found);
     } else {
         result = (HashEntryNameToFuncGroup *)hash_search(nameHash, &temp_name, action, &found);
@@ -144,7 +144,7 @@ static const Builtin_func* OidHashTableAccess(HASHACTION action, Oid oid, const
     bool found = false;
     Assert(oid > 0);
 
-    if (DB_IS_CMPT(B_FORMAT) && b_oidHash != NULL && u_sess->attr.attr_sql.b_sql_plugin) {
+    if (DB_IS_CMPT(B_FORMAT) && b_oidHash != NULL && u_sess->attr.attr_sql.dolphin) {
         result = (HashEntryOidToBuiltinFunc *)hash_search(b_oidHash, &oid, action, &found);
     } else {
         result = (HashEntryOidToBuiltinFunc *)hash_search(oidHash, &oid, action, &found);

src/common/backend/catalog/pg_subscription.cpp

@@ -28,7 +28,6 @@
 #include "utils/builtins.h"
 #include "utils/fmgroids.h"
 #include "utils/syscache.h"
-#include "replication/worker_internal.h"
 
 static List *textarray_to_stringlist(ArrayType *textarray);
 
@@ -91,6 +90,13 @@ Subscription *GetSubscription(Oid subid, bool missing_ok)
     }
     sub->publications = textarray_to_stringlist(DatumGetArrayTypeP(datum));
 
+    datum = SysCacheGetAttr(SUBSCRIPTIONOID, tup, Anum_pg_subscription_subbinary, &isnull);
+    if (unlikely(isnull)) {
+        sub->binary = false;
+    } else {
+        sub->binary = DatumGetBool(datum);
+    }
+
     ReleaseSysCache(tup);
 
     return sub;
@@ -183,7 +189,7 @@ char *get_subscription_name(Oid subid, bool missing_ok)
 }
 
 /* Clear the list content, only deal with DefElem and string content */
-static void ClearListContent(List *list)
+void ClearListContent(List *list)
 {
     ListCell *cell = NULL;
     foreach(cell, list) {
@@ -203,25 +209,6 @@ static void ClearListContent(List *list)
     }
 }
 
-/*
- * Decrypt conninfo for subscription.
- * IMPORTANT: caller should clear and free the memory after using it immediately
- */
-char *DecryptConninfo(char *encryptConninfo)
-{
-    const char* sensitiveOptionsArray[] = {"password"};
-    const int sensitiveArrayLength = lengthof(sensitiveOptionsArray);
-    List *defList = ConninfoToDefList(encryptConninfo);
-    DecryptOptions(defList, sensitiveOptionsArray, sensitiveArrayLength, SUBSCRIPTION_MODE);
-    char *decryptConninfo = DefListToString(defList);
-
-    /* defList has plain content, clear it before free */
-    ClearListContent(defList);
-    list_free_ext(defList);
-    /* IMPORTANT: caller should clear and free the memory after using it immediately */
-    return decryptConninfo;
-}
-
 /*
  * Convert text array to list of strings.
  *

src/common/backend/nodes/readfuncs.cpp

@@ -70,6 +70,7 @@
 
 THR_LOCAL bool skip_read_extern_fields = false;
 
+#define IS_DATANODE_BUT_NOT_SINGLENODE (IS_PGXC_DATANODE && !IS_SINGLE_NODE)
 /*
  * Macros to simplify reading of different kinds of fields.  Use these
  * wherever possible to reduce the chance for silly typos.	Note that these
@@ -401,24 +402,27 @@ THR_LOCAL bool skip_read_extern_fields = false;
     token = pg_strtok(&length); /* skip :fldname */ \
     local_node->fldname = _readBitmapset()
 
-#define READ_TYPEINFO_FIELD(fldname)                                                                          \
-    do {                                                                                                      \
-        if (local_node->fldname >= FirstBootstrapObjectId) {                                                  \
-            IF_EXIST(exprtypename)                                                                            \
-            {                                                                                                 \
-                char* exprtypename = NULL;                                                                    \
-                char* exprtypenamespace = NULL;                                                               \
-                token = pg_strtok(&length);                                                                   \
-                token = pg_strtok(&length);                                                                   \
-                exprtypename = nullable_string(token, length);                                                \
-                token = pg_strtok(&length);                                                                   \
-                token = pg_strtok(&length);                                                                   \
-                exprtypenamespace = nullable_string(token, length);                                           \
-                local_node->fldname = get_typeoid(get_namespace_oid(exprtypenamespace, false), exprtypename); \
-                pfree_ext(exprtypename);                                                                      \
-                pfree_ext(exprtypenamespace);                                                                 \
-            }                                                                                                 \
-        }                                                                                                     \
+#define READ_TYPEINFO_FIELD(fldname)                                                                              \
+    do {                                                                                                          \
+        if (local_node->fldname >= FirstBootstrapObjectId) {                                                      \
+            IF_EXIST(exprtypename)                                                                                \
+            {                                                                                                     \
+                char* exprtypename = NULL;                                                                        \
+                char* exprtypenamespace = NULL;                                                                   \
+                token = pg_strtok(&length);                                                                       \
+                token = pg_strtok(&length);                                                                       \
+                exprtypename = nullable_string(token, length);                                                    \
+                token = pg_strtok(&length);                                                                       \
+                token = pg_strtok(&length);                                                                       \
+                exprtypenamespace = nullable_string(token, length);                                               \
+                /* No need to reset field on CN or singlenode, keep pg_strtok() for forward compatibility */      \
+                if (IS_DATANODE_BUT_NOT_SINGLENODE) {                                                             \
+                    local_node->fldname = get_typeoid(get_namespace_oid(exprtypenamespace, false), exprtypename); \
+                }                                                                                                 \
+                pfree_ext(exprtypename);                                                                          \
+                pfree_ext(exprtypenamespace);                                                                     \
+            }                                                                                                     \
+        }                                                                                                         \
     } while (0)
 
 #define READ_TYPEINFO(typePtr)                                                                \
@@ -493,9 +497,30 @@ THR_LOCAL bool skip_read_extern_fields = false;
                 token = pg_strtok(&length);                                                                 \
                 token = pg_strtok(&length);                                                                 \
                 funcnamespace = nullable_string(token, length);                                             \
-                if (IS_PGXC_DATANODE && !skip_read_extern_fields) {                                         \
-                    local_node->fldname =                                                                   \
-                        get_func_oid(funcname, get_namespace_oid(funcnamespace, false), (Expr*)local_node); \
+                bool notfound = false;                                                                      \
+                if (IS_DATANODE_BUT_NOT_SINGLENODE && !skip_read_extern_fields) {                           \
+                    Oid funcoid = InvalidOid;                                                               \
+                    do {                                                                                    \
+                        Oid nspid = get_namespace_oid(funcnamespace, true);                                 \
+                        if (!OidIsValid(nspid)) {                                                           \
+                            notfound = true;                                                                \
+                            break;                                                                          \
+                        }                                                                                   \
+                        funcoid = get_func_oid(funcname, nspid, (Expr*)local_node);                         \
+                    } while (0);                                                                            \
+                    if (notfound || !OidIsValid(funcoid)) {                                                 \
+                        ereport(ERROR,                                                                      \
+                            (errmodule(MOD_OPT), errcode(ERRCODE_UNDEFINED_OBJECT),                         \
+                                errmsg("Cannot identify function %s.%s while deserializing field.",         \
+                                    funcname, funcnamespace),                                               \
+                                errdetail("Function with oid %u or its namespace may be renamed",           \
+                                    local_node->fldname),                                                   \
+                                errhint("Please rebuild column defalt expression, views etc. that are"      \
+                                    " related to this renamed object."),                                    \
+                                errcause("Object renamed after recorded as nodetree."),                     \
+                                erraction("Rebuild relevant object.")));                                    \
+                    }                                                                                       \
+                    local_node->fldname = funcoid;                                                          \
                 }                                                                                           \
                 pfree_ext(funcname);                                                                        \
                 pfree_ext(funcnamespace);                                                                   \
@@ -525,7 +550,7 @@ THR_LOCAL bool skip_read_extern_fields = false;
             token = pg_strtok(&length);                                                         \
             token = pg_strtok(&length);                                                         \
             oprrightname = nullable_string(token, length);                                      \
-            if (IS_PGXC_DATANODE) {                                                             \
+            if (IS_DATANODE_BUT_NOT_SINGLENODE) {                                               \
                 namespaceId = get_namespace_oid(opnamespace, false);                            \
                 oprleft = get_typeoid(namespaceId, oprleftname);                                \
                 oprright = oprleft;                                                             \
@@ -568,7 +593,7 @@ THR_LOCAL bool skip_read_extern_fields = false;
                 token = pg_strtok(&length);                                                            \
                 token = pg_strtok(&length);                                                            \
                 oprrightname = nullable_string(token, length);                                         \
-                if (IS_PGXC_DATANODE) {                                                                \
+                if (IS_DATANODE_BUT_NOT_SINGLENODE) {                                                  \
                     namespaceId = get_namespace_oid(opnamespace, false);                               \
                     oprleft = get_typeoid(namespaceId, oprleftname);                                   \
                     oprright = oprleft;                                                                \
@@ -2126,14 +2151,21 @@ static FuncExpr* _readFuncExpr(void)
             ereport(ERROR, (errcode(ERRCODE_UNEXPECTED_NULL_VALUE), errmsg("NULL seqNamespace for nextval()")));
         }
 
-        if (!IS_PGXC_COORDINATOR && !skip_read_extern_fields) {
-            Oid seqid = get_valid_relname_relid(seqNamespace, seqName);
+        if (IS_DATANODE_BUT_NOT_SINGLENODE && !skip_read_extern_fields) {
 
+            Oid seqid = get_valid_relname_relid(seqNamespace, seqName, true);
+            Const* firstArg = (Const*)linitial(local_node->args);
             if (OidIsValid(seqid)) {
-                Const* firstArg = (Const*)linitial(local_node->args);
                 if (firstArg != NULL) {
                     firstArg->constvalue = ObjectIdGetDatum(seqid);
                 }
+            } else {
+                ereport(ERROR, (errmodule(MOD_OPT), errcode(ERRCODE_UNDEFINED_OBJECT),
+                    errmsg("Cannot identify sequence %s.%s while deserializing field.", seqNamespace, seqName),
+                    errdetail("Sequence with oid %u or its namespace may be renamed",
+                        DatumGetObjectId(firstArg->constvalue)),
+                    errhint("Please rebuild column defalt expression, views etc. that are related to this sequence"),
+                    errcause("Object renamed after recorded as nodetree."), erraction("Rebuild relevant object.")));
             }
         }
         pfree_ext(seqName);

src/common/backend/parser/parse_func.cpp

@@ -1454,7 +1454,14 @@ FuncCandidateList sort_candidate_func_list(FuncCandidateList oldCandidates)
         }
         candidates[smallestIndex] = NULL;
     }
-
+    
+    for (int i = 0; i < size; i++) {
+        if (candidates[i] != NULL) {
+            lastCandidate->next = candidates[i];
+            lastCandidate = lastCandidate->next;
+        }
+    }
+    lastCandidate->next = NULL;
     pfree(candidates);
     return sortedCandidates;
 }

src/common/backend/utils/cache/knl_globalbasedefcache.cpp

@@ -61,7 +61,7 @@ void GlobalBaseDefCache::RemoveElemFromBucket(GlobalBaseEntry *base)
     if (is_relation) {
         GlobalRelationEntry *entry = (GlobalRelationEntry *)base;
         uint64 rel_size = GetRelEstimateSize(entry);
-        pg_atomic_fetch_sub_u64(&m_base_space, rel_size);
+        pg_atomic_fetch_sub_u64(&m_base_space, AllocSetContextUsedSpace(((AllocSet)entry->rel_mem_manager)));
         m_db_entry->MemoryEstimateSub(rel_size);
     } else {
         GlobalPartitionEntry *entry = (GlobalPartitionEntry *)base;
@@ -77,7 +77,7 @@ void GlobalBaseDefCache::AddHeadToBucket(Index hash_index, GlobalBaseEntry *base
     if (is_relation) {
         GlobalRelationEntry *entry = (GlobalRelationEntry *)base;
         uint64 rel_size = GetRelEstimateSize(entry);
-        pg_atomic_fetch_add_u64(&m_base_space, rel_size);
+        pg_atomic_fetch_add_u64(&m_base_space, AllocSetContextUsedSpace(((AllocSet)entry->rel_mem_manager)));
         m_db_entry->MemoryEstimateAdd(rel_size);
     } else {
         GlobalPartitionEntry *entry = (GlobalPartitionEntry *)base;
@@ -400,4 +400,4 @@ GlobalBaseDefCache::GlobalBaseDefCache(Oid db_oid, bool is_shared, GlobalSysDBCa
     m_base_space = 0;
     m_obj_locks = NULL;
     m_db_entry = entry;
-}
+}

src/common/backend/utils/cache/knl_globalsysdbcache.cpp

@@ -659,7 +659,27 @@ void GlobalSysDBCache::InitSysCacheRelIds()
  */
 void GlobalSysDBCache::RefreshHotStandby()
 {
+    if (!EnableGlobalSysCache()) {
+	    return;
+    }
     hot_standby = (t_thrd.postmaster_cxt.HaShmData->current_mode != STANDBY_MODE || XLogStandbyInfoActive());
+    if (hot_standby || !m_is_inited) {
+	    return;
+    }
+    /* clean all */
+    for (int hash_index = 0; hash_index < m_nbuckets; hash_index ++) {
+        PthreadRWlockRdlock(LOCAL_SYSDB_RESOWNER, &m_db_locks[hash_index]);
+	for (Dlelem * elt = DLGetTail(m_bucket_list.GetBucket(hash_index)); elt != NULL;) {
+            GlobalSysDBCacheEntry *entry = (GlobalSysDBCacheEntry *)DLE_VAL(elt);
+	    elt = DLGetPred(elt);
+	    entry->ResetDBCache<true>();
+	}
+        PthreadRWlockUnlock(LOCAL_SYSDB_RESOWNER, &m_db_locks[hash_index]);
+    }
+    if (m_global_shared_db_entry != NULL) {
+        m_global_shared_db_entry->ResetDBCache<true>();
+    }
+
 }
 
 void GlobalSysDBCache::Init(MemoryContext parent)
@@ -1272,9 +1292,18 @@ int ResizeHashBucket(int origin_nbucket, DynamicHashBucketStrategy strategy)
     return cc_nbuckets;
 }
 
+void NotifyGscRecoveryStarted()
+{
+    if (!EnableGlobalSysCache()) {
+	    return;
+    }
+    g_instance.global_sysdbcache.recovery_finished = false;
+
+}
+
 void NotifyGscRecoveryFinished()
 {
     if (EnableGlobalSysCache()) {
         g_instance.global_sysdbcache.recovery_finished = true;
     }
-}
+}

src/common/backend/utils/cache/knl_globalsystabcache.cpp

@@ -183,7 +183,7 @@ void GlobalSysTabCache::InvalidTuples(int cache_id, uint32 hash_value, bool rese
 
     /* maybe upgrade from version before v5r2c00, the cacheid is out of order
      * whatever, we cache nothing except relmap, so just ignore the catcache invalmsg */
-    if (unlikely(!g_instance.global_sysdbcache.recovery_finished) && m_global_systupcaches[cache_id] == NULL) {
+    if (unlikely(!g_instance.global_sysdbcache.recovery_finished && m_global_systupcaches[cache_id] == NULL)) {
         return;
     }
 

src/common/backend/utils/cache/knl_localpartdefcache.cpp

@@ -74,7 +74,7 @@ Partition LocalPartDefCache::SearchPartitionFromGlobalCopy(Oid part_oid)
     if (!g_instance.global_sysdbcache.hot_standby) {
         return NULL;
     }
-    if (unlikely(!g_instance.global_sysdbcache.recovery_finished)) {
+    if (unlikely(!IsPrimaryRecoveryFinished())) {
         return NULL;
     }
     uint32 hash_value = oid_hash((void *)&(part_oid), sizeof(Oid));
@@ -165,7 +165,7 @@ static bool IsPartOidStoreInGlobal(Oid part_oid)
     if (!g_instance.global_sysdbcache.hot_standby) {
         return false;
     }
-    if (unlikely(!g_instance.global_sysdbcache.recovery_finished)) {
+    if (unlikely(!IsPrimaryRecoveryFinished())) {
         return false;
     }
     if (g_instance.global_sysdbcache.StopInsertGSC()) {
@@ -456,4 +456,4 @@ Partition LocalPartDefCache::PartitionIdGetPartition(Oid part_oid, StorageType s
     }
 
     return pd;
-}
+}

src/common/backend/utils/cache/knl_localsystupcache.cpp

@@ -433,7 +433,7 @@ LocalCatCTup *LocalSysTupCache::SearchTupleFromGlobal(Datum *arguments, uint32 h
     bool bypass_gsc = HistoricSnapshotActive() ||
         m_global_systupcache->enable_rls ||
         !g_instance.global_sysdbcache.hot_standby ||
-        unlikely(!g_instance.global_sysdbcache.recovery_finished);
+        unlikely(!IsPrimaryRecoveryFinished());
     if (invalid_entries.ExistTuple(hash_value) || bypass_gsc) {
         global_ct = m_global_systupcache->SearchTupleFromFile(hash_value, arguments, true);
     } else {
@@ -585,7 +585,7 @@ LocalCatCList *LocalSysTupCache::SearchListFromGlobal(int nkeys, Datum *argument
     bool bypass_gsc = HistoricSnapshotActive() ||
         m_global_systupcache->enable_rls ||
         !g_instance.global_sysdbcache.hot_standby ||
-        unlikely(!g_instance.global_sysdbcache.recovery_finished);
+        unlikely(!IsPrimaryRecoveryFinished());
     GlobalCatCList *global_cl;
     if (invalid_entries.ExistList() || bypass_gsc) {
         global_cl = m_global_systupcache->SearchListFromFile(hash_value, nkeys, arguments, true);
@@ -703,7 +703,7 @@ LocalCatCTup *LocalSysTupCache::SearchTupleFromGlobalForProcAllArgs(
     bool bypass_gsc = HistoricSnapshotActive() ||
         m_global_systupcache->enable_rls ||
         !g_instance.global_sysdbcache.hot_standby ||
-        unlikely(!g_instance.global_sysdbcache.recovery_finished);
+        unlikely(!IsPrimaryRecoveryFinished());
     if (invalid_entries.ExistTuple(hash_value) || bypass_gsc) {
         global_ct = m_global_systupcache->SearchTupleFromFileWithArgModes(hash_value, arguments, argModes, true);
     } else {

src/common/backend/utils/cache/knl_localtabdefcache.cpp

@@ -93,7 +93,7 @@ Relation LocalTabDefCache::SearchRelationFromGlobalCopy(Oid rel_oid)
     if (!g_instance.global_sysdbcache.hot_standby) {
         return NULL;
     }
-    if (unlikely(!g_instance.global_sysdbcache.recovery_finished)) {
+    if (unlikely(!IsPrimaryRecoveryFinished())) {
         return NULL;
     }
     uint32 hash_value = oid_hash((void *)&(rel_oid), sizeof(Oid));
@@ -190,7 +190,7 @@ static bool IsRelOidStoreInGlobal(Oid rel_oid)
     if (!g_instance.global_sysdbcache.hot_standby) {
         return false;
     }
-    if (unlikely(!g_instance.global_sysdbcache.recovery_finished)) {
+    if (unlikely(!IsPrimaryRecoveryFinished())) {
         return false;
     }
     if (g_instance.global_sysdbcache.StopInsertGSC()) {
@@ -1137,4 +1137,4 @@ void LocalTabDefCache::ResetInitFlag()
     m_is_inited_phase3 = false;
 
     m_db_id = InvalidOid;
-}
+}

src/common/backend/utils/cache/lsyscache.cpp

@@ -1723,7 +1723,7 @@ char* get_relname_relid_extend(
 extern bool StreamTopConsumerAmI();
 
 /* same as get_relname_relid except we check for cache invalidation here */
-Oid get_valid_relname_relid(const char* relnamespace, const char* relname)
+Oid get_valid_relname_relid(const char* relnamespace, const char* relname, bool nsp_missing_ok)
 {
     Oid nspid = InvalidOid;
     Oid oldnspid = InvalidOid;
@@ -1747,7 +1747,10 @@ Oid get_valid_relname_relid(const char* relnamespace, const char* relname)
         if (EnableLocalSysCache()) {
             thrd_inval_count = t_thrd.lsc_cxt.lsc->inval_cxt.SIMCounter;
         }
-        nspid = get_namespace_oid(relnamespace, false);
+        nspid = get_namespace_oid(relnamespace, nsp_missing_ok);
+        if (!OidIsValid(nspid)) {
+            return InvalidOid;
+        }
         relid = get_relname_relid(relname, nspid);
         /*
          * In bootstrap processing mode, we don't bother with locking

src/common/backend/utils/init/globals.cpp

@@ -59,7 +59,7 @@ bool open_join_children = true;
 bool will_shutdown = false;
 
 /* hard-wired binary version number */
-const uint32 GRAND_VERSION_NUM = 92605;
+const uint32 GRAND_VERSION_NUM = 92606;
 
 const uint32 PREDPUSH_SAME_LEVEL_VERSION_NUM = 92522;
 const uint32 UPSERT_WHERE_VERSION_NUM = 92514;
@@ -101,6 +101,7 @@ const uint32 PRIVS_DIRECTORY_VERSION_NUM = 92460;
 const uint32 COMMENT_RECORD_PARAM_VERSION_NUM = 92484;
 const uint32 SCAN_BATCH_MODE_VERSION_NUM = 92568;
 const uint32 PUBLICATION_VERSION_NUM = 92580;
+const uint32 SUBSCRIPTION_BINARY_VERSION_NUM = 92606;
 
 /* Version number of the guc parameter backend_version added in V500R001C20 */
 const uint32 V5R1C20_BACKEND_VERSION_NUM = 92305;

src/common/backend/utils/init/postinit.cpp

@@ -2712,8 +2712,8 @@ void PostgresInitializer::InitExtensionVariable()
     }
     
     /* check whether the extension has been created */
-    const char* b_sql_plugin = "b_sql_plugin";
-    u_sess->attr.attr_sql.b_sql_plugin = CheckIfExtensionExists(b_sql_plugin);
+    const char* dolphin = "dolphin";
+    u_sess->attr.attr_sql.dolphin = CheckIfExtensionExists(dolphin);
 }
 
 void PostgresInitializer::FinishInit()

src/common/pl/plpgsql/src/gram.y

@@ -10340,7 +10340,16 @@ check_sql_expr(const char *stmt, int location, int leaderlen)
 
     oldCxt = MemoryContextSwitchTo(u_sess->plsql_cxt.curr_compile_context->compile_tmp_cxt);
     u_sess->plsql_cxt.plpgsql_yylloc = plpgsql_yylloc;
-    (void) raw_parser(stmt);
+    RawParserHook parser_hook= raw_parser;
+#ifndef ENABLE_MULTIPLE_NODES
+    if (u_sess->attr.attr_sql.dolphin) {
+        int id = GetCustomParserId();
+        if (id >= 0 && g_instance.raw_parser_hook[id] != NULL) {
+            parser_hook = (RawParserHook)g_instance.raw_parser_hook[id];
+        }
+    }
+#endif
+    (void)parser_hook(stmt, NULL);
     MemoryContextSwitchTo(oldCxt);
 
     /* Restore former ereport callback */

src/gausskernel/cbb/instruments/percentile/percentile.cpp

@@ -222,6 +222,7 @@ NON_EXEC_STATIC void PercentileMain()
         g_instance.stat_cxt.force_process = false;
         sleep(SLEEP_INTERVAL);
     }
+    elog(LOG, "instrumention percentile ended");
     gs_thread_exit(0);
 }
 

src/gausskernel/dbmind/tools/components/index_advisor/index_advisor_workload.py

@@ -26,9 +26,11 @@ import logging
 try:
     from .dao.gsql_execute import GSqlExecute
     from .dao.execute_factory import ExecuteFactory
+    from .mcts import MCTS
 except ImportError:
     from dao.gsql_execute import GSqlExecute
     from dao.execute_factory import ExecuteFactory
+    from mcts import MCTS
 
 ENABLE_MULTI_NODE = False
 SAMPLE_NUM = 5
@@ -192,9 +194,12 @@ class IndexAdvisor:
                                                     self.workload_used_index))
         if DRIVER:
             self.db.close_conn()
-
-        opt_config = greedy_determine_opt_config(self.workload_info[0], atomic_config_total,
-                                                 candidate_indexes, self.index_cost_total[0])
+        if MAX_INDEX_STORAGE:
+            opt_config = MCTS(self.workload_info[0], atomic_config_total, candidate_indexes,
+                              MAX_INDEX_STORAGE, MAX_INDEX_NUM)
+        else:
+            opt_config = greedy_determine_opt_config(self.workload_info[0], atomic_config_total,
+                                                     candidate_indexes, self.index_cost_total[0])
         self.retain_lower_cost_index(candidate_indexes)
         if len(opt_config) == 0:
             print("No optimal indexes generated!")
@@ -943,7 +948,7 @@ def check_parameter(args):
         raise argparse.ArgumentTypeError("%s is an invalid positive int value" %
                                          args.max_index_num)
     if args.max_index_storage is not None and args.max_index_storage <= 0:
-        raise argparse.ArgumentTypeError("%s is an invalid positive int value" %
+        raise argparse.ArgumentTypeError("%s is an invalid positive float value" %
                                          args.max_index_storage)
     JSON_TYPE = args.json
     MAX_INDEX_NUM = args.max_index_num
@@ -971,7 +976,7 @@ def main(argv):
     arg_parser.add_argument(
         "--max_index_num", help="Maximum number of suggested indexes", type=int)
     arg_parser.add_argument("--max_index_storage",
-                            help="Maximum storage of suggested indexes/MB", type=int)
+                            help="Maximum storage of suggested indexes/MB", type=float)
     arg_parser.add_argument("--multi_iter_mode", action='store_true',
                             help="Whether to use multi-iteration algorithm", default=False)
     arg_parser.add_argument("--multi_node", action='store_true',

src/gausskernel/dbmind/tools/components/index_advisor/mcts.py

@@ -0,0 +1,397 @@
+import sys
+import math
+import random
+import copy
+
+STORAGE_THRESHOLD = 0
+AVAILABLE_CHOICES = None
+ATOMIC_CHOICES = None
+WORKLOAD_INFO = None
+MAX_INDEX_NUM = 0
+
+
+def is_same_index(index, compared_index):
+    return index.table == compared_index.table and \
+           index.columns == compared_index.columns and \
+           index.index_type == compared_index.index_type
+
+
+def atomic_config_is_valid(atomic_config, config):
+    # if candidate indexes contains all atomic index of current config1, then record it
+    for atomic_index in atomic_config:
+        is_exist = False
+        for index in config:
+            if is_same_index(index, atomic_index):
+                index.storage = atomic_index.storage
+                is_exist = True
+                break
+        if not is_exist:
+            return False
+    return True
+
+
+def find_subsets_num(choice):
+    atomic_subsets_num = []
+    for pos, atomic in enumerate(ATOMIC_CHOICES):
+        if not atomic or len(atomic) > len(choice):
+            continue
+        # find valid atomic index
+        if atomic_config_is_valid(atomic, choice):
+            atomic_subsets_num.append(pos)
+        # find the same atomic index as the candidate index
+        if len(atomic) == 1 and (is_same_index(choice[-1], atomic[0])):
+            choice[-1].atomic_pos = pos
+    return atomic_subsets_num
+
+
+def find_best_benefit(choice):
+    atomic_subsets_num = find_subsets_num(choice)
+    total_benefit = 0
+    for ind, obj in enumerate(WORKLOAD_INFO):
+        # calculate the best benefit for the current sql
+        max_benefit = 0
+        for pos in atomic_subsets_num:
+            if (obj.cost_list[0] - obj.cost_list[pos]) > max_benefit:
+                max_benefit = obj.cost_list[0] - obj.cost_list[pos]
+        total_benefit += max_benefit
+    return total_benefit
+
+
+def get_diff(available_choices, choices):
+    except_choices = copy.copy(available_choices)
+    for i in available_choices:
+        for j in choices:
+            if is_same_index(i, j):
+                except_choices.remove(i)
+    return except_choices
+
+
+class State(object):
+    """
+    The game state of the Monte Carlo tree search,
+    the state data recorded under a certain Node node,
+    including the current game score, the current number of game rounds,
+    and the execution record from the beginning to the current.
+
+    It is necessary to realize whether the current state has reached the end of the game state,
+    and support the operation of randomly fetching from the Action collection.
+    """
+
+    def __init__(self):
+        self.current_storage = 0.0
+        self.current_benefit = 0.0
+        # record the sum of choices up to the current state
+        self.accumulation_choices = []
+        # record available choices of current state
+        self.available_choices = []
+        self.displayable_choices = []
+
+    def get_available_choices(self):
+        return self.available_choices
+
+    def set_available_choices(self, choices):
+        self.available_choices = choices
+
+    def get_current_storage(self):
+        return self.current_storage
+
+    def set_current_storage(self, value):
+        self.current_storage = value
+
+    def get_current_benefit(self):
+        return self.current_benefit
+
+    def set_current_benefit(self, value):
+        self.current_benefit = value
+
+    def get_accumulation_choices(self):
+        return self.accumulation_choices
+
+    def set_accumulation_choices(self, choices):
+        self.accumulation_choices = choices
+
+    def is_terminal(self):
+        # the current node is a leaf node
+        return len(self.accumulation_choices) == MAX_INDEX_NUM
+
+    def compute_benefit(self):
+        return self.current_benefit
+
+    def get_next_state_with_random_choice(self):
+        # ensure that the choices taken are not repeated
+        if not self.available_choices:
+            return None
+        random_choice = random.choice([choice for choice in self.available_choices])
+        self.available_choices.remove(random_choice)
+        choice = copy.copy(self.accumulation_choices)
+        choice.append(random_choice)
+        benefit = find_best_benefit(choice)
+        # if current choice not satisfy restrictions, then continue get next choice
+        if benefit <= self.current_benefit or \
+                self.current_storage + random_choice.storage > STORAGE_THRESHOLD:
+            return self.get_next_state_with_random_choice()
+
+        next_state = State()
+        # initialize the properties of the new state
+        next_state.set_accumulation_choices(choice)
+        next_state.set_current_benefit(benefit)
+        next_state.set_current_storage(self.current_storage + random_choice.storage)
+        next_state.set_available_choices(get_diff(AVAILABLE_CHOICES, choice))
+        return next_state
+
+    def __repr__(self):
+        self.displayable_choices = ['{}: {}'.format(choice.table, choice.columns)
+                                    for choice in self.accumulation_choices]
+        return "reward: {}, storage :{}, choices: {}".format(
+            self.current_benefit, self.current_storage, self.displayable_choices)
+
+
+class Node(object):
+    """
+    The Node of the Monte Carlo tree search tree contains the parent node and
+     current point information,
+    which is used to calculate the traversal times and quality value of the UCB,
+    and the State of the Node selected by the game.
+    """
+    def __init__(self):
+        self.visit_number = 0
+        self.quality = 0.0
+
+        self.parent = None
+        self.children = []
+        self.state = None
+
+    def get_parent(self):
+        return self.parent
+
+    def set_parent(self, parent):
+        self.parent = parent
+
+    def get_children(self):
+        return self.children
+
+    def expand_child(self, node):
+        node.set_parent(self)
+        self.children.append(node)
+
+    def set_state(self, state):
+        self.state = state
+
+    def get_state(self):
+        return self.state
+
+    def get_visit_number(self):
+        return self.visit_number
+
+    def set_visit_number(self, number):
+        self.visit_number = number
+
+    def update_visit_number(self):
+        self.visit_number += 1
+
+    def get_quality_value(self):
+        return self.quality
+
+    def set_quality_value(self, value):
+        self.quality = value
+
+    def update_quality_value(self, reward):
+        self.quality += reward
+
+    def is_all_expand(self):
+        return len(self.children) == \
+               len(AVAILABLE_CHOICES) - len(self.get_state().get_accumulation_choices())
+
+    def __repr__(self):
+        return "Node: {}, Q/N: {}/{}, State: {}".format(
+            hash(self), self.quality, self.visit_number, self.state)
+
+
+def tree_policy(node):
+    """
+    In the Selection and Expansion stages of Monte Carlo tree search,
+    the node that needs to be searched (such as the root node) is passed in,
+    and the best node that needs to be expanded is returned
+    according to the exploration/exploitation algorithm.
+    Note that if the node is a leaf node, it will be returned directly.
+
+   The basic strategy is to first find the child nodes that have not been selected at present,
+   and select them randomly if there are more than one. If both are selected,
+   find the one with the largest UCB value that has weighed exploration/exploitation,
+   and randomly select if the UCB values are equal.
+    """
+
+    # check if the current node is leaf node
+    while node and not node.get_state().is_terminal():
+
+        if node.is_all_expand():
+            node = best_child(node, True)
+        else:
+            # return the new sub node
+            sub_node = expand(node)
+            # when there is no node that satisfies the condition in the remaining nodes,
+            # this state is empty
+            if sub_node.get_state():
+                return sub_node
+
+    # return the leaf node
+    return node
+
+
+def default_policy(node):
+    """
+    In the Simulation stage of Monte Carlo tree search, input a node that needs to be expanded,
+    create a new node after random operation, and return the reward of the new node.
+    Note that the input node should not be a child node,
+    and there are unexecuted Actions that can be expendable.
+
+    The basic strategy is to choose the Action at random.
+    """
+
+    # get the state of the game
+    current_state = copy.deepcopy(node.get_state())
+
+    # run until the game over
+    while not current_state.is_terminal():
+        # pick one random action to play and get next state
+        next_state = current_state.get_next_state_with_random_choice()
+        if not next_state:
+            break
+        current_state = next_state
+
+    final_state_reward = current_state.compute_benefit()
+    return final_state_reward
+
+
+def expand(node):
+    """
+    Enter a node, expand a new node on the node, use the random method to execute the Action,
+    and return the new node. Note that it is necessary to ensure that the newly
+     added nodes are different from other node Action
+    """
+
+    new_state = node.get_state().get_next_state_with_random_choice()
+    sub_node = Node()
+    sub_node.set_state(new_state)
+    node.expand_child(sub_node)
+
+    return sub_node
+
+
+def best_child(node, is_exploration):
+    """
+    Using the UCB algorithm,
+    select the child node with the highest score after weighing the exploration and exploitation.
+    Note that if it is the prediction stage,
+    the current Q-value score with the highest score is directly selected.
+    """
+
+    best_score = -sys.maxsize
+    best_sub_node = None
+
+    # travel all sub nodes to find the best one
+    for sub_node in node.get_children():
+        # The children nodes of the node contains the children node whose state is empty,
+        # this kind of node comes from the node that does not meet the conditions.
+        if not sub_node.get_state():
+            continue
+        # ignore exploration for inference
+        if is_exploration:
+            C = 1 / math.sqrt(2.0)
+        else:
+            C = 0.0
+
+        # UCB = quality / times + C * sqrt(2 * ln(total_times) / times)
+        left = sub_node.get_quality_value() / sub_node.get_visit_number()
+        right = 2.0 * math.log(node.get_visit_number()) / sub_node.get_visit_number()
+        score = left + C * math.sqrt(right)
+        # get the maximum score, while filtering nodes that do not meet the space constraints and
+        # nodes that have no revenue
+        if score > best_score \
+                and sub_node.get_state().get_current_storage() <= STORAGE_THRESHOLD \
+                and sub_node.get_state().get_current_benefit() > 0:
+            best_sub_node = sub_node
+            best_score = score
+
+    return best_sub_node
+
+
+def backpropagate(node, reward):
+    """
+    In the Backpropagation stage of Monte Carlo tree search,
+    input the node that needs to be expended and the reward of the newly executed Action,
+    feed it back to the expend node and all upstream nodes,
+    and update the corresponding data.
+    """
+
+    # update util the root node
+    while node is not None:
+        # update the visit number
+        node.update_visit_number()
+
+        # update the quality value
+        node.update_quality_value(reward)
+
+        # change the node to the parent node
+        node = node.parent
+
+
+def monte_carlo_tree_search(node):
+    """
+    Implement the Monte Carlo tree search algorithm, pass in a root node,
+    expand new nodes and update data according to the
+     tree structure that has been explored before in a limited time,
+    and then return as long as the child node with the highest exploitation.
+
+    When making predictions,
+    you only need to select the node with the largest exploitation according to the Q value,
+    and find the next optimal node.
+    """
+
+    computation_budget = len(AVAILABLE_CHOICES) * 3
+
+    # run as much as possible under the computation budget
+    for i in range(computation_budget):
+        # 1. find the best node to expand
+        expand_node = tree_policy(node)
+        if not expand_node:
+            # when it is None, it means that all nodes are added but no nodes meet the space limit
+            break
+        # 2. random get next action and get reward
+        reward = default_policy(expand_node)
+
+        # 3. update all passing nodes with reward
+        backpropagate(expand_node, reward)
+
+    # get the best next node
+    best_next_node = best_child(node, False)
+
+    return best_next_node
+
+
+def MCTS(workload_info, atomic_choices, available_choices, storage_threshold, max_index_num):
+    global ATOMIC_CHOICES, STORAGE_THRESHOLD, WORKLOAD_INFO, AVAILABLE_CHOICES, MAX_INDEX_NUM
+    WORKLOAD_INFO = workload_info
+    AVAILABLE_CHOICES = available_choices
+    ATOMIC_CHOICES = atomic_choices
+    STORAGE_THRESHOLD = storage_threshold
+    MAX_INDEX_NUM = max_index_num if max_index_num else len(available_choices)
+
+    # create the initialized state and initialized node
+    init_state = State()
+    choices = copy.copy(available_choices)
+    init_state.set_available_choices(choices)
+    init_node = Node()
+    init_node.set_state(init_state)
+    current_node = init_node
+
+    opt_config = []
+    # set the rounds to play
+    for i in range(len(AVAILABLE_CHOICES)):
+        if current_node:
+            current_node = monte_carlo_tree_search(current_node)
+            if current_node:
+                opt_config = current_node.state.accumulation_choices
+            else:
+                break
+    return opt_config

src/gausskernel/dbmind/tools/tests/test_index_advisor_workload.py

@@ -27,6 +27,7 @@ from collections.abc import Iterable
 from collections import defaultdict
 
 import index_advisor_workload as iaw
+import mcts
 
 
 def hash_any(obj):
@@ -227,6 +228,32 @@ select * from student_range_part1 where credit=1;
 
 class IndexAdvisorTester(unittest.TestCase):
 
+    def test_mcts(self):
+        storage_threshold = 12
+        index1 = iaw.IndexItem('public.a', 'col1', index_type='global')
+        index2 = iaw.IndexItem('public.b', 'col1', index_type='global')
+        index3 = iaw.IndexItem('public.c', 'col1', index_type='global')
+        index4 = iaw.IndexItem('public.d', 'col1', index_type='global')
+
+        atomic_index1 = iaw.IndexItem('public.a', 'col1', index_type='global')
+        atomic_index2 = iaw.IndexItem('public.b', 'col1', index_type='global')
+        atomic_index3 = iaw.IndexItem('public.c', 'col1', index_type='global')
+        atomic_index4 = iaw.IndexItem('public.d', 'col1', index_type='global')
+
+        atomic_index1.storage = 10
+        atomic_index2.storage = 4
+        atomic_index3.storage = 7
+        available_choices = [index1, index2, index3, index4]
+        atomic_choices = [[], [atomic_index2], [atomic_index1], [atomic_index3],
+                          [atomic_index2, atomic_index3], [atomic_index4]]
+        query = iaw.QueryItem('select * from gia_01', 1)
+        query.cost_list = [10, 7, 5, 9, 4, 11]
+        workload_info = [query]
+
+        results = mcts.MCTS(workload_info, atomic_choices, available_choices, storage_threshold, 2)
+        self.assertLessEqual([index1.atomic_pos, index2.atomic_pos, index3.atomic_pos], [2, 1, 3])
+        self.assertSetEqual({results[0].table, results[1].table}, {'public.b', 'public.c'})
+
     def test_get_indexable_columns(self):
         tables = 'table1 table2 table2 table3 table3 table3'.split()
         columns = 'col1,col2 col2 col3 col1,col2 col2,col3 col2,col5'.split()

src/gausskernel/optimizer/commands/dropcmds.cpp

@@ -90,7 +90,10 @@ static void DropExtensionInListIsSupported(List* objname)
         }
     }
 
-    ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("EXTENSION is not yet supported.")));
+    /* Enable DROP operation of the above objects during inplace upgrade or support_extended_features is true */
+    if (!u_sess->attr.attr_common.IsInplaceUpgrade && !g_instance.attr.attr_common.support_extended_features) {
+        ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("EXTENSION is not yet supported.")));
+    }
 }
 
 /*

src/gausskernel/optimizer/commands/extension.cpp

@@ -1175,7 +1175,7 @@ void CreateExtension(CreateExtensionStmt* stmt)
         FEATURE_NOT_PUBLIC_ERROR("EXTENSION is not yet supported.");
     }
 
-    if (pg_strcasecmp(stmt->extname, "b_sql_plugin") == 0 && !DB_IS_CMPT(B_FORMAT)) {
+    if (pg_strcasecmp(stmt->extname, "dolphin") == 0 && !DB_IS_CMPT(B_FORMAT)) {
         ereport(ERROR,
             (errmsg("please create extension \"%s\" with B type DBCOMPATIBILITY", stmt->extname)));
     }
@@ -1418,8 +1418,8 @@ void CreateExtension(CreateExtensionStmt* stmt)
 
     u_sess->exec_cxt.extension_is_valid = true;
 
-    if (pg_strcasecmp(stmt->extname, "b_sql_plugin") == 0) {
-         u_sess->attr.attr_sql.b_sql_plugin = true;
+    if (pg_strcasecmp(stmt->extname, "dolphin") == 0) {
+         u_sess->attr.attr_sql.dolphin = true;
     }
 
     /*

src/gausskernel/optimizer/commands/subscriptioncmds.cpp

@@ -44,7 +44,7 @@
 #include "utils/array.h"
 #include "utils/acl.h"
 
-static void ConnectPublisher(char *conninfo, char* slotname);
+static bool ConnectPublisher(char* conninfo, char* slotname);
 static void CreateSlotInPublisher(char *slotname);
 static void ValidateReplicationSlot(char *slotname, List *publications);
 
@@ -56,7 +56,7 @@ static void ValidateReplicationSlot(char *slotname, List *publications);
  * accommodate that.
  */
 static void parse_subscription_options(const List *options, char **conninfo, List **publications, bool *enabled_given,
-    bool *enabled, bool *slot_name_given, char **slot_name, char **synchronous_commit)
+    bool *enabled, bool *slot_name_given, char **slot_name, char **synchronous_commit, bool *binary_given, bool *binary)
 {
     ListCell *lc;
 
@@ -76,6 +76,10 @@ static void parse_subscription_options(const List *options, char **conninfo, Lis
     if (synchronous_commit) {
         *synchronous_commit = NULL;
     }
+    if (binary) {
+        *binary_given = false;
+        *binary = false;
+    }
 
     /* Parse options */
     foreach (lc, options) {
@@ -124,6 +128,15 @@ static void parse_subscription_options(const List *options, char **conninfo, Lis
             /* Test if the given value is valid for synchronous_commit GUC. */
             (void)set_config_option("synchronous_commit", *synchronous_commit, PGC_BACKEND, PGC_S_TEST, GUC_ACTION_SET,
                 false, 0, false);
+        } else if (strcmp(defel->defname, "binary") == 0 && binary) {
+            if (*binary_given) {
+                ereport(ERROR,
+                        (errcode(ERRCODE_SYNTAX_ERROR),
+                         errmsg("conflicting or redundant options")));
+            }
+
+            *binary_given = true;
+            *binary = defGetBoolean(defel);
         } else {
             ereport(ERROR,
                 (errcode(ERRCODE_SYNTAX_ERROR), errmsg("unrecognized subscription parameter: %s", defel->defname)));
@@ -197,26 +210,82 @@ static Datum publicationListToArray(List *publist)
 }
 
 /*
- * connect publisher and create slot.
- * the input conninfo should be encrypt, we will decrypt password inside
+ * Parse the original connection string which is encrypted, poll all hosts and ports,
+ * and try to connect to the publisher.
+ * When checkRemoteMode is true, the remotemode must be normal or primary.
+ * Return true to indicate successful connection.
  */
-static void ConnectPublisher(char *conninfo, char *slotname)
+bool AttemptConnectPublisher(const char *conninfoOriginal, char* slotname, bool checkRemoteMode)
+{
+    size_t conninfoLen = strlen(conninfoOriginal) + 1;
+
+    char* conninfo = NULL;
+    StringInfoData conninfoWithoutHostport;
+    initStringInfo(&conninfoWithoutHostport);
+    HostPort* hostPortList[MAX_REPLNODE_NUM] = {NULL};
+    ParseConninfo(conninfoOriginal, &conninfoWithoutHostport, hostPortList);
+    if (hostPortList[0] == NULL) {
+        ereport(ERROR, (errcode(ERRCODE_SYNTAX_ERROR), errmsg(
+            "invalid connection string syntax, missing host and port")));
+    }
+    bool connectSuccess = false;
+    conninfo = (char*)palloc(conninfoLen * sizeof(char));
+    for (int i = 0; i < MAX_REPLNODE_NUM; ++i) {
+        if (hostPortList[i] == NULL) {
+            break;
+        }
+        int ret = snprintf_s(conninfo, conninfoLen, conninfoLen - 1,
+            "%s host=%s port=%s", conninfoWithoutHostport.data,
+            hostPortList[i]->host, hostPortList[i]->port);
+        securec_check_ss(ret, "\0", "\0");
+
+        connectSuccess = ConnectPublisher(conninfo, slotname);
+        if (!connectSuccess) {
+            /* try next host */
+            continue;
+        }
+        if (!checkRemoteMode) {
+            break;
+        }
+        ServerMode publisherServerMde = IdentifyRemoteMode();
+        if (publisherServerMde == NORMAL_MODE || publisherServerMde == PRIMARY_MODE) {
+            break;
+        }
+        /* it's a standby, try next host */
+        (WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_disconnect();
+        connectSuccess = false;
+    }
+    pfree_ext(conninfo);
+
+    /* clean up */
+    FreeStringInfo(&conninfoWithoutHostport);
+    for (int i = 0; i < MAX_REPLNODE_NUM; ++i) {
+        if (hostPortList[i] == NULL) {
+            break;
+        }
+        pfree_ext(hostPortList[i]->host);
+        pfree_ext(hostPortList[i]->port);
+        pfree_ext(hostPortList[i]);
+    }
+    return connectSuccess;
+}
+
+/*
+ * connect to publisher with conninfo
+ */
+static bool ConnectPublisher(char* conninfo, char* slotname)
 {
     /* Try to connect to the publisher. */
     volatile WalRcvData *walrcv = t_thrd.walreceiverfuncs_cxt.WalRcv;
     SpinLockAcquire(&walrcv->mutex);
     walrcv->conn_target = REPCONNTARGET_PUBLICATION;
     SpinLockRelease(&walrcv->mutex);
-
-    char *decryptConninfo = DecryptConninfo(conninfo);
+    char* decryptConninfo = EncryptOrDecryptConninfo(conninfo, 'D');
     bool connectSuccess = (WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_connect(decryptConninfo, NULL, slotname, -1);
     int rc = memset_s(decryptConninfo, strlen(decryptConninfo), 0, strlen(decryptConninfo));
     securec_check(rc, "", "");
     pfree_ext(decryptConninfo);
-
-    if (!connectSuccess) {
-        ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), errmsg("could not connect to the publisher")));
-    }
+    return connectSuccess;
 }
 
 /*
@@ -293,9 +362,10 @@ ObjectAddress CreateSubscription(CreateSubscriptionStmt *stmt, bool isTopLevel)
     bool enabled_given = false;
     bool enabled = true;
     char *synchronous_commit;
-    char *conninfo;
     char *slotname;
     bool slotname_given;
+    bool binary;
+    bool binary_given;
     char originname[NAMEDATALEN];
     List *publications;
     int rc;
@@ -305,7 +375,7 @@ ObjectAddress CreateSubscription(CreateSubscriptionStmt *stmt, bool isTopLevel)
      * Connection and publication should not be specified here.
      */
     parse_subscription_options(stmt->options, NULL, NULL, &enabled_given, &enabled, &slotname_given, &slotname,
-        &synchronous_commit);
+        &synchronous_commit, &binary_given, &binary);
 
     /*
      * Since creating a replication slot is not transactional, rolling back
@@ -333,11 +403,10 @@ ObjectAddress CreateSubscription(CreateSubscriptionStmt *stmt, bool isTopLevel)
         synchronous_commit = "off";
     }
 
-    conninfo = stmt->conninfo;
     publications = stmt->publication;
 
     /* Check the connection info string. */
-    libpqrcv_check_conninfo(conninfo);
+    libpqrcv_check_conninfo(stmt->conninfo);
 
     /* Everything ok, form a new tuple. */
     rc = memset_s(values, sizeof(values), 0, sizeof(values));
@@ -349,18 +418,12 @@ ObjectAddress CreateSubscription(CreateSubscriptionStmt *stmt, bool isTopLevel)
     values[Anum_pg_subscription_subname - 1] = DirectFunctionCall1(namein, CStringGetDatum(stmt->subname));
     values[Anum_pg_subscription_subowner - 1] = ObjectIdGetDatum(owner);
     values[Anum_pg_subscription_subenabled - 1] = BoolGetDatum(enabled);
+    values[Anum_pg_subscription_subbinary - 1] = BoolGetDatum(binary);
 
     /* encrypt conninfo */
-    List *conninfoList = ConninfoToDefList(stmt->conninfo);
-    /* Sensitive options for subscription, will be encrypted when saved to catalog. */
-    const char* sensitiveOptionsArray[] = {"password"};
-    const int sensitiveArrayLength = lengthof(sensitiveOptionsArray);
-    EncryptGenericOptions(conninfoList, sensitiveOptionsArray, sensitiveArrayLength, SUBSCRIPTION_MODE);
-    char *encryptConninfo = DefListToString(conninfoList);
-
+    char *encryptConninfo = EncryptOrDecryptConninfo(stmt->conninfo, 'E');
     values[Anum_pg_subscription_subconninfo - 1] = CStringGetTextDatum(encryptConninfo);
 
-    pfree_ext(conninfoList);
     if (enabled) {
         if (!slotname_given) {
             slotname = stmt->subname;
@@ -396,11 +459,14 @@ ObjectAddress CreateSubscription(CreateSubscriptionStmt *stmt, bool isTopLevel)
      */
     if (enabled) {
         Assert(slotname);
-        ConnectPublisher(encryptConninfo, slotname);
+
+        if (!AttemptConnectPublisher(encryptConninfo, slotname, true)) {
+            ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), errmsg("Failed to connect to publisher.")));
+        }
+
         CreateSlotInPublisher(slotname);
         (WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_disconnect();
     }
-
     pfree_ext(encryptConninfo);
     heap_close(rel, RowExclusiveLock);
     rc = memset_s(stmt->conninfo, strlen(stmt->conninfo), 0, strlen(stmt->conninfo));
@@ -439,6 +505,8 @@ ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt)
     Oid subid;
     bool enabled_given = false;
     bool enabled;
+    bool binary_given;
+    bool binary;
     char *synchronous_commit;
     char *conninfo;
     char *slot_name;
@@ -473,7 +541,7 @@ ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt)
 
     /* Parse options. */
     parse_subscription_options(stmt->options, &conninfo, &publications, &enabled_given, &enabled, &slotname_given,
-        &slot_name, &synchronous_commit);
+        &slot_name, &synchronous_commit, &binary_given, &binary);
 
     /* Form a new tuple. */
     rc = memset_s(nulls, sizeof(nulls), false, sizeof(nulls));
@@ -490,23 +558,15 @@ ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt)
     if (conninfo) {
         /* Check the connection info string. */
         libpqrcv_check_conninfo(conninfo);
-
-        /* encrypt conninfo */
-        List *conninfoList = ConninfoToDefList(conninfo);
-        /* Sensitive options for subscription, will be encrypted when saved to catalog. */
-        const char* sensitiveOptionsArray[] = {"password"};
-        const int sensitiveArrayLength = lengthof(sensitiveOptionsArray);
-        EncryptGenericOptions(conninfoList, sensitiveOptionsArray, sensitiveArrayLength, SUBSCRIPTION_MODE);
-        encryptConninfo = DefListToString(conninfoList);
-        needFreeConninfo = true;
-
+        encryptConninfo = EncryptOrDecryptConninfo(conninfo, 'E');
+        rc = memset_s(conninfo, strlen(conninfo), 0, strlen(conninfo));
+        securec_check(rc, "\0", "\0");
         values[Anum_pg_subscription_subconninfo - 1] = CStringGetTextDatum(encryptConninfo);
         replaces[Anum_pg_subscription_subconninfo - 1] = true;
+        needFreeConninfo = true;
 
-        pfree_ext(conninfoList);
-
+        /* need to check whether new conninfo can be used to connect to new publisher */
         if (sub->enabled || (enabled_given && enabled)) {
-            /* we need to check whether new conninfo can be used to connect to new publisher */
             checkConn = true;
         }
     }
@@ -548,6 +608,10 @@ ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt)
         values[Anum_pg_subscription_subsynccommit - 1] = CStringGetTextDatum(synchronous_commit);
         replaces[Anum_pg_subscription_subsynccommit - 1] = true;
     }
+    if (binary_given) {
+        values[Anum_pg_subscription_subbinary - 1] = BoolGetDatum(binary);
+        replaces[Anum_pg_subscription_subbinary - 1] = true;
+    }
     if (publications != NIL) {
         values[Anum_pg_subscription_subpublications - 1] = publicationListToArray(publications);
         replaces[Anum_pg_subscription_subpublications - 1] = true;
@@ -570,16 +634,18 @@ ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt)
     if (sub->enabled && !enabled) {
         ereport(ERROR, (errmsg("If you want to deactivate this subscription, use DROP SUBSCRIPTION.")));
     }
-    /* enable subscription */
-    if (!sub->enabled && enabled) {
-        /* if slot hasn't been created, then create it */
-        if (!sub->slotname || !*(sub->slotname)) {
+    /* enabling subscription, but slot hasn't been created,
+     * then mark createSlot to true.
+     */
+    if (!sub->enabled && enabled && (!sub->slotname || !*(sub->slotname))) {
             createSlot = true;
-        }
     }
 
     if (checkConn || createSlot || validateSlot) {
-        ConnectPublisher(encryptConninfo, finalSlotName);
+        if (!AttemptConnectPublisher(encryptConninfo, finalSlotName, true)) {
+            ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), errmsg(
+                checkConn ? "The new conninfo cannot connect to new publisher." : "Failed to connect to publisher.")));
+        }
 
         if (createSlot) {
             CreateSlotInPublisher(finalSlotName);
@@ -597,12 +663,6 @@ ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt)
     if (needFreeConninfo) {
         pfree_ext(encryptConninfo);
     }
-
-    if (conninfo) {
-        rc = memset_s(conninfo, strlen(conninfo), 0, strlen(conninfo));
-        securec_check(rc, "", "");
-    }
-
     return myself;
 }
 
@@ -753,7 +813,11 @@ void DropSubscription(DropSubscriptionStmt *stmt, bool isTopLevel)
     initStringInfo(&cmd);
     appendStringInfo(&cmd, "DROP_REPLICATION_SLOT %s", quote_identifier(slotname));
 
-    ConnectPublisher(conninfo, slotname);
+    if (!AttemptConnectPublisher(conninfo, slotname, true)) {
+        ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), errmsg(
+            "could not connect to publisher.")));
+    }
+
     PG_TRY();
     {
         int sqlstate = 0;
@@ -779,6 +843,7 @@ void DropSubscription(DropSubscriptionStmt *stmt, bool isTopLevel)
 
     (WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_disconnect();
 
+    pfree_ext(conninfo);
     pfree(cmd.data);
     heap_close(rel, NoLock);
 }
@@ -908,3 +973,149 @@ void RenameSubscription(List *oldname, const char *newname)
 
     return;
 }
+
+/*
+ * Parse the host or port string into a string array,
+ * where host and port are separated by ",".
+ * input: conn --- host or port string separated by ","
+ * output: connArray --- host or port string array
+ * return: the length of connArray
+ * for example:
+ * (1):
+ * conn = 1.1.1.1,2.2.2.2,...,9.9.9.9
+ * connArray = {
+ *              1,.1.1.1,
+ *              2.2.2.2,
+ *              ...,
+ *              9.9.9.9
+ *              }
+ * return 9
+ * (2):
+ * conn = 1,2,...,9
+ * connArray = {1,2,...,9}
+ * return 9
+ */
+static int HostsPortsToArray(const char* conn, char** connArray)
+{
+    if (conn == NULL) {
+        return 0;
+    }
+    char* cp = NULL;
+    char* cur = NULL;
+    char *buf = pstrdup(conn);
+
+    cp = buf;
+    int i = 0;
+    while (*cp) {
+        cur = cp;
+        while (*cp && *cp != ',') {
+            ++cp;
+        }
+        if (*cp == ',') {
+            *cp = '\0';
+            ++cp;
+        }
+        if (i >= MAX_REPLNODE_NUM) {
+            ereport(ERROR, (errmsg("Currently, a maximum of %d servers are "
+                "supported.", MAX_REPLNODE_NUM)));
+        }
+        connArray[i++] = pstrdup(cur);
+
+        if (*cp == 0) {
+            break;
+        }
+    }
+    pfree(buf);
+    return i;
+}
+
+/*
+ * parse host and port
+ */
+static void ParseHostPort(char* hoststr, char* portstr, HostPort** hostPortList)
+{
+    char* hosts[MAX_REPLNODE_NUM] = {NULL};
+    char* ports[MAX_REPLNODE_NUM] = {NULL};
+    int hostNum = HostsPortsToArray(hoststr, hosts);
+    int portNum = HostsPortsToArray(portstr, ports);
+    if (hostNum != portNum) {
+        ereport(ERROR, (errcode(ERRCODE_SYNTAX_ERROR), errmsg("The number of host and port are inconsistent.")));
+    }
+
+    for (int i = 0; i < hostNum; ++i) {
+        hostPortList[i] = (HostPort*)palloc(sizeof(HostPort));
+        hostPortList[i]->host = hosts[i];
+        hostPortList[i]->port = ports[i];
+    }
+}
+
+/*
+ * Parse conninfo
+ * conninfo format:
+ *      'dbname=abc user=username password=xxxx host=ip1,ip2,...,ip9 port=p1,p2,...,p9'
+ * after parsing:
+ * conninfoWithoutHostPort:
+ *      'dbname=abc user=username password=xxxx'
+ * hostPortList:
+ *      {
+ *          {host=ip1, port=p1},
+ *          {host=ip2, port=p2},
+ *          ...
+ *          {host=ip9, port=p9}
+ *      }
+ */
+void ParseConninfo(const char* conninfo, StringInfoData* conninfoWithoutHostPort, HostPort** hostPortList)
+{
+    List* conninfoList = ConninfoToDefList(conninfo);
+    ListCell* l = NULL;
+
+    char* hostStr = NULL;
+    char* portStr = NULL;
+    foreach (l, conninfoList) {
+        DefElem* defel = (DefElem*)lfirst(l);
+        if (pg_strcasecmp(defel->defname, "host") == 0) {
+            hostStr = defGetString(defel);
+        } else if (pg_strcasecmp(defel->defname, "port") == 0) {
+            portStr = defGetString(defel);
+        } else {
+            appendStringInfo(conninfoWithoutHostPort, "%s=%s ", defel->defname, defGetString(defel));
+        }
+    }
+    if (hostPortList != NULL) {
+        ParseHostPort(hostStr, portStr, hostPortList);
+    }
+}
+
+/*
+ * encrypt conninfo when action = 'E'
+ * decrypt conninfo when action = 'D'
+ * conninfoNew: encrypted or decrypted conninfo
+ */
+char* EncryptOrDecryptConninfo(const char* conninfo, const char action)
+{
+    /* parse conninfo to list */
+    List *conninfoList = ConninfoToDefList(conninfo);
+    /* Sensitive options for subscription */
+    const char* sensitiveOptionsArray[] = {"password"};
+    const int sensitiveArrayLength = lengthof(sensitiveOptionsArray);
+    switch (action) {
+        /* Encrypt */
+        case 'E':
+            EncryptGenericOptions(conninfoList, sensitiveOptionsArray, sensitiveArrayLength, SUBSCRIPTION_MODE);
+            break;
+
+        /* Decrypt */
+        case 'D':
+            DecryptOptions(conninfoList, sensitiveOptionsArray, sensitiveArrayLength, SUBSCRIPTION_MODE);
+            break;
+
+        default:
+            break;
+    }
+
+    char* conninfoNew = DefListToString(conninfoList);
+    ClearListContent(conninfoList);
+    list_free_ext(conninfoList);
+
+    return conninfoNew;
+}

src/gausskernel/optimizer/commands/tablecmds.cpp

@@ -23175,7 +23175,7 @@ static void checkValidationForExchangeTable(Relation partTableRel, Relation ordT
                 int2 bucketId = InvalidBktId;
 
                 // get right partition oid for the tuple
-                targetPartOid = heapTupleGetPartitionId(partTableRel, (HeapTuple) tuple);
+                targetPartOid = heapTupleGetPartitionId(partTableRel, (HeapTuple)tuple, true);
 
                 searchFakeReationForPartitionOid(
                     partRelHTAB, CurrentMemoryContext, partTableRel, targetPartOid, partRel, part, RowExclusiveLock);
@@ -24797,7 +24797,8 @@ static Oid AddTemporaryPartitionForAlterPartitions(const AlterTableCmd* cmd, Rel
             destPartOid = AddTemporaryHashPartitionForAlterPartitions(cmd, partTableRel, partSeq, renameTargetPart);
             break;
         }
-        case PART_TYPE_RANGE: {
+        case PART_TYPE_RANGE:
+        case PART_TYPE_INTERVAL: {
             destPartOid = AddTemporaryRangePartitionForAlterPartitions(cmd, partTableRel, partSeq, renameTargetPart);
             break;
         }
@@ -25098,11 +25099,11 @@ static void readTuplesAndInsertInternal(Relation tempTableRel, Relation partTabl
 
         /* tableam_tops_copy_tuple is not ready so we add UStore hack path */
         copyTuple = tableam_tops_copy_tuple(tuple);
-        targetPartOid = heapTupleGetPartitionId(partTableRel, (void *)tuple);
+        targetPartOid = heapTupleGetPartitionId(partTableRel, (void *)tuple, true);
         searchFakeReationForPartitionOid(
             partRelHTAB, CurrentMemoryContext, partTableRel, targetPartOid, partRel, part, RowExclusiveLock);
         if (RelationIsSubPartitioned(partTableRel)) {
-            targetPartOid = heapTupleGetPartitionId(partRel, (void *)tuple);
+            targetPartOid = heapTupleGetPartitionId(partRel, (void *)tuple, true);
             searchFakeReationForPartitionOid(partRelHTAB, CurrentMemoryContext, partRel, targetPartOid, subPartRel,
                                              subPart, RowExclusiveLock);
             partRel = subPartRel;

src/gausskernel/optimizer/prep/prepjointree.cpp

@@ -1181,6 +1181,17 @@ static Node* pull_up_simple_subquery(PlannerInfo* root, Node* jtnode, RangeTblEn
         return jtnode;
     }
 
+    /*
+     * We must flatten any join alias Vars in the subquery's targetlist,
+     * because pulling up the subquery's subqueries might have changed their
+     * expansions into arbitrary expressions, which could affect
+     * pullup_replace_vars' decisions about whether PlaceHolderVar wrappers
+     * are needed for tlist entries.  (Likely it'd be better to do
+     * flatten_join_alias_vars on the whole query tree at some earlier stage,
+     * maybe even in the rewriter; but for now let's just fix this case here.)
+     */
+    subquery->targetList = (List *) flatten_join_alias_vars(subroot, (Node *) subquery->targetList);
+
     /*
      * Adjust level-0 varnos in subquery so that we can append its rangetable
      * to upper query's.  We have to fix the subquery's append_rel_list as

src/gausskernel/process/postmaster/pagewriter.cpp

@@ -1263,7 +1263,7 @@ static void ckpt_pagewriter_main_thread_loop(void)
         HandlePageWriterMainInterrupts();
 
         candidate_num = get_curr_candidate_nums(false) + get_curr_candidate_nums(true);
-        if (candidate_num == 0) {
+        if (candidate_num == 0 && !t_thrd.pagewriter_cxt.shutdown_requested) {
             /* wakeup sub thread scan the buffer pool, init the candidate list */
             wakeup_sub_thread();
         }

src/gausskernel/process/tcop/postgres.cpp

@@ -825,10 +825,10 @@ void client_read_ended(void)
 #define INIT_PLUGIN_OBJECT "init_plugin_object"
 void InitBSqlPluginHookIfNeeded()
 {
-    const char* b_sql_plugin = "b_sql_plugin";
+    const char* dolphin = "dolphin";
     CFunInfo tmpCF;
 
-    tmpCF = load_external_function(b_sql_plugin, INIT_PLUGIN_OBJECT, false, false);
+    tmpCF = load_external_function(dolphin, INIT_PLUGIN_OBJECT, false, false);
     if (tmpCF.user_fn != NULL) {
         ((void* (*)(void))(tmpCF.user_fn))();
     }
@@ -862,9 +862,11 @@ List* pg_parse_query(const char* query_string, List** query_string_locationlist)
 
     List* (*parser_hook)(const char*, List**) = raw_parser;
 #ifndef ENABLE_MULTIPLE_NODES
-    int id = GetCustomParserId();
-    if (id >= 0 && g_instance.raw_parser_hook[id] != NULL) {
-        parser_hook = (List* (*)(const char*, List**))g_instance.raw_parser_hook[id];
+    if (u_sess->attr.attr_sql.dolphin) {
+        int id = GetCustomParserId();
+        if (id >= 0 && g_instance.raw_parser_hook[id] != NULL) {
+            parser_hook = (List* (*)(const char*, List**))g_instance.raw_parser_hook[id];
+        }
     }
 #endif
     raw_parsetree_list = parser_hook(query_string, query_string_locationlist);
@@ -6112,6 +6114,9 @@ void ProcessInterrupts(void)
 
             /* The logical replication launcher can be stopped at any time. */
             proc_exit(0);
+        } else if (IsLogicalWorker()) {
+            ereport(FATAL, (errcode(ERRCODE_ADMIN_SHUTDOWN),
+                errmsg("terminating logical replication worker due to administrator command")));
 #endif
         } else if (IsTxnSnapCapturerProcess()) {
             ereport(FATAL,
@@ -7569,7 +7574,7 @@ int PostgresMain(int argc, char* argv[], const char* dbname, const char* usernam
         init_set_params_htab();
 
 #ifndef ENABLE_MULTIPLE_NODES
-    if (u_sess->proc_cxt.MyDatabaseId != InvalidOid && DB_IS_CMPT(B_FORMAT) && u_sess->attr.attr_sql.b_sql_plugin) {
+    if (u_sess->proc_cxt.MyDatabaseId != InvalidOid && DB_IS_CMPT(B_FORMAT) && u_sess->attr.attr_sql.dolphin) {
         InitBSqlPluginHookIfNeeded();
     }
 #endif

src/gausskernel/process/threadpool/threadpool_worker.cpp

@@ -848,7 +848,7 @@ static bool InitSession(knl_session_context* session)
     t_thrd.proc_cxt.PostInit->InitSession();
 
 #ifndef ENABLE_MULTIPLE_NODES
-    if (u_sess->proc_cxt.MyDatabaseId != InvalidOid && DB_IS_CMPT(B_FORMAT) && u_sess->attr.attr_sql.b_sql_plugin) {
+    if (u_sess->proc_cxt.MyDatabaseId != InvalidOid && DB_IS_CMPT(B_FORMAT) && u_sess->attr.attr_sql.dolphin) {
         InitBSqlPluginHookIfNeeded();
     }
 #endif

src/gausskernel/runtime/executor/execQual.cpp

@@ -720,7 +720,6 @@ static Datum ExecEvalWholeRowVar(
 {
     Var* variable = (Var*)wrvstate->xprstate.expr;
     TupleTableSlot* slot = NULL;
-    TupleDesc slot_tupdesc;
     bool needslow = false;
 
     if (isDone != NULL)
@@ -802,21 +801,13 @@ static Datum ExecEvalWholeRowVar(
     if (wrvstate->wrv_junkFilter != NULL)
         slot = ExecFilterJunk(wrvstate->wrv_junkFilter, slot);
 
-    slot_tupdesc = slot->tts_tupleDescriptor;
-
     /*
-     * If it's a RECORD Var, we'll use the slot's type ID info.  It's likely
-     * that the slot's type is also RECORD; if so, make sure it's been
-     * "blessed", so that the Datum can be interpreted later.
-     *
      * If the Var identifies a named composite type, we must check that the
      * actual tuple type is compatible with it.
      */
-    if (variable->vartype == RECORDOID) {
-        if (slot_tupdesc->tdtypeid == RECORDOID && slot_tupdesc->tdtypmod < 0)
-            assign_record_type_typmod(slot_tupdesc);
-    } else {
+    if (variable->vartype != RECORDOID) {
         TupleDesc var_tupdesc;
+        TupleDesc slot_tupdesc;
         int i;
 
         /*
@@ -833,6 +824,8 @@ static Datum ExecEvalWholeRowVar(
          */
         var_tupdesc = lookup_rowtype_tupdesc(variable->vartype, -1);
 
+        slot_tupdesc = slot->tts_tupleDescriptor;
+
         if (var_tupdesc->natts != slot_tupdesc->natts)
             ereport(ERROR,
                 (errcode(ERRCODE_DATATYPE_MISMATCH),
@@ -886,6 +879,7 @@ static Datum ExecEvalWholeRowFast(
 {
     Var* variable = (Var*)wrvstate->xprstate.expr;
     TupleTableSlot* slot = NULL;
+    TupleDesc slot_tupdesc;
     HeapTuple tuple;
     TupleDesc tupleDesc;
     HeapTupleHeader dtuple;
@@ -915,6 +909,17 @@ static Datum ExecEvalWholeRowFast(
     if (wrvstate->wrv_junkFilter != NULL)
         slot = ExecFilterJunk(wrvstate->wrv_junkFilter, slot);
 
+    /*
+     * If it's a RECORD Var, we'll use the slot's type ID info.  It's likely
+     * that the slot's type is also RECORD; if so, make sure it's been
+     * "blessed", so that the Datum can be interpreted later.
+     */        
+    slot_tupdesc = slot->tts_tupleDescriptor;
+    if (variable->vartype == RECORDOID) {
+        if (slot_tupdesc->tdtypeid == RECORDOID && slot_tupdesc->tdtypmod < 0)
+            assign_record_type_typmod(slot_tupdesc);
+    }
+
     tuple = ExecFetchSlotTuple(slot);
     tupleDesc = slot->tts_tupleDescriptor;
 

src/gausskernel/runtime/executor/execReplication.cpp

@@ -322,7 +322,6 @@ static bool RelationFindReplTupleByIndex(EState *estate, Relation rel, Relation
         }
         if (found) {
             /* Found tuple, try to lock it in the lockmode. */
-            outslot->tts_tuple = ExecMaterializeSlot(outslot);
             xwait = TransactionIdIsValid(snap.xmin) ? snap.xmin : snap.xmax;
             /*
              * If the tuple is locked, wait for locking transaction to finish
@@ -346,12 +345,16 @@ static bool RelationFindReplTupleByIndex(EState *estate, Relation rel, Relation
             ItemPointer tid = tableam_tops_get_t_self(targetRel, outslot->tts_tuple);
 
             if (RelationIsUstoreFormat(targetRel)) {
+                /* materialize the slot, so we can visit it after the scan is end */
+                outslot->tts_tuple = UHeapMaterialize(outslot);
                 ItemPointerCopy(tid, &UHeaplocktup.ctid);
                 rc = memset_s(&tbuf, sizeof(tbuf), 0, sizeof(tbuf));
                 securec_check(rc, "\0", "\0");
                 UHeaplocktup.disk_tuple = &tbuf.hdr;
                 locktup = &UHeaplocktup;
             } else {
+                /* materialize the slot, so we can visit it after the scan is end */
+                outslot->tts_tuple = ExecMaterializeSlot(outslot);
                 ItemPointerCopy(tid, &heaplocktup.t_self);
                 locktup = &heaplocktup;
             }
@@ -401,6 +404,10 @@ static bool tuple_equals_slot(TupleDesc desc, const Tuple tup, TupleTableSlot *s
     /* Check equality of the attributes. */
     for (attrnum = 0; attrnum < desc->natts; attrnum++) {
         TypeCacheEntry *typentry;
+        /* skip generate column */
+        if (GetGeneratedCol(desc, attrnum)) {
+            continue;
+        }
         /*
          * If one value is NULL and other is not, then they are certainly not
          * equal
@@ -478,7 +485,6 @@ static bool RelationFindReplTupleSeq(Relation rel, LockTupleMode lockmode, Tuple
 
             found = true;
             ExecStoreTuple(scantuple, outslot, InvalidBuffer, false);
-            outslot->tts_tuple = ExecMaterializeSlot(outslot);
 
             xwait = TransactionIdIsValid(snap.xmin) ? snap.xmin : snap.xmax;
             /*
@@ -511,12 +517,16 @@ static bool RelationFindReplTupleSeq(Relation rel, LockTupleMode lockmode, Tuple
             ItemPointer tid = tableam_tops_get_t_self(rel, outslot->tts_tuple);
 
             if (RelationIsUstoreFormat(targetRel)) {
+                /* materialize the slot, so we can visit it after the scan is end */
+                outslot->tts_tuple = UHeapMaterialize(outslot);
                 ItemPointerCopy(tid, &UHeaplocktup.ctid);
                 rc = memset_s(&tbuf, sizeof(tbuf), 0, sizeof(tbuf));
                 securec_check(rc, "\0", "\0");
                 UHeaplocktup.disk_tuple = &tbuf.hdr;
                 locktup = &UHeaplocktup;
             } else {
+                /* materialize the slot, so we can visit it after the scan is end */
+                outslot->tts_tuple = ExecMaterializeSlot(outslot);
                 ItemPointerCopy(tid, &heaplocktup.t_self);
                 locktup = &heaplocktup;
             }
@@ -653,7 +663,6 @@ void ExecSimpleRelationUpdate(EState *estate, EPQState *epqstate, TupleTableSlot
     /* Compute stored generated columns */
     if (rel->rd_att->constr && rel->rd_att->constr->has_generated_stored) {
         ExecComputeStoredGenerated(resultRelInfo, estate, slot, tuple, CMD_UPDATE);
-        tuple = slot->tts_tuple;
     }
 
     /* Check the constraints of the tuple */
@@ -672,10 +681,11 @@ void ExecSimpleRelationUpdate(EState *estate, EPQState *epqstate, TupleTableSlot
         rowMovement = true;
     }
 
+    tuple = slot->tts_tuple;
     CommandId cid = GetCurrentCommandId(true);
     /* OK, update the tuple and index entries for it */
     if (!rowMovement) {
-        res = tableam_tuple_update(targetRelation, parentRelation, searchSlotTid, slot->tts_tuple, cid,
+        res = tableam_tuple_update(targetRelation, parentRelation, searchSlotTid, tuple, cid,
             InvalidSnapshot, estate->es_snapshot, true, &oldslot, &tmfd, &updateIndexes, &modifiedIdxAttrs,
             false, allowInplaceUpdate);
         CheckTupleModifyRes(res);

src/gausskernel/storage/access/common/reloptions.cpp

@@ -2947,9 +2947,9 @@ void SetOneOfCompressOption(DefElem* defElem, TableCreateSupport* tableCreateSup
     } else if (pg_strcasecmp(defname, "compress_level") == 0) {
         tableCreateSupport->compressLevel = true;
     } else if (pg_strcasecmp(defname, "compress_byte_convert") == 0) {
-        tableCreateSupport->compressByteConvert = true;
+        tableCreateSupport->compressByteConvert = defGetBoolean(defElem);
     } else if (pg_strcasecmp(defname, "compress_diff_convert") == 0) {
-        tableCreateSupport->compressDiffConvert = true;
+        tableCreateSupport->compressDiffConvert = defGetBoolean(defElem);
     }
 }
 

src/gausskernel/storage/access/transam/xlog.cpp

@@ -8965,6 +8965,8 @@ void StartupXLOG(void)
     g_instance.comm_cxt.predo_cxt.redoPf.redo_done_time = 0;
     pg_atomic_write_u32(&(g_instance.comm_cxt.localinfo_cxt.is_finish_redo), 0);
 
+    NotifyGscRecoveryStarted();
+
     /*
      * Initialize WAL insert status array and the flush index - lastWalStatusEntryFlushed.
      */

src/gausskernel/storage/access/ustore/knl_utuple.cpp

@@ -1818,3 +1818,47 @@ void UHeapSlotStoreUHeapTuple(UHeapTuple utuple, TupleTableSlot *slot, bool shou
     /* Mark extracted state invalid */
     slot->tts_nvalid = 0;
 }
+
+/*
+ * Make the contents of the uheap table's slot contents solely depend on the slot(make them a local copy),
+ * and not on underlying external resources like another memory context, buffers etc.
+ *
+ * @pram slot: slot to be materialized.
+ */
+Tuple UHeapMaterialize(TupleTableSlot *slot)
+{
+    Assert(!slot->tts_isempty);
+    Assert(slot->tts_tupslotTableAm == TAM_USTORE);
+    Assert(slot->tts_tupleDescriptor != NULL);
+    /*
+     * If we have a regular physical tuple, and it's locally palloc'd, we have
+     * nothing to do.
+     */
+    if (slot->tts_tuple && slot->tts_shouldFree) {
+        return slot->tts_tuple;
+    }
+
+    /*
+     * Otherwise, copy or build a physical tuple, and store it into the slot.
+     *
+     * We may be called in a context that is shorter-lived than the tuple
+     * slot, but we have to ensure that the materialized tuple will survive
+     * anyway.
+     */
+    MemoryContext old_context = MemoryContextSwitchTo(slot->tts_mcxt);
+    if (slot->tts_tuple != NULL) {
+        slot->tts_tuple = UHeapCopyTuple((UHeapTuple)slot->tts_tuple);
+    } else {
+        slot->tts_tuple = UHeapFormTuple(slot->tts_tupleDescriptor, slot->tts_values, slot->tts_isnull);
+    }
+    slot->tts_shouldFree = true;
+    MemoryContextSwitchTo(old_context);
+
+    /*
+     * Have to deform from scratch, otherwise tts_values[] entries could point
+     * into the non-materialized tuple (which might be gone when accessed).
+     */
+    slot->tts_nvalid = 0;
+    return slot->tts_tuple;
+}
+

src/gausskernel/storage/mot/core/memory/garbage_collector/mm_gc_manager.h

@@ -238,15 +238,19 @@ public:
         return m_purpose;
     }
 
-    void GcStartTxnMTtests()
+    GcEpochType GcStartInnerTxn()
     {
-        if (m_gcEpoch != GetGlobalEpoch())
-            m_gcEpoch = GetGlobalEpoch();
+        m_gcEpoch = GetGlobalEpoch();
+
+        return m_gcEpoch;
     }
 
-    void GcEndTxnMTtests()
+    void GcEndInnerTxn(bool clean_gc)
     {
-        RunQuicese();
+        if (clean_gc) {
+            RunQuicese();
+        }
+        m_gcEpoch = 0;
     }
 
     void GcStartTxn()
@@ -272,6 +276,7 @@ public:
             RunQuicese();
             m_managerLock.unlock();
         }
+        m_gcEpoch = 0;
         m_isTxnStarted = false;
     }
 
@@ -313,7 +318,6 @@ public:
     {
         if (m_performGcEpoch != g_gcActiveEpoch)
             HardQuiesce(m_rcuFreeCount);
-        m_gcEpoch = 0;
     }
 
     /** @brief Clean all object at the end of the session */

src/gausskernel/storage/mot/core/storage/index/index.cpp

@@ -258,6 +258,15 @@ bool Index::IndexInsert(Sentinel*& outputSentinel, const Key* key, uint32_t pid,
         outputSentinel = IndexInsertImpl(key, sentinel, inserted, pid);
         // sync between rollback/delete and insert
         if (inserted == false) {
+            if (unlikely(outputSentinel == nullptr)) {
+                MOT_REPORT_ERROR(
+                    MOT_ERROR_OOM, "Index Insert", "Failed to insert sentinel to index %s", m_name.c_str());
+                rc = RC_MEMORY_ALLOCATION_ERROR;
+                m_sentinelPool->Release<Sentinel>(sentinel);
+                sentinel = nullptr;
+                return false;
+            }
+
             // Spin if the counter is 0 - aborting in parallel or sentinel is marks for commit
             if (outputSentinel->RefCountUpdate(INC, pid) == RC_OK)
                 retryInsert = false;
@@ -300,8 +309,16 @@ Sentinel* Index::IndexInsert(const Key* key, Row* row, uint32_t pid)
         // no need to report to full error stack
         SetLastError(MOT_ERROR_UNIQUE_VIOLATION, MOT_SEVERITY_NORMAL);
         m_sentinelPool->Release<Sentinel>(sentinel);
+        sentinel = nullptr;
         return nullptr;
     } else {
+        if (inserted == false) {
+            MOT_REPORT_ERROR(MOT_ERROR_OOM, "Index Insert", "Failed to insert sentinel to index %s", m_name.c_str());
+            m_sentinelPool->Release<Sentinel>(sentinel);
+            sentinel = nullptr;
+            return nullptr;
+        }
+
         if (GetIndexOrder() == IndexOrder::INDEX_ORDER_PRIMARY) {
             sentinel->SetPrimaryIndex();
             sentinel->SetNextPtr(row);
@@ -343,9 +360,6 @@ Sentinel* Index::IndexReadHeader(const Key* key, uint32_t pid) const
 Sentinel* Index::IndexRemove(const Key* key, uint32_t pid)
 {
     Sentinel* sentinel = IndexRemoveImpl(key, pid);
-
-    MOT_ASSERT(sentinel != nullptr);
-    MOT_ASSERT(sentinel->GetCounter() == 0);
     return sentinel;
 }
 

src/gausskernel/storage/mot/core/storage/index/masstree/mot_masstree_insert.hpp

@@ -56,12 +56,22 @@ void* basic_table<P>::insert(MOT::Key const* const& key, void* const& entry, boo
           5. Update the the key slice, keylen, key suffix and key's value in the
        leaf
           6. Add the key's location in permutation's back (key is not visible for
-       readers yet) As key's location is not part of the permutation yet, the key
+       readers yet) as key's location is not part of the permutation yet, the key
        is not reachable (aka not present). In addition, the leaf is still locked.
         Unlocking the node and enter the key into the permutation will be done
-       later in finish_insert (called from lp.finish). */
+       later in finish_insert (done in lp.finish function). */
 
-    bool found = lp.find_insert(*mtSessionThreadInfo);
+    bool found = false;
+    if (!lp.find_insert(*mtSessionThreadInfo, found)) {
+        // Failed to insert key due to memory allocation failure.
+        MOT_ASSERT(!mtSessionThreadInfo->non_disruptive_error());
+        MOT_ASSERT(found == false);
+        lp.finish(0, *mtSessionThreadInfo);
+        result = false;
+        return nullptr;
+    }
+
+    MOT_ASSERT(mtSessionThreadInfo->non_disruptive_error());
 
     // If the key is new (not previously existing) then we record the entry under
     // that key

src/gausskernel/storage/mot/core/storage/index/masstree/mot_masstree_remove.hpp

@@ -33,15 +33,15 @@ namespace Masstree {
 template <typename P>
 struct gc_layer_rcu_callback_ng : public P::threadinfo_type::mrcu_callback {
     typedef typename P::threadinfo_type threadinfo;
-    node_base<P>* root_;
+    node_base<P>** root_ref_;
     int len_;
     size_t size_;
     MOT::MasstreePrimaryIndex* index_;
     char s_[0];
-    gc_layer_rcu_callback_ng(node_base<P>* root, Str prefix, size_t size)
-        : root_(root), len_(prefix.length()), size_(size), index_(mtSessionThreadInfo->get_working_index())
+    gc_layer_rcu_callback_ng(node_base<P>** root_ref, Str prefix, size_t size)
+        : root_ref_(root_ref), len_(prefix.length()), size_(size), index_(mtSessionThreadInfo->get_working_index())
     {
-        errno_t erc = memcpy_s(s_, size_, prefix.data(), len_);
+        errno_t erc = memcpy_s(s_, len_, prefix.data(), len_);
         securec_check(erc, "\0", "\0");
     }
     size_t operator()(bool drop_index);
@@ -51,7 +51,7 @@ struct gc_layer_rcu_callback_ng : public P::threadinfo_type::mrcu_callback {
         return size_;
     }
 
-    static void make(node_base<P>* root, Str prefix, threadinfo& ti);
+    static void make(node_base<P>** root_ref, Str prefix, threadinfo& ti);
 };
 
 template <typename P>
@@ -60,8 +60,8 @@ size_t gc_layer_rcu_callback_ng<P>::operator()(bool drop_index)
     // If drop_index == true, all index's pools are going to be cleaned, so we can skip gc_layer call (which might add
     // more elements into GC)
     if (drop_index == false) {
-        // GC layer remove might delete elements from tree and add them to the limbolist. Index must be provided to
-        // allow access to the memory pools.
+        // GC layer remove might delete elements from tree and might create new gc layer removal requests and add them to GC.
+        // Index must be provided to allow access to the memory pools.
         mtSessionThreadInfo->set_working_index(index_);
         (*this)(*mtSessionThreadInfo);
         mtSessionThreadInfo->set_working_index(NULL);
@@ -73,30 +73,33 @@ size_t gc_layer_rcu_callback_ng<P>::operator()(bool drop_index)
 template <typename P>
 void gc_layer_rcu_callback_ng<P>::operator()(threadinfo& ti)
 {
-    // root_ node while creating gc_layer_rcu_callback_ng might not be the current root. Find updated tree's root.
-    while (!root_->is_root()) {
-        root_ = root_->maybe_parent();
-    }
+    masstree_invariant(root_ref_);
 
-    // If root was already deleted, do nothing.
-    if (root_->deleted()) {
-        return;
-    }
-
-    tcursor<P> node_cursor(root_, s_, len_);
-    if (!node_cursor.gc_layer(ti) || !node_cursor.finish_remove(ti)) {
+    tcursor<P> node_cursor(root_ref_, s_, len_);
+    bool do_remove = node_cursor.gc_layer(ti);
+    if (!do_remove || !node_cursor.finish_remove(ti)) {
         node_cursor.n_->unlock();
     }
+    ti.add_nodes_to_gc();
 }
 
 template <typename P>
-void gc_layer_rcu_callback_ng<P>::make(node_base<P>* root, Str prefix, threadinfo& ti)
+void gc_layer_rcu_callback_ng<P>::make(node_base<P>** root_ref, Str prefix, threadinfo& ti)
 {
     size_t sz = prefix.len + sizeof(gc_layer_rcu_callback_ng<P>);
-    // As we are using slab allocator to allocate the memory, sz is will updated in ti.allocate with the real allocated
-    // size
-    void* data = ti.allocate(sz, memtag_masstree_gc, &sz /*OUT PARAM*/);
-    gc_layer_rcu_callback_ng<P>* cb = new (data) gc_layer_rcu_callback_ng<P>(root, prefix, sz);
+    // As we are using slab allocator for allocation, sz is will be updated by ti.allocate with the real allocation
+    // size. We need this size for GC deallocation size report
+    void* data = ti.allocate(sz, memtag_masstree_gc, &sz /* IN/OUT PARAM */);
+    if (!data) {
+        // If allocation fails, gc layer removal command will not be added to GC and this layer wont be removed.
+        // We might deal with this issue in the future by replacing the current mechanism with one of the following options:
+        //    1. Use thread local GC layer removal object (per threadinfo) and keep list of key suffixes to clean (also in threadinfo)
+        //    2. Move this feature to VACUUM process: Create special iterator that adds GC Layer callbacks when it finds empty layers
+        ti.set_last_error(MT_MERR_GC_LAYER_REMOVAL_MAKE);
+        return;
+    }
+
+    gc_layer_rcu_callback_ng<P>* cb = new (data) gc_layer_rcu_callback_ng<P>(root_ref, prefix, sz);
     ti.rcu_register(cb, sz);
 }
 

src/gausskernel/storage/mot/core/storage/index/masstree_index.cpp

@@ -69,14 +69,17 @@ Sentinel* MasstreePrimaryIndex::IndexInsertImpl(const Key* key, Sentinel* sentin
     mtSessionThreadInfo->set_gc_session(
         MOTEngine::GetInstance()->GetCurrentGcSession());  // set current GC session in thread-pooled envelope
 
+    mtSessionThreadInfo->set_last_error(MT_MERR_OK);
+
     existingItem = m_index.insert(key, sentinel, inserted, pid);
 
     mtSessionThreadInfo->set_gc_session(NULL);
     mtSessionThreadInfo->set_working_index(NULL);
 
-    if (!inserted) {  // key mapping already exists in unique index
+    if (!inserted && existingItem) {  // key mapping already exists in unique index
         result = reinterpret_cast<Sentinel*>(existingItem);
-    }  // otherwise return null pointer
+    }  // otherwise return null pointer (if !inserted && !existingItem, Key does not exist and insertation failed due to
+       // memory issue)
 
     return result;
 }
@@ -108,6 +111,8 @@ Sentinel* MasstreePrimaryIndex::IndexRemoveImpl(const Key* key, uint32_t pid)
     mtSessionThreadInfo->set_gc_session(
         MOTEngine::GetInstance()->GetCurrentGcSession());  // set current GC session in thread-pooled envelope
 
+    mtSessionThreadInfo->set_last_error(MT_MERR_OK);
+
     output = m_index.remove(key->GetKeyBuf(), key->GetKeyLength(), result, pid);
 
     mtSessionThreadInfo->set_gc_session(NULL);

src/gausskernel/storage/mot/core/storage/index/masstree_index.h

@@ -37,6 +37,7 @@
 #include "masstree/mot_masstree_struct.hpp"
 #include "masstree/mot_masstree_iterator.hpp"
 #include <cmath>
+#include "mot_engine.h"
 
 namespace MOT {
 /**
@@ -302,11 +303,38 @@ public:
     /**
      * @brief Print Masstree pools memory consumption details to log.
      */
-    virtual void PrintPoolsStats()
+    virtual void PrintPoolsStats(LogLevel level = LogLevel::LL_DEBUG)
     {
-        m_leafsPool->Print("Leafs pool: ");
-        m_internodesPool->Print("Internode pool: ");
-        m_ksuffixSlab->Print("Ksuffix slab: ");
+        m_leafsPool->Print("Leafs pool", level);
+        m_internodesPool->Print("Internode pool", level);
+        m_ksuffixSlab->Print("Ksuffix slab", level);
+    }
+
+    virtual void GetLeafsPoolStats(uint64_t& objSize, uint64_t& numUsedObj, uint64_t& totalSize, uint64_t& netto)
+    {
+        PoolStatsSt stats = {};
+        m_leafsPool->GetStats(stats);
+
+        objSize = stats.m_objSize;
+        numUsedObj = stats.m_totalObjCount - stats.m_freeObjCount;
+        totalSize = stats.m_poolCount * stats.m_poolGrossSize;
+        netto = numUsedObj * objSize;
+    }
+
+    virtual void GetInternodesPoolStats(uint64_t& objSize, uint64_t& numUsedObj, uint64_t& totalSize, uint64_t& netto)
+    {
+        PoolStatsSt stats = {};
+        m_internodesPool->GetStats(stats);
+
+        objSize = stats.m_objSize;
+        numUsedObj = stats.m_totalObjCount - stats.m_freeObjCount;
+        totalSize = stats.m_poolCount * stats.m_poolGrossSize;
+        netto = numUsedObj * objSize;
+    }
+
+    virtual PoolStatsSt* GetKsuffixSlabStats()
+    {
+        return m_ksuffixSlab->GetStats();
     }
 
     /**
@@ -316,6 +344,8 @@ public:
     {
         m_initialized = false;
         DestroyPools();
+        // remove masstree's root pointer (not valid anymore)
+        *(m_index.root_ref()) = nullptr;
 
         return IndexInitImpl(NULL);
     }
@@ -332,7 +362,7 @@ public:
      * @param tag Hint to determine which pool to use.
      * @return Pointer to allocated memory.
      */
-    void* AllocateMem(int& size, enum memtag tag)
+    virtual void* AllocateMem(int& size, enum memtag tag)
     {
         switch (tag) {
             case memtag_masstree_leaf:
@@ -360,7 +390,7 @@ public:
      * @param Pointer to allocated memory.
      * @return True if deallocation succeeded.
      */
-    bool DeallocateMem(void* ptr, int size, enum memtag tag)
+    virtual bool DeallocateMem(void* ptr, int size, enum memtag tag)
     {
         switch (tag) {
             case memtag_masstree_leaf:
@@ -431,10 +461,16 @@ public:
     {
         // If dropIndex == true, all index's pools are going to be cleaned, so we skip the release here
         mtSessionThreadInfo->set_gc_session(GetCurrentGcSession());
+        GcEpochType local_epoch =
+            GetSessionManager()->GetCurrentSessionContext()->GetTxnManager()->GetGcSession()->GcStartInnerTxn();
+
         size_t allocationSize = (*static_cast<mrcu_callback*>(gcRemoveLayerFuncObjPtr))(dropIndex);
+
         if (dropIndex == false) {
             ((SlabAllocator*)slab)->Release(gcRemoveLayerFuncObjPtr, allocationSize);
         }
+
+        GetSessionManager()->GetCurrentSessionContext()->GetTxnManager()->GetGcSession()->GcEndInnerTxn(false);
         mtSessionThreadInfo->set_gc_session(NULL);
         return allocationSize;
     }

src/gausskernel/storage/mot/core/storage/table.cpp

@@ -582,6 +582,7 @@ Row* Table::RemoveKeyFromIndex(Row* row, Sentinel* sentinel, uint64_t tid, GcMan
 #endif
         currSentinel = ix->IndexRemove(&key, tid);
         MOT_ASSERT(currSentinel == sentinel);
+        MOT_ASSERT(currSentinel->GetCounter() == 0);
         if (likely(gc != nullptr)) {
             if (ix->GetIndexOrder() == IndexOrder::INDEX_ORDER_PRIMARY) {
                 OutputRow = currSentinel->GetData();

src/gausskernel/storage/mot/core/system/transaction/txn.cpp

@@ -413,6 +413,8 @@ RC TxnManager::RollbackInsert(Access* ac)
 #endif
         outputSen = index_->IndexRemove(&m_key, GetThdId());
         MOT_ASSERT(outputSen != nullptr);
+        MOT_ASSERT(outputSen->GetCounter() == 0);
+
         GcSessionRecordRcu(index_->GetIndexId(), outputSen, nullptr, Index::SentinelDtor, SENTINEL_SIZE(index_));
         // If we are the owner of the key and insert on top of a deleted row,
         // lets check if we can reclaim the deleted row
@@ -820,6 +822,8 @@ void TxnInsertAction::CleanupOptimisticInsert(
                 MOT_ASSERT(pIndexInsertResult->GetCounter() == 0);
                 Sentinel* outputSen = index_->IndexRemove(currentItem->m_key, m_manager->GetThdId());
                 MOT_ASSERT(outputSen != nullptr);
+                MOT_ASSERT(outputSen->GetCounter() == 0);
+
                 m_manager->GcSessionRecordRcu(
                     index_->GetIndexId(), outputSen, nullptr, Index::SentinelDtor, SENTINEL_SIZE(index_));
                 m_manager->m_accessMgr->IncreaseTableStat(table);

src/gausskernel/storage/replication/libpqwalreceiver.cpp

@@ -255,8 +255,14 @@ void StartRemoteStreaming(const LibpqrcvConnectParam *options)
             stringlist_to_identifierstr(t_thrd.libwalreceiver_cxt.streamConn, options->publicationNames);
         appendStringInfo(&cmd, ", publication_names %s",
             PQescapeLiteral(t_thrd.libwalreceiver_cxt.streamConn, pubnames_str, strlen(pubnames_str)));
-        appendStringInfoChar(&cmd, ')');
         pfree(pubnames_str);
+
+        if (options->binary && PQserverVersion(t_thrd.libwalreceiver_cxt.streamConn) >= 90204) {
+            appendStringInfoString(&cmd, ", binary 'true'");
+            ereport(DEBUG5, (errmsg("append binary true")));
+        }
+
+        appendStringInfoChar(&cmd, ')');
     }
 
     PGresult *res = libpqrcv_PQexec(cmd.data);
@@ -420,7 +426,7 @@ void IdentifyRemoteSystem(bool checkRemote)
 }
 
 /* identify remote mode, should do this after connect success. */
-static ServerMode IdentifyRemoteMode()
+ServerMode IdentifyRemoteMode()
 {
     Assert(t_thrd.libwalreceiver_cxt.streamConn != NULL);
     volatile WalRcvData *walrcv = t_thrd.walreceiverfuncs_cxt.WalRcv;
@@ -443,7 +449,9 @@ static ServerMode IdentifyRemoteMode()
                                   num_fields)));
     }
     remoteMode = (ServerMode)pg_strtoint32(PQgetvalue(res, 0, 0));
-    if (!t_thrd.walreceiver_cxt.AmWalReceiverForFailover && (!IS_PRIMARY_NORMAL(remoteMode)) &&
+    if (walrcv->conn_target != REPCONNTARGET_PUBLICATION &&
+        !t_thrd.walreceiver_cxt.AmWalReceiverForFailover &&
+        (!IS_PRIMARY_NORMAL(remoteMode)) &&
         /* remoteMode of cascade standby is a standby */
         !t_thrd.xlog_cxt.is_cascade_standby && !IS_SHARED_STORAGE_MODE) {
         PQclear(res);

src/gausskernel/storage/replication/logical/decode.cpp

@@ -223,6 +223,7 @@ void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *
         return;
     }
 
+    ResourceOwner tmpOwner = t_thrd.utils_cxt.CurrentResourceOwner;
     /* cast so we get a warning when new rmgrs are added */
     switch ((RmgrIds)XLogRecGetRmid(record)) {
         /*
@@ -257,6 +258,7 @@ void LogicalDecodingProcessRecord(LogicalDecodingContext *ctx, XLogReaderState *
         default:
             break;
     }
+    t_thrd.utils_cxt.CurrentResourceOwner = tmpOwner;
 }
 
 /*

src/gausskernel/storage/replication/logical/proto.cpp

@@ -18,7 +18,6 @@
 #include "catalog/pg_type.h"
 #include "libpq/pqformat.h"
 #include "replication/logicalproto.h"
-#include "utils/builtins.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -28,7 +27,7 @@
 static const int LOGICALREP_IS_REPLICA_IDENTITY = 1;
 
 static void logicalrep_write_attrs(StringInfo out, Relation rel);
-static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple);
+static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple, bool binary);
 
 static void logicalrep_read_attrs(StringInfo in, LogicalRepRelation *rel);
 static void logicalrep_read_tuple(StringInfo in, LogicalRepTupleData *tuple);
@@ -115,7 +114,7 @@ void logicalrep_write_origin(StringInfo out, const char *origin, XLogRecPtr orig
 /*
  * Write INSERT to the output stream.
  */
-void logicalrep_write_insert(StringInfo out, Relation rel, HeapTuple newtuple)
+void logicalrep_write_insert(StringInfo out, Relation rel, HeapTuple newtuple, bool binary)
 {
     pq_sendbyte(out, 'I'); /* action INSERT */
 
@@ -123,7 +122,7 @@ void logicalrep_write_insert(StringInfo out, Relation rel, HeapTuple newtuple)
     pq_sendint32(out, RelationGetRelid(rel));
 
     pq_sendbyte(out, 'N'); /* new tuple follows */
-    logicalrep_write_tuple(out, rel, newtuple);
+    logicalrep_write_tuple(out, rel, newtuple, binary);
 }
 
 /*
@@ -151,7 +150,7 @@ LogicalRepRelId logicalrep_read_insert(StringInfo in, LogicalRepTupleData *newtu
 /*
  * Write UPDATE to the output stream.
  */
-void logicalrep_write_update(StringInfo out, Relation rel, HeapTuple oldtuple, HeapTuple newtuple)
+void logicalrep_write_update(StringInfo out, Relation rel, HeapTuple oldtuple, HeapTuple newtuple, bool binary)
 {
     pq_sendbyte(out, 'U'); /* action UPDATE */
 
@@ -166,11 +165,11 @@ void logicalrep_write_update(StringInfo out, Relation rel, HeapTuple oldtuple, H
             pq_sendbyte(out, 'O'); /* old tuple follows */
         else
             pq_sendbyte(out, 'K'); /* old key follows */
-        logicalrep_write_tuple(out, rel, oldtuple);
+        logicalrep_write_tuple(out, rel, oldtuple, binary);
     }
 
     pq_sendbyte(out, 'N'); /* new tuple follows */
-    logicalrep_write_tuple(out, rel, newtuple);
+    logicalrep_write_tuple(out, rel, newtuple, binary);
 }
 
 /*
@@ -213,7 +212,7 @@ LogicalRepRelId logicalrep_read_update(StringInfo in, bool *has_oldtuple, Logica
 /*
  * Write DELETE to the output stream.
  */
-void logicalrep_write_delete(StringInfo out, Relation rel, HeapTuple oldtuple)
+void logicalrep_write_delete(StringInfo out, Relation rel, HeapTuple oldtuple, bool binary)
 {
     char relreplident = RelationGetRelReplident(rel);
     Assert(relreplident == REPLICA_IDENTITY_DEFAULT ||
@@ -229,7 +228,7 @@ void logicalrep_write_delete(StringInfo out, Relation rel, HeapTuple oldtuple)
     else
         pq_sendbyte(out, 'K'); /* old key follows */
 
-    logicalrep_write_tuple(out, rel, oldtuple);
+    logicalrep_write_tuple(out, rel, oldtuple, binary);
 }
 
 /*
@@ -344,7 +343,7 @@ void logicalrep_read_typ(StringInfo in, LogicalRepTyp *ltyp)
 /*
  * Write a tuple to the outputstream, in the most efficient format possible.
  */
-static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple)
+static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple, bool binary)
 {
     TupleDesc desc;
     Datum values[MaxTupleAttributeNumber];
@@ -371,7 +370,6 @@ static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple
         HeapTuple typtup;
         Form_pg_type typclass;
         Form_pg_attribute att = desc->attrs[i];
-        char *outputstr;
 
         /* skip dropped columns */
         if (att->attisdropped || GetGeneratedCol(desc, i)) {
@@ -379,7 +377,7 @@ static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple
         }
 
         if (isnull[i]) {
-            pq_sendbyte(out, 'n'); /* null column */
+            pq_sendbyte(out, LOGICALREP_COLUMN_NULL); /* null column */
             continue;
         }
 
@@ -388,61 +386,91 @@ static void logicalrep_write_tuple(StringInfo out, Relation rel, HeapTuple tuple
             elog(ERROR, "cache lookup failed for type %u", att->atttypid);
         typclass = (Form_pg_type)GETSTRUCT(typtup);
 
-        pq_sendbyte(out, 't'); /* 'text' data follows */
-        if (!typclass->typbyval && typclass->typlen == -1) {
-            /* definitely detoasted Datum */
-            Datum val = PointerGetDatum(PG_DETOAST_DATUM(values[i]));
-            outputstr = OidOutputFunctionCall(typclass->typoutput, val);
+        /*
+        * Send in binary if requested and type has suitable send function.
+        */
+        if (binary && OidIsValid(typclass->typsend)) {
+            bytea* outputbytes = NULL;
+            pq_sendbyte(out, LOGICALREP_COLUMN_BINARY);
+            if (!typclass->typbyval && typclass->typlen == -1) {
+                /* definitely detoasted Datum */
+                Datum val = PointerGetDatum(PG_DETOAST_DATUM(values[i]));
+                outputbytes = OidSendFunctionCall(typclass->typsend, val);
+            } else {
+                outputbytes = OidSendFunctionCall(typclass->typsend, values[i]);
+            }
+            int len = VARSIZE(outputbytes) - VARHDRSZ;
+            pq_sendint(out, len, 4);    /* length */
+            pq_sendbytes(out, VARDATA(outputbytes), len);    /* data */
+            if (outputbytes != NULL) {
+                pfree(outputbytes);
+            }
         } else {
-            outputstr = OidOutputFunctionCall(typclass->typoutput, values[i]);
+            char* outputstr = NULL;
+            pq_sendbyte(out, LOGICALREP_COLUMN_TEXT);
+            if (!typclass->typbyval && typclass->typlen == -1) {
+                /* definitely detoasted Datum */
+                Datum val = PointerGetDatum(PG_DETOAST_DATUM(values[i]));
+                outputstr = OidOutputFunctionCall(typclass->typoutput, val);
+            } else {
+                outputstr = OidOutputFunctionCall(typclass->typoutput, values[i]);
+            }
+            pq_sendcountedtext(out, outputstr, strlen(outputstr), false);
+            if (outputstr != NULL) {
+                pfree(outputstr);
+            }
         }
-        pq_sendcountedtext(out, outputstr, strlen(outputstr), false);
-        pfree(outputstr);
         ReleaseSysCache(typtup);
     }
 }
 
 /*
- * Read tuple in remote format from stream.
- *
- * The returned tuple points into the input stringinfo.
+ * Read tuple in logical replication format from stream.
  */
 static void logicalrep_read_tuple(StringInfo in, LogicalRepTupleData *tuple)
 {
-    uint16 i;
-    uint16 natts;
-    int rc;
-
     /* Get number of attributes. */
-    natts = pq_getmsgint(in, sizeof(uint16));
+    uint16 natts = pq_getmsgint(in, sizeof(uint16));
 
-    rc = memset_s(tuple->changed, sizeof(tuple->changed), 0, sizeof(tuple->changed));
-    securec_check(rc, "", "");
+    /* Allocate space for per-column values; zero out unused StringInfoDatas */
+    tuple->colvalues = (StringInfoData *) palloc0(natts * sizeof(StringInfoData));
+    tuple->colstatus = (char *) palloc(natts * sizeof(char));
     tuple->ncols = natts;
 
     /* Read the data */
-    for (i = 0; i < natts; i++) {
-        char kind;
-
-        kind = pq_getmsgbyte(in);
+    for (uint16 i = 0; i < natts; i++) {
+        char kind = pq_getmsgbyte(in);
+        tuple->colstatus[i] = kind;
+        uint32 len;
+        StringInfo value = &tuple->colvalues[i];
 
         switch (kind) {
-            case 'n': /* null */
-                tuple->values[i] = NULL;
-                tuple->changed[i] = true;
+            case LOGICALREP_COLUMN_NULL: /* null */
+                /* nothing more to do */
                 break;
-            case 't': { /* text formatted value */
-                uint32 len;
-                tuple->changed[i] = true;
-
-                len = pq_getmsgint(in, sizeof(uint32)); /* read length */
+            case LOGICALREP_COLUMN_TEXT:
+                len = pq_getmsgint(in, sizeof(uint32));    /* read length */
 
                 /* and data */
-                tuple->values[i] = (char *)palloc(len + 1);
-                pq_copymsgbytes(in, tuple->values[i], len);
-                tuple->values[i][len] = '\0';
+                value->data = (char *) palloc((len + 1) * sizeof(char));
+                pq_copymsgbytes(in, value->data, len);
+                value->data[len] = '\0';
+                /* make StringInfo fully valid */
+                value->len = len;
+                value->cursor = 0;
+                value->maxlen = len;
+                break;
+            case LOGICALREP_COLUMN_BINARY:
+                len = pq_getmsgint(in, sizeof(uint32));    /* read length */
+
+                /* and data */
+                value->data = (char *)palloc0((len + 1) * sizeof(char));
+                pq_copymsgbytes(in, value->data, len);
+                /* make StringInfo fully valid */
+                value->len = len;
+                value->cursor = 0;
+                value->maxlen = len;
                 break;
-            }
             default:
                 elog(ERROR, "unrecognized data representation type '%c'", kind);
                 break;
@@ -451,7 +479,7 @@ static void logicalrep_read_tuple(StringInfo in, LogicalRepTupleData *tuple)
 }
 
 /*
- * Write relation attributes to the stream.
+ * Write relation attributes metadata to the stream.
  */
 static void logicalrep_write_attrs(StringInfo out, Relation rel)
 {
@@ -504,7 +532,7 @@ static void logicalrep_write_attrs(StringInfo out, Relation rel)
 }
 
 /*
- * Read relation attribute names from the stream.
+ * Read relation attribute metadata from the stream.
  */
 static void logicalrep_read_attrs(StringInfo in, LogicalRepRelation *rel)
 {
@@ -573,3 +601,25 @@ static const char *logicalrep_read_namespace(StringInfo in)
 
     return nspname;
 }
+
+/*
+ * Write conninfo to the output stream.
+ */
+void logicalrep_write_conninfo(StringInfo out, char* conninfo)
+{
+    pq_sendbyte(out, 'S'); /* action */
+
+    pq_writestring(out, conninfo); /* conninfo follows */
+}
+
+/*
+ * Read conninfo from stream.
+ */
+void logicalrep_read_conninfo(StringInfo in, char** conninfo)
+{
+    const char* conninfoTemp = pq_getmsgstring(in);
+    size_t conninfoLen = strlen(conninfoTemp) + 1;
+    *conninfo = (char*)palloc(conninfoLen);
+    int rc = strcpy_s(*conninfo, conninfoLen, conninfoTemp);
+    securec_check(rc, "", "");
+}

src/gausskernel/storage/replication/logical/worker.cpp

@@ -41,6 +41,7 @@
 #include "catalog/pg_partition_fn.h"
 
 #include "commands/trigger.h"
+#include "commands/subscriptioncmds.h"
 
 #include "executor/executor.h"
 #include "executor/node/nodeModifyTable.h"
@@ -111,6 +112,8 @@ static void store_flush_position(XLogRecPtr remote_lsn);
 static void reread_subscription(void);
 static void ApplyWorkerProcessMsg(char type, StringInfo s, XLogRecPtr *lastRcv);
 static void apply_dispatch(StringInfo s);
+static void apply_handle_conninfo(StringInfo s);
+static void UpdateConninfo(char* standbysInfo);
 
 /* SIGHUP: set flag to re-read config file at next convenient time */
 static void LogicalrepWorkerSighub(SIGNAL_ARGS)
@@ -118,18 +121,6 @@ static void LogicalrepWorkerSighub(SIGNAL_ARGS)
     t_thrd.applyworker_cxt.got_SIGHUP = true;
 }
 
-/* SIGTERM: time to die */
-static void LogicalrepWorkerSigterm(SIGNAL_ARGS)
-{
-    int saveErrno = errno;
-
-    t_thrd.applyworker_cxt.got_SIGTERM = true;
-    if (t_thrd.proc)
-        SetLatch(&t_thrd.proc->procLatch);
-
-    errno = saveErrno;
-}
-
 /*
  * Make sure that we started local transaction.
  *
@@ -265,11 +256,11 @@ static void slot_store_error_callback(void *arg)
 }
 
 /*
- * Store data in C string form into slot.
- * This is similar to BuildTupleFromCStrings but TupleTableSlot fits our
- * use better.
+ * Store tuple data into slot.
+ *
+ * Incoming data can be either text or binary format.
  */
-static void slot_store_cstrings(TupleTableSlot *slot, LogicalRepRelMapEntry *rel, char **values)
+static void slot_store_data(TupleTableSlot *slot, LogicalRepRelMapEntry *rel, LogicalRepTupleData *tupleData)
 {
     int natts = slot->tts_tupleDescriptor->natts;
     int i;
@@ -286,19 +277,52 @@ static void slot_store_cstrings(TupleTableSlot *slot, LogicalRepRelMapEntry *rel
     errcallback.previous = t_thrd.log_cxt.error_context_stack;
     t_thrd.log_cxt.error_context_stack = &errcallback;
 
-    /* Call the "in" function for each non-dropped attribute */
+    /* Call the "in" function for each non-dropped, non-null attribute */
     for (i = 0; i < natts; i++) {
         Form_pg_attribute att = slot->tts_tupleDescriptor->attrs[i];
         int remoteattnum = rel->attrmap[i];
-        Oid typinput;
-        Oid typioparam;
 
-        if (!att->attisdropped && remoteattnum >= 0 && values[remoteattnum] != NULL) {
+        if (!att->attisdropped && remoteattnum >= 0) {
+            StringInfo colvalue = &tupleData->colvalues[remoteattnum];
             errarg.remote_attnum = remoteattnum;
 
-            getTypeInputInfo(att->atttypid, &typinput, &typioparam);
-            slot->tts_values[i] = OidInputFunctionCall(typinput, values[remoteattnum], typioparam, att->atttypmod);
-            slot->tts_isnull[i] = false;
+            if (tupleData->colstatus[remoteattnum] == LOGICALREP_COLUMN_TEXT) {
+                Oid typinput;
+                Oid typioparam;
+
+                getTypeInputInfo(att->atttypid, &typinput, &typioparam);
+                slot->tts_values[i] = OidInputFunctionCall(typinput, colvalue->data, typioparam, att->atttypmod);
+                slot->tts_isnull[i] = false;
+            } else if (tupleData->colstatus[remoteattnum] == LOGICALREP_COLUMN_BINARY) {
+                Oid typreceive;
+                Oid typioparam;
+
+                /*
+                 * In some code paths we may be asked to re-parse the same
+                 * tuple data.  Reset the StringInfo's cursor so that works.
+                 */
+                colvalue->cursor = 0;
+
+                getTypeBinaryInputInfo(att->atttypid, &typreceive, &typioparam);
+                slot->tts_values[i] = OidReceiveFunctionCall(typreceive, colvalue, typioparam, att->atttypmod);
+
+                /* Trouble if it didn't eat the whole buffer */
+                if (colvalue->cursor != colvalue->len) {
+                    ereport(ERROR, (errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
+                        errmsg("incorrect binary data format in logical replication column %d", 
+                        remoteattnum + 1)));
+                }
+                slot->tts_isnull[i] = false;
+            } else {
+                /*
+                 * NULL value from remote.  (We don't expect to see
+                 * LOGICALREP_COLUMN_UNCHANGED here, but if we do, treat it as
+                 * NULL.)
+                 */
+                slot->tts_values[i] = (Datum) 0;
+                slot->tts_isnull[i] = true;
+            }
+            /* Reset attnum for error callback */
             errarg.remote_attnum = -1;
         } else {
             /*
@@ -318,18 +342,19 @@ static void slot_store_cstrings(TupleTableSlot *slot, LogicalRepRelMapEntry *rel
 }
 
 /*
- * Replace selected columns with user data provided as C strings.
+ * Replace updated columns with data from the LogicalRepTupleData struct.
  * This is somewhat similar to heap_modify_tuple but also calls the type
  * input functions on the user data.
- * "slot" is filled with a copy of the tuple in "srcslot", with
- * columns selected by the "replaces" array replaced with data values
- * from "values".
+ *
+ * "slot" is filled with a copy of the tuple in "srcslot", replacing
+ * columns provided in "tupleData" and leaving others as-is.
+ *
  * Caution: unreplaced pass-by-ref columns in "slot" will point into the
  * storage for "srcslot".  This is OK for current usage, but someday we may
  * need to materialize "slot" at the end to make it independent of "srcslot".
  */
-static void slot_modify_cstrings(TupleTableSlot *slot, TupleTableSlot *srcslot, LogicalRepRelMapEntry *rel,
-    char **values, const bool *replaces)
+static void slot_modify_data(TupleTableSlot *slot, TupleTableSlot *srcslot, LogicalRepRelMapEntry *rel,
+    LogicalRepTupleData *tupleData)
 {
     int natts = slot->tts_tupleDescriptor->natts;
     int i;
@@ -364,23 +389,47 @@ static void slot_modify_cstrings(TupleTableSlot *slot, TupleTableSlot *srcslot,
         Form_pg_attribute att = slot->tts_tupleDescriptor->attrs[i];
         int remoteattnum = rel->attrmap[i];
 
-        if (remoteattnum < 0 || !replaces[remoteattnum]) {
+        if (remoteattnum < 0) {
             continue;
         }
 
-        if (values[remoteattnum] != NULL) {
-            Oid typinput;
-            Oid typioparam;
-
+        if (tupleData->colstatus[remoteattnum] != LOGICALREP_COLUMN_UNCHANGED) {
+            StringInfo colvalue = &tupleData->colvalues[remoteattnum];
             errarg.remote_attnum = remoteattnum;
 
-            getTypeInputInfo(att->atttypid, &typinput, &typioparam);
-            slot->tts_values[i] = OidInputFunctionCall(typinput, values[remoteattnum], typioparam, att->atttypmod);
-            slot->tts_isnull[i] = false;
+            if (tupleData->colstatus[remoteattnum] == LOGICALREP_COLUMN_TEXT) {
+                Oid typinput;
+                Oid typioparam;
+
+                getTypeInputInfo(att->atttypid, &typinput, &typioparam);
+                slot->tts_values[i] = OidInputFunctionCall(typinput, colvalue->data, typioparam, att->atttypmod);
+                slot->tts_isnull[i] = false;
+            } else if (tupleData->colstatus[remoteattnum] == LOGICALREP_COLUMN_BINARY) {
+                Oid typreceive;
+                Oid typioparam;
+
+                /*
+                 * In some code paths we may be asked to re-parse the same
+                 * tuple data.  Reset the StringInfo's cursor so that works.
+                 */
+                colvalue->cursor = 0;
+
+                getTypeBinaryInputInfo(att->atttypid, &typreceive, &typioparam);
+                slot->tts_values[i] = OidReceiveFunctionCall(typreceive, colvalue, typioparam, att->atttypmod);
+
+                /* Trouble if it didn't eat the whole buffer */
+                if (colvalue->cursor != colvalue->len) {
+                    ereport(ERROR, (errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
+                        errmsg("incorrect binary data format in logical replication column %d", remoteattnum + 1)));
+                }
+                slot->tts_isnull[i] = false;
+            } else {
+                /* must be LOGICALREP_COLUMN_NULL */
+                slot->tts_values[i] = (Datum) 0;
+                slot->tts_isnull[i] = true;
+            }
+            
             errarg.remote_attnum = -1;
-        } else {
-            slot->tts_values[i] = (Datum)0;
-            slot->tts_isnull[i] = true;
         }
     }
 
@@ -516,7 +565,7 @@ static void apply_handle_insert(StringInfo s)
     PushActiveSnapshot(GetTransactionSnapshot());
     /* Process and store remote tuple in the slot */
     oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
-    slot_store_cstrings(remoteslot, rel, newtup.values);
+    slot_store_data(remoteslot, rel, &newtup);
     slot_fill_defaults(rel, estate, remoteslot);
     MemoryContextSwitchTo(oldctx);
 
@@ -646,7 +695,7 @@ static void apply_handle_update(StringInfo s)
         int remoteattnum = rel->attrmap[i];
         if (!att->attisdropped && remoteattnum >= 0) {
             Assert(remoteattnum < newtup.ncols);
-            if (newtup.changed[i]) {
+            if (newtup.colstatus[i] != LOGICALREP_COLUMN_UNCHANGED) {
                 target_rte->updatedCols = bms_add_member(target_rte->updatedCols,
                                                      i + 1 - FirstLowInvalidHeapAttributeNumber);
             }
@@ -660,7 +709,7 @@ static void apply_handle_update(StringInfo s)
 
     /* Build the search tuple. */
     oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
-    slot_store_cstrings(remoteslot, rel, has_oldtup ? oldtup.values : newtup.values);
+    slot_store_data(remoteslot, rel, has_oldtup ? &oldtup : &newtup);
     MemoryContextSwitchTo(oldctx);
 
     /*
@@ -683,7 +732,7 @@ static void apply_handle_update(StringInfo s)
     if (found) {
         /* Process and store remote tuple in the slot */
         oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
-        slot_modify_cstrings(remoteslot, localslot, rel, newtup.values, newtup.changed);
+        slot_modify_data(remoteslot, localslot, rel, &newtup);
         MemoryContextSwitchTo(oldctx);
 
         EvalPlanQualSetSlot(&epqstate, remoteslot);
@@ -748,7 +797,7 @@ static void apply_handle_delete(StringInfo s)
 
     /* Find the tuple using the replica identity index. */
     oldctx = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
-    slot_store_cstrings(remoteslot, rel, oldtup.values);
+    slot_store_data(remoteslot, rel, &oldtup);
     MemoryContextSwitchTo(oldctx);
 
     /*
@@ -824,6 +873,9 @@ static void apply_dispatch(StringInfo s)
         case 'O':
             apply_handle_origin(s);
             break;
+        case 'S':
+            apply_handle_conninfo(s);
+            break;
         default:
             ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION),
                 errmsg("invalid logical replication message type \"%c\"", action)));
@@ -1004,13 +1056,15 @@ static void ApplyLoop(void)
     /* mark as idle, before starting to loop */
     pgstat_report_activity(STATE_IDLE, NULL);
 
-    while (!t_thrd.applyworker_cxt.got_SIGTERM) {
+    for (;;) {
         MemoryContextSwitchTo(t_thrd.applyworker_cxt.messageContext);
 
         int len;
         char *buf = NULL;
         unsigned char type;
 
+        CHECK_FOR_INTERRUPTS();
+
         /* Wait a while for data to arrive */
         if ((WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_receive(NAPTIME_PER_CYCLE, &type, &buf, &len)) {
             StringInfoData s;
@@ -1224,6 +1278,17 @@ static void reread_subscription(void)
         proc_exit(0);
     }
 
+    /*
+    * Exit if any parameter that affects the remote connection was changed.
+    * The launcher will start a new worker.
+    */
+    if (strcmp(newsub->name, t_thrd.applyworker_cxt.mySubscription->name) != 0 ||
+        newsub->binary != t_thrd.applyworker_cxt.mySubscription->binary) {
+            ereport(LOG, (errmsg("logical replication apply worker for subscription \"%s\" "
+                "will restart because of a parameter change", t_thrd.applyworker_cxt.mySubscription->name)));
+            proc_exit(0);
+    }
+
     /* !slotname should never happen when enabled is true. */
     Assert(newsub->slotname);
     
@@ -1296,7 +1361,7 @@ void ApplyWorkerMain()
      */
     gspqsignal(SIGHUP, LogicalrepWorkerSighub);
     gspqsignal(SIGINT, StatementCancelHandler);
-    gspqsignal(SIGTERM, LogicalrepWorkerSigterm);
+    gspqsignal(SIGTERM, die);
 
     gspqsignal(SIGQUIT, quickdie);
     gspqsignal(SIGALRM, handle_sig_alarm);
@@ -1422,14 +1487,9 @@ void ApplyWorkerMain()
 
     CommitTransactionCommand();
 
-    char *decryptConnInfo = DecryptConninfo(t_thrd.applyworker_cxt.mySubscription->conninfo);
-    bool connectSuccess = (WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_connect(decryptConnInfo, NULL,
-        t_thrd.applyworker_cxt.mySubscription->name, -1);
-    rc = memset_s(decryptConnInfo, strlen(decryptConnInfo), 0, strlen(decryptConnInfo));
-    securec_check(rc, "", "");
-    pfree_ext(decryptConnInfo);
-    if (!connectSuccess) {
-        ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), errmsg("could not connect to the publisher")));
+    if (!AttemptConnectPublisher(t_thrd.applyworker_cxt.mySubscription->conninfo,
+        t_thrd.applyworker_cxt.mySubscription->name, true)) {
+        ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), errmsg("Failed to connect to publisher.")));
     }
 
     /*
@@ -1447,6 +1507,7 @@ void ApplyWorkerMain()
     options.slotname = t_thrd.applyworker_cxt.mySubscription->slotname;
     options.protoVersion = LOGICALREP_PROTO_VERSION_NUM;
     options.publicationNames = t_thrd.applyworker_cxt.mySubscription->publications;
+    options.binary = t_thrd.applyworker_cxt.mySubscription->binary;
 
     /* Start streaming from the slot. */
     (WalReceiverFuncTable[GET_FUNC_IDX]).walrcv_startstreaming(&options);
@@ -1595,3 +1656,72 @@ char* DefListToString(const List *defList)
     return buf.data;
 }
 
+/*
+ * Handle conninfo update message.
+ */
+static void apply_handle_conninfo(StringInfo s)
+{
+    char* standbysInfo = NULL;
+    logicalrep_read_conninfo(s, &standbysInfo);
+    UpdateConninfo(standbysInfo);
+    pfree_ext(standbysInfo);
+}
+
+static void UpdateConninfo(char* standbysInfo)
+{
+    Relation rel;
+    bool nulls[Natts_pg_subscription];
+    bool replaces[Natts_pg_subscription];
+    Datum values[Natts_pg_subscription];
+    HeapTuple tup;
+    Subscription* sub = t_thrd.applyworker_cxt.mySubscription;
+    Oid subid = sub->oid;
+
+    StartTransactionCommand();
+    rel = heap_open(SubscriptionRelationId, RowExclusiveLock);
+    /* Fetch the existing tuple. */
+    tup = SearchSysCacheCopy2(SUBSCRIPTIONNAME, u_sess->proc_cxt.MyDatabaseId,
+        CStringGetDatum(t_thrd.applyworker_cxt.mySubscription->name));
+    if (!HeapTupleIsValid(tup)) {
+        ereport(ERROR, (errcode(ERRCODE_UNDEFINED_OBJECT), errmsg("subscription \"%s\" does not exist",
+            t_thrd.applyworker_cxt.mySubscription->name)));
+    }
+    subid = HeapTupleGetOid(tup);
+
+    /* Form a new tuple. */
+    int rc = memset_s(nulls, sizeof(nulls), false, sizeof(nulls));
+    securec_check(rc, "", "");
+    rc = memset_s(values, sizeof(values), 0, sizeof(values));
+    securec_check(rc, "", "");
+    rc = memset_s(replaces, sizeof(replaces), false, sizeof(replaces));
+    securec_check(rc, "", "");
+
+    /* get conninfoWithoutHostport */
+    StringInfoData conninfoWithoutHostport;
+    initStringInfo(&conninfoWithoutHostport);
+    ParseConninfo(sub->conninfo, &conninfoWithoutHostport, (HostPort**)NULL);
+
+    /* join conninfoWithoutHostport together with standbysinfo */
+    appendStringInfo(&conninfoWithoutHostport, " %s", standbysInfo);
+    /* Replace connection information */
+    values[Anum_pg_subscription_subconninfo - 1] = CStringGetTextDatum(conninfoWithoutHostport.data);
+    replaces[Anum_pg_subscription_subconninfo - 1] = true;
+    tup = heap_modify_tuple(tup, RelationGetDescr(rel), values, nulls, replaces);
+
+    /* Update the catalog. */
+    simple_heap_update(rel, &tup->t_self, tup);
+    CatalogUpdateIndexes(rel, tup);
+
+    heap_close(rel, RowExclusiveLock);
+    CommitTransactionCommand();
+
+    ereport(LOG, (errmsg("Update conninfo successfully, new conninfo %s.", standbysInfo)));
+}
+
+/*
+ * Is current process a logical replication worker?
+ */
+bool IsLogicalWorker(void)
+{
+    return t_thrd.applyworker_cxt.curWorker != NULL;
+}

src/gausskernel/storage/replication/pgoutput/pgoutput.cpp

@@ -15,6 +15,8 @@
 
 #include "catalog/pg_publication.h"
 
+#include "commands/defrem.h"
+
 #include "replication/logical.h"
 #include "replication/logicalproto.h"
 #include "replication/origin.h"
@@ -44,6 +46,8 @@ static bool pgoutput_origin_filter(LogicalDecodingContext *ctx, RepOriginId orig
 
 static List *LoadPublications(List *pubnames);
 static void publication_invalidation_cb(Datum arg, int cacheid, uint32 hashvalue);
+static bool ReplconninfoChanged();
+static void GetConninfo(StringInfoData* standbysInfo);
 
 /* Entry in the map used to remember which relation schemas we sent. */
 typedef struct RelationSyncEntry {
@@ -79,6 +83,9 @@ static void parse_output_parameters(List *options, PGOutputData *data)
     ListCell *lc;
     bool protocol_version_given = false;
     bool publication_names_given = false;
+    bool binary_option_given = false;
+
+    data->binary = false;
 
     foreach (lc, options) {
         DefElem *defel = (DefElem *)lfirst(lc);
@@ -108,6 +115,12 @@ static void parse_output_parameters(List *options, PGOutputData *data)
 
             if (!SplitIdentifierString(strVal(defel->arg), ',', &(data->publication_names)))
                 ereport(ERROR, (errcode(ERRCODE_INVALID_NAME), errmsg("invalid publication_names syntax")));
+        } else if (strcmp(defel->defname, "binary") == 0) {
+            if (binary_option_given)
+                ereport(ERROR, (errcode(ERRCODE_SYNTAX_ERROR), errmsg("conflicting or redundant options")));
+            binary_option_given = true;
+
+            data->binary = defGetBoolean(defel);
         } else
             elog(ERROR, "unrecognized pgoutput option: %s", defel->defname);
     }
@@ -206,6 +219,22 @@ static void pgoutput_commit_txn(LogicalDecodingContext *ctx, ReorderBufferTXN *t
     OutputPluginPrepareWrite(ctx, true);
     logicalrep_write_commit(ctx->out, txn, commit_lsn);
     OutputPluginWrite(ctx, true);
+
+    /*
+     * Send the newest connecttion information to the subscriber,
+     * when the connection information about the standby changes.
+     */
+    if (ReplconninfoChanged()) {
+        StringInfoData standbysInfo;
+        initStringInfo(&standbysInfo);
+
+        GetConninfo(&standbysInfo);
+        OutputPluginPrepareWrite(ctx, true);
+        logicalrep_write_conninfo(ctx->out, standbysInfo.data);
+        OutputPluginWrite(ctx, true);
+
+        FreeStringInfo(&standbysInfo);
+    }
 }
 
 /*
@@ -300,19 +329,19 @@ static void pgoutput_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
     switch (change->action) {
         case REORDER_BUFFER_CHANGE_INSERT:
             OutputPluginPrepareWrite(ctx, true);
-            logicalrep_write_insert(ctx->out, relation, &change->data.tp.newtuple->tuple);
+            logicalrep_write_insert(ctx->out, relation, &change->data.tp.newtuple->tuple, data->binary);
             OutputPluginWrite(ctx, true);
             break;
         case REORDER_BUFFER_CHANGE_UINSERT:
             OutputPluginPrepareWrite(ctx, true);
-            logicalrep_write_insert(ctx->out, relation, (HeapTuple)(&change->data.utp.newtuple->tuple));
+            logicalrep_write_insert(ctx->out, relation, (HeapTuple)(&change->data.utp.newtuple->tuple), data->binary);
             OutputPluginWrite(ctx, true);
             break;
         case REORDER_BUFFER_CHANGE_UPDATE: {
             HeapTuple oldtuple = change->data.tp.oldtuple ? &change->data.tp.oldtuple->tuple : NULL;
 
             OutputPluginPrepareWrite(ctx, true);
-            logicalrep_write_update(ctx->out, relation, oldtuple, &change->data.tp.newtuple->tuple);
+            logicalrep_write_update(ctx->out, relation, oldtuple, &change->data.tp.newtuple->tuple, data->binary);
             OutputPluginWrite(ctx, true);
             break;
         }
@@ -320,14 +349,14 @@ static void pgoutput_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
             HeapTuple oldtuple = change->data.utp.oldtuple ? ((HeapTuple)(&change->data.utp.oldtuple->tuple)) : NULL;
 
             OutputPluginPrepareWrite(ctx, true);
-            logicalrep_write_update(ctx->out, relation, oldtuple, (HeapTuple)(&change->data.utp.newtuple->tuple));
+            logicalrep_write_update(ctx->out, relation, oldtuple, (HeapTuple)(&change->data.utp.newtuple->tuple), data->binary);
             OutputPluginWrite(ctx, true);
             break;
         }
         case REORDER_BUFFER_CHANGE_DELETE:
             if (change->data.tp.oldtuple) {
                 OutputPluginPrepareWrite(ctx, true);
-                logicalrep_write_delete(ctx->out, relation, &change->data.tp.oldtuple->tuple);
+                logicalrep_write_delete(ctx->out, relation, &change->data.tp.oldtuple->tuple, data->binary);
                 OutputPluginWrite(ctx, true);
             } else
                 elog(DEBUG1, "didn't send DELETE change because of missing oldtuple");
@@ -335,7 +364,7 @@ static void pgoutput_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
         case REORDER_BUFFER_CHANGE_UDELETE:
             if (change->data.utp.oldtuple) {
                 OutputPluginPrepareWrite(ctx, true);
-                logicalrep_write_delete(ctx->out, relation, (HeapTuple)(&change->data.utp.oldtuple->tuple));
+                logicalrep_write_delete(ctx->out, relation, (HeapTuple)(&change->data.utp.oldtuple->tuple), data->binary);
                 OutputPluginWrite(ctx, true);
             } else
                 elog(DEBUG1, "didn't send DELETE change because of missing oldtuple");
@@ -605,3 +634,43 @@ static void rel_sync_cache_publication_cb(Datum arg, int cacheid, uint32 hashval
         entry->pubactions.pubdelete = false;
     }
 }
+
+static void GetConninfo(StringInfoData* standbysInfo)
+{
+    bool primaryJoined = false;
+    StringInfoData hosts;
+    StringInfoData ports;
+    initStringInfo(&hosts);
+    initStringInfo(&ports);
+    for (int i = 1; i < MAX_REPLNODE_NUM + 1; ++i) {
+        t_thrd.postmaster_cxt.ReplConnChangeType[i] = 0;
+        if (t_thrd.postmaster_cxt.ReplConnArray[i] == NULL) {
+            continue;
+        }
+        if (!primaryJoined) {
+            appendStringInfo(&hosts, "%s,%s",
+                t_thrd.postmaster_cxt.ReplConnArray[i]->localhost,
+                t_thrd.postmaster_cxt.ReplConnArray[i]->remotehost);
+            appendStringInfo(&ports, "%d,%d",
+                t_thrd.postmaster_cxt.ReplConnArray[i]->localport,
+                t_thrd.postmaster_cxt.ReplConnArray[i]->remoteport);
+            primaryJoined = true;
+        } else {
+            appendStringInfo(&hosts, ",%s",
+                t_thrd.postmaster_cxt.ReplConnArray[i]->remotehost);
+            appendStringInfo(&ports, ",%d",
+                t_thrd.postmaster_cxt.ReplConnArray[i]->remoteport);
+        }
+    }
+    appendStringInfo(standbysInfo, "host=%s port=%s", hosts.data, ports.data);
+}
+
+static inline bool ReplconninfoChanged()
+{
+    for (int i = 1; i < MAX_REPLNODE_NUM; ++i) {
+        if (t_thrd.postmaster_cxt.ReplConnChangeType[i]) {
+            return true;
+        }
+    }
+    return false;
+}

src/gausskernel/storage/replication/slot.cpp

@@ -1934,14 +1934,18 @@ ArchiveConfig* formArchiveConfigFromStr(char *content, bool encrypted)
     ArchiveConnConfig* conn_config = NULL;
     char* media_type = NULL;
     errno_t rc = EOK;
-    /* SplitIdentifierString will change origin string */
-    char *content_copy = pstrdup(content);
+    char *content_copy = NULL;
     char *tmp = NULL;
     List* elemlist = NIL;
     int param_num = 0;
     size_t elem_index = 0;
     archive_config = (ArchiveConfig *)palloc0(sizeof(ArchiveConfig));
     char decryptSecretAccessKeyStr[DEST_CIPHER_LENGTH] = {'\0'};
+    if (content == NULL || strlen(content) == 0) {
+        goto FAILURE;
+    }
+    /* SplitIdentifierString will change origin string */
+    content_copy = pstrdup(content);
     /* Parse string into list of identifiers */
     if (!SplitIdentifierString(content_copy, ';', &elemlist, false, false)) {
         goto FAILURE;

src/gausskernel/storage/replication/walsender.cpp

@@ -401,6 +401,7 @@ int WalSenderMain(void)
         t_thrd.walsender_cxt.server_run_mode = t_thrd.postmaster_cxt.HaShmData->current_mode;
 
     SetHaWalSenderChannel();
+    Assert(t_thrd.utils_cxt.CurrentResourceOwner != NULL);
     /* Handle handshake messages before streaming */
     WalSndHandshake();
 
@@ -496,6 +497,7 @@ static void WalSndHandshake(void)
 
     while (!repCxt.replicationStarted) {
         int firstchar;
+        Assert(t_thrd.utils_cxt.CurrentResourceOwner != NULL);
 
         WalSndSetState(WALSNDSTATE_STARTUP);
         set_ps_display("idle", false);
@@ -2033,6 +2035,8 @@ bool isLogicalSlotExist(char* slotName)
 }
 
 static void IdentifyCommand(Node* cmd_node, ReplicationCxt* repCxt, const char *cmd_string){
+    ResourceOwner tmpOwner = t_thrd.utils_cxt.CurrentResourceOwner;
+    Assert(tmpOwner != NULL);
     switch (cmd_node->type) {
         case T_IdentifySystemCmd:
             IdentifySystem();
@@ -2140,7 +2144,7 @@ static void IdentifyCommand(Node* cmd_node, ReplicationCxt* repCxt, const char *
             ereport(FATAL,
                     (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("invalid standby query string: %s", cmd_string)));
     }
-
+    t_thrd.utils_cxt.CurrentResourceOwner = tmpOwner;
 }
 
 /*
@@ -3596,8 +3600,12 @@ static int WalSndLoop(WalSndSendDataCallback send_data)
     t_thrd.walsender_cxt.last_logical_xlog_advanced_timestamp = GetCurrentTimestamp();
     t_thrd.walsender_cxt.waiting_for_ping_response = false;
 
+    ResourceOwner tmpOwner = t_thrd.utils_cxt.CurrentResourceOwner;
+    Assert(!IsTransactionOrTransactionBlock() &&
+        strcmp(ResourceOwnerGetName(tmpOwner), "walsender top-level resource owner") == 0);
     /* Loop forever, unless we get an error */
     for (;;) {
+        t_thrd.utils_cxt.CurrentResourceOwner = tmpOwner;
         TimestampTz now;
 
         /* Clear any already-pending wakeups */
@@ -4006,6 +4014,7 @@ static int WalSndLoop(WalSndSendDataCallback send_data)
         }
     }
 
+    t_thrd.utils_cxt.CurrentResourceOwner = tmpOwner;
     WalSndShutdown();
     return 1; /* keep the compiler quiet */
 }

src/gausskernel/storage/smgr/md.cpp

@@ -394,8 +394,6 @@ void mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
     }
 
     if (fd < 0) {
-        int save_errno = errno;
-
         /*
          * During bootstrap, there are cases where a system relation will be
          * accessed (by internal backend processes) before the bootstrap
@@ -407,26 +405,7 @@ void mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
          * new catalogs can by no means be used by other relations, we simply
          * truncate them.
          */
-        if (isRedo || IsBootstrapProcessingMode() ||
-            (u_sess->attr.attr_common.IsInplaceUpgrade && filenode.rnode.node.relNode < FirstNormalObjectId)) {
-            ADIO_RUN()
-            {
-                flags = O_RDWR | PG_BINARY | O_DIRECT | (u_sess->attr.attr_common.IsInplaceUpgrade ? O_TRUNC : 0);
-            }
-            ADIO_ELSE()
-            {
-                flags = O_RDWR | PG_BINARY | (u_sess->attr.attr_common.IsInplaceUpgrade ? O_TRUNC : 0);
-            }
-            ADIO_END();
-
-            fd = DataFileIdOpenFile(path, filenode, flags, 0600);
-        }
-
-        if (fd < 0) {
-            /* be sure to report the error reported by create, not open */
-            errno = save_errno;
-            ereport(ERROR, (errcode_for_file_access(), errmsg("could not create file \"%s\": %m", path)));
-        }
+        fd = RetryDataFileIdOpenFile(isRedo, path, filenode, flags);
     }
 
     File fd_pca = -1;

src/get_PlatForm_str.sh

@@ -145,6 +145,29 @@ then
     plat_form_str=ubuntu18.04_"$cpu_bit"
 fi
 
+##################################################################################
+# the modification here will also lead to the synchronous modification of the 3rd-party compilation library path
+# PR link : https://gitee.com/opengauss/openGauss-third_party/pulls/130
+# redflag platform  
+# the result form like this: asianux_x86_64
+##################################################################################
+if [ "$kernel"x = "redflag"x ]
+then
+    plat_form_str=asianux7.6_"$cpu_bit"
+fi
+
+
+##################################################################################
+# the modification here will also lead to the synchronous modification of the 3rd-party compilation library path
+# PR link : https://gitee.com/opengauss/openGauss-third_party/pulls/130
+# asianux platform
+# the result form like this: asianux_aarch64
+##################################################################################
+if [ "$kernel"x = "asianux"x ]
+then
+    plat_form_str=asianux7.5_"$cpu_bit"
+fi
+
 ##################################################################################
 #
 # other platform 

src/include/access/ustore/knl_utuple.h

@@ -314,4 +314,5 @@ template<bool hasnull>
 void UHeapFillDiskTuple(TupleDesc tupleDesc, Datum *values, const bool *isnull, UHeapDiskTupleData *diskTuple,
     uint32 dataSize, bool enableReverseBitmap, bool enableReserve);
 void CheckTupleValidity(Relation rel, UHeapTuple utuple);
+Tuple UHeapMaterialize(TupleTableSlot *slot);
 #endif

src/include/catalog/heap.h

@@ -165,7 +165,7 @@ extern void addNewPartitionTuple(Relation pg_part_desc, Partition new_part_desc,
     Datum interval, Datum maxValues,  Datum transitionPoint, Datum reloptions);
 
 extern void heap_truncate_one_part(Relation rel , Oid partOid);
-extern Oid heapTupleGetPartitionId(Relation rel, void *tuple);
+extern Oid heapTupleGetPartitionId(Relation rel, void *tuple, bool isDDL = false);
 extern Oid heapTupleGetSubPartitionId(Relation rel, void *tuple);
 extern void heap_truncate(List *relids);
 extern void heap_truncate_one_rel(Relation rel);
@@ -240,7 +240,7 @@ extern char* make_column_map(TupleDesc tuple_desc);
  */
 extern bool* CheckPartkeyHasTimestampwithzone(Relation partTableRel, bool isForSubPartition = false);
 
-extern Oid AddNewIntervalPartition(Relation rel, void* insertTuple);
+extern Oid AddNewIntervalPartition(Relation rel, void* insertTuple, bool isDDL = false);
 
 extern int GetIndexKeyAttsByTuple(Relation relation, HeapTuple indexTuple);
 

src/include/catalog/pg_subscription.h

@@ -50,13 +50,15 @@ CATALOG(pg_subscription,6126) BKI_SHARED_RELATION BKI_ROWTYPE_OID(6128) BKI_SCHE
     NameData subslotname;    /* Slot name on publisher */
     text subsynccommit;      /* Synchronous commit setting for worker */
     text subpublications[1]; /* List of publications subscribed to */
+    bool subbinary;          /* True if the subscription wants the
+                              * publisher to send data in binary */
 #endif
 }
 FormData_pg_subscription;
 
 typedef FormData_pg_subscription *Form_pg_subscription;
 
-#define Natts_pg_subscription 8
+#define Natts_pg_subscription 9
 #define Anum_pg_subscription_subdbid 1
 #define Anum_pg_subscription_subname 2
 #define Anum_pg_subscription_subowner 3
@@ -65,6 +67,7 @@ typedef FormData_pg_subscription *Form_pg_subscription;
 #define Anum_pg_subscription_subslotname 6
 #define Anum_pg_subscription_subsynccommit 7
 #define Anum_pg_subscription_subpublications 8
+#define Anum_pg_subscription_subbinary 9
 
 
 typedef struct Subscription {
@@ -77,6 +80,7 @@ typedef struct Subscription {
     char *slotname;     /* Name of the replication slot */
     char *synccommit;   /* Synchronous commit setting for worker */
     List *publications; /* List of publication names to subscribe to */
+    bool binary;        /* Indicates if the subscription wants data in binary format */
 } Subscription;
 
 
@@ -86,6 +90,7 @@ extern Oid get_subscription_oid(const char *subname, bool missing_ok);
 extern char *get_subscription_name(Oid subid, bool missing_ok);
 
 extern int CountDBSubscriptions(Oid dbid);
-extern char *DecryptConninfo(char *encryptConninfo);
+extern void ClearListContent(List *list);
+
 
 #endif /* PG_SUBSCRIPTION_H */

src/include/commands/subscriptioncmds.h

@@ -17,6 +17,11 @@
 
 #include "nodes/parsenodes.h"
 
+typedef struct HostPort {
+    char* host;
+    char* port;
+} HostPort;
+
 extern ObjectAddress CreateSubscription(CreateSubscriptionStmt *stmt, bool isTopLevel);
 extern ObjectAddress AlterSubscription(AlterSubscriptionStmt *stmt);
 extern void DropSubscription(DropSubscriptionStmt *stmt, bool isTopLevel);
@@ -24,6 +29,12 @@ extern void DropSubscription(DropSubscriptionStmt *stmt, bool isTopLevel);
 extern ObjectAddress AlterSubscriptionOwner(const char *name, Oid newOwnerId);
 extern void AlterSubscriptionOwner_oid(Oid subid, Oid newOwnerId);
 extern void RenameSubscription(List* oldname, const char* newname);
+extern void AddStandbysInfo(char* standbysInfo);
+extern void DropStandbysInfo(char* standbysInfo);
+
+extern void ParseConninfo(const char* conninfo, StringInfoData* conninfoWithoutHostPort, HostPort** hostPortList);
+extern char* EncryptOrDecryptConninfo(const char* conninfo, const char action);
+extern bool AttemptConnectPublisher(const char *conninfoOriginal, char* slotname, bool checkRemoteMode);
 
 #endif							/* SUBSCRIPTIONCMDS_H */
 

src/include/knl/knl_guc/knl_session_attr_sql.h

@@ -238,7 +238,7 @@ typedef struct knl_session_attr_sql {
 #ifndef ENABLE_MULTIPLE_NODES
     bool enable_custom_parser;
 #endif
-    bool b_sql_plugin;
+    bool dolphin;
 } knl_session_attr_sql;
 
 #endif /* SRC_INCLUDE_KNL_KNL_SESSION_ATTR_SQL */

src/include/miscadmin.h

@@ -90,6 +90,7 @@ extern const uint32 SUPPORT_DATA_REPAIR;
 extern const uint32 SCAN_BATCH_MODE_VERSION_NUM;
 extern const uint32 RELMAP_4K_VERSION_NUM;
 extern const uint32 PUBLICATION_VERSION_NUM;
+extern const uint32 SUBSCRIPTION_BINARY_VERSION_NUM;
 extern const uint32 ANALYZER_HOOK_VERSION_NUM;
 extern const uint32 SUPPORT_HASH_XLOG_VERSION_NUM;
 extern const uint32 PITR_INIT_VERSION_NUM;

src/include/replication/libpqwalreceiver.h

@@ -35,6 +35,7 @@ typedef struct LibpqrcvConnectParam {
     bool logical;
     uint32 protoVersion;    /* Logical protocol version */
     List *publicationNames; /* String list of publications */
+    bool binary;            /* Ask publisher to use binary */
 }LibpqrcvConnectParam;
 
 extern int32 pg_atoi(char* s, int size, int c);
@@ -53,5 +54,6 @@ extern bool libpqrcv_command(const char *cmd, char **err, int *sqlstate);
 extern void IdentifyRemoteSystem(bool checkRemote);
 extern void CreateRemoteReplicationSlot(XLogRecPtr startpoint, const char* slotname, bool isLogical);
 extern void StartRemoteStreaming(const LibpqrcvConnectParam *options);
+extern ServerMode IdentifyRemoteMode();
 
 #endif

src/include/replication/logical.h

@@ -31,6 +31,7 @@
 #include "access/extreme_rto/spsc_blocking_queue.h"
 #include "access/parallel_recovery/redo_item.h"
 
+#include "nodes/parsenodes_common.h"
 #include "nodes/replnodes.h"
 #include "access/ustore/knl_utuple.h"
 #include "replication/logical_queue.h"

src/include/replication/logicalproto.h

@@ -35,12 +35,21 @@
  * Keep in mind that the columns correspond to the *remote* table.
  */
 typedef struct LogicalRepTupleData {
-    char *values[MaxTupleAttributeNumber]; /* value in out function format or NULL if values is NULL */
-    bool changed[MaxTupleAttributeNumber]; /* marker for changed/unchanged values */
+    /* Array of StringInfos, one per column; some may be unused */
+    StringInfoData *colvalues;
+    /* Array of markers for null/unchanged/text/binary, one per column */
+    char       *colstatus;
     /* Length of above arrays */
     int ncols;
 } LogicalRepTupleData;
 
+/* Possible values for LogicalRepTupleData.colstatus[colnum] */
+/* These values are also used in the on-the-wire protocol */
+#define LOGICALREP_COLUMN_NULL        'n'
+#define LOGICALREP_COLUMN_UNCHANGED    'u'
+#define LOGICALREP_COLUMN_TEXT        't'
+#define LOGICALREP_COLUMN_BINARY    'b' /* added in PG14 */
+
 typedef uint32 LogicalRepRelId;
 
 /* Relation information */
@@ -81,16 +90,18 @@ extern void logicalrep_read_begin(StringInfo in, LogicalRepBeginData *begin_data
 extern void logicalrep_write_commit(StringInfo out, ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 extern void logicalrep_read_commit(StringInfo in, LogicalRepCommitData *commit_data);
 extern void logicalrep_write_origin(StringInfo out, const char *origin, XLogRecPtr origin_lsn);
-extern void logicalrep_write_insert(StringInfo out, Relation rel, HeapTuple newtuple);
+extern void logicalrep_write_insert(StringInfo out, Relation rel, HeapTuple newtuple, bool binary);
 extern LogicalRepRelId logicalrep_read_insert(StringInfo in, LogicalRepTupleData *newtup);
-extern void logicalrep_write_update(StringInfo out, Relation rel, HeapTuple oldtuple, HeapTuple newtuple);
+extern void logicalrep_write_update(StringInfo out, Relation rel, HeapTuple oldtuple, HeapTuple newtuple, bool binary);
 extern LogicalRepRelId logicalrep_read_update(StringInfo in, bool *has_oldtuple, LogicalRepTupleData *oldtup,
     LogicalRepTupleData *newtup);
-extern void logicalrep_write_delete(StringInfo out, Relation rel, HeapTuple oldtuple);
+extern void logicalrep_write_delete(StringInfo out, Relation rel, HeapTuple oldtuple, bool binary);
 extern LogicalRepRelId logicalrep_read_delete(StringInfo in, LogicalRepTupleData *oldtup);
 extern void logicalrep_write_rel(StringInfo out, Relation rel);
 extern LogicalRepRelation *logicalrep_read_rel(StringInfo in);
 extern void logicalrep_write_typ(StringInfo out, Oid typoid);
 extern void logicalrep_read_typ(StringInfo out, LogicalRepTyp *ltyp);
+extern void logicalrep_write_conninfo(StringInfo out, char* conninfo);
+extern void logicalrep_read_conninfo(StringInfo in, char** conninfo);
 
 #endif /* LOGICALREP_PROTO_H */

src/include/replication/logicalworker.h

@@ -13,5 +13,6 @@
 #define LOGICALWORKER_H
 
 extern void ApplyWorkerMain();
+extern bool IsLogicalWorker(void);
 
 #endif /* LOGICALWORKER_H */

src/include/replication/pgoutput.h

@@ -24,6 +24,7 @@ typedef struct PGOutputData {
 
     List *publication_names;
     List *publications;
+    bool binary;
 } PGOutputData;
 
 #endif /* PGOUTPUT_H */

src/include/utils/knl_globalsysdbcache.h

@@ -229,10 +229,15 @@ private:
     bool m_syscache_relids[FirstNormalObjectId];
 
 };
+void NotifyGscRecoveryStarted();
 void NotifyGscRecoveryFinished();
 extern Datum gs_gsc_dbstat_info(PG_FUNCTION_ARGS);
 extern Datum gs_gsc_clean(PG_FUNCTION_ARGS);
 extern Datum gs_gsc_catalog_detail(PG_FUNCTION_ARGS);
 extern Datum gs_gsc_table_detail(PG_FUNCTION_ARGS);
 extern int ResizeHashBucket(int origin_nbucket, DynamicHashBucketStrategy strategy);
-#endif
+
+/* only primary node check recovery, recovery from upgrade may have misorder cacheids */
+#define IsPrimaryRecoveryFinished() \
+    (g_instance.global_sysdbcache.recovery_finished || t_thrd.postmaster_cxt.HaShmData->current_mode == STANDBY_MODE)
+#endif

src/include/utils/lsyscache.h

@@ -213,7 +213,7 @@ extern bool check_rel_is_partitioned(Oid relid);
 extern Oid partid_get_parentid(Oid partid);
 extern bool is_not_strict_agg(Oid funcOid);
 extern bool is_pgxc_class_table(Oid tableoid);
-extern Oid get_valid_relname_relid(const char* relnamespace, const char* relname);
+extern Oid get_valid_relname_relid(const char* relnamespace, const char* relname, bool nsp_missing_ok = false);
 extern bool get_func_iswindow(Oid funcid);
 extern char get_func_prokind(Oid funcid);
 extern char get_typecategory(Oid typid);

src/include/utils/plpgsql.h

@@ -1385,6 +1385,8 @@ typedef struct PlDebugEntry {
     PLpgSQL_function* func;
 } PlDebugEntry;
 
+typedef List* (*RawParserHook)(const char*, List**);
+
 const int MAXINT8LEN = 25;
 const int DEFAULT_DEBUG_BUF_SIZE = 1024;
 const long NANOSECOND_PER_SECOND = 1000000000L; /* 1s */

src/test/regress/expected/composite_datum_record.out

@@ -0,0 +1,37 @@
+--
+-- check for over-optimization of whole-row Var referencing an Append plan
+--
+create table int4_test(f1 int4);
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6 where f1<=0) q) from int4_test;
+ q 
+---
+(0 rows)
+
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6.0 where f1<=0) q) from int4_test;
+ q 
+---
+(0 rows)
+
+insert into int4_test(f1) values(' 0 ');
+insert into int4_test(f1) values('123456 ');
+select * from int4_test;
+   f1   
+--------
+      0
+ 123456
+(2 rows)
+
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6 where f1<=0) q) from int4_test;
+    q    
+---------
+ (4,5,6)
+ (1,2,3)
+(2 rows)
+
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6.0 where f1<=0) q) from int4_test;
+     q     
+-----------
+ (4,5,6.0)
+ (1,2,3)
+(2 rows)
+

src/test/regress/expected/deserialize_func.out

@@ -0,0 +1,116 @@
+create schema deserialize_func1;
+set current_schema = deserialize_func1;
+create sequence test_seq;
+create or replace function test_func() return int
+as
+begin
+return 1;
+end;
+/
+create table t1(key int, id int default nextval('test_seq'));
+create table t2(key int, id int default test_func());
+select * from pg_get_tabledef('deserialize_func1.t1');
+                   pg_get_tabledef                    
+------------------------------------------------------
+ SET search_path = deserialize_func1;                +
+ CREATE TABLE t1 (                                   +
+     key integer,                                    +
+     id integer DEFAULT nextval('test_seq'::regclass)+
+ )                                                   +
+ WITH (orientation=row, compression=no);
+(1 row)
+
+select * from pg_get_tabledef('deserialize_func1.t2');
+             pg_get_tabledef             
+-----------------------------------------
+ SET search_path = deserialize_func1;   +
+ CREATE TABLE t2 (                      +
+     key integer,                       +
+     id integer DEFAULT test_func()     +
+ )                                      +
+ WITH (orientation=row, compression=no);
+(1 row)
+
+alter schema deserialize_func1 rename to deserialize_func2;
+select * from pg_get_tabledef('deserialize_func2.t1');
+                            pg_get_tabledef                             
+------------------------------------------------------------------------
+ SET search_path = deserialize_func2;                                  +
+ CREATE TABLE t1 (                                                     +
+     key integer,                                                      +
+     id integer DEFAULT nextval('deserialize_func2.test_seq'::regclass)+
+ )                                                                     +
+ WITH (orientation=row, compression=no);
+(1 row)
+
+select * from pg_get_tabledef('deserialize_func2.t2');
+                   pg_get_tabledef                    
+------------------------------------------------------
+ SET search_path = deserialize_func2;                +
+ CREATE TABLE t2 (                                   +
+     key integer,                                    +
+     id integer DEFAULT deserialize_func2.test_func()+
+ )                                                   +
+ WITH (orientation=row, compression=no);
+(1 row)
+
+alter table deserialize_func2.t1 alter column id drop default;
+alter table deserialize_func2.t1 alter column id set default nextval('deserialize_func2.test_seq');
+select * from pg_get_tabledef('deserialize_func2.t1');
+                            pg_get_tabledef                             
+------------------------------------------------------------------------
+ SET search_path = deserialize_func2;                                  +
+ CREATE TABLE t1 (                                                     +
+     key integer,                                                      +
+     id integer DEFAULT nextval('deserialize_func2.test_seq'::regclass)+
+ )                                                                     +
+ WITH (orientation=row, compression=no);
+(1 row)
+
+alter table deserialize_func2.t2 alter column id drop default;
+alter table deserialize_func2.t2 alter column id set default deserialize_func2.test_func();
+select * from pg_get_tabledef('deserialize_func2.t2');
+                   pg_get_tabledef                    
+------------------------------------------------------
+ SET search_path = deserialize_func2;                +
+ CREATE TABLE t2 (                                   +
+     key integer,                                    +
+     id integer DEFAULT deserialize_func2.test_func()+
+ )                                                   +
+ WITH (orientation=row, compression=no);
+(1 row)
+
+set current_schema = deserialize_func2;
+create type type1 as (a int);
+create view v1 as select '(1)'::type1;
+alter type type1 rename to type2;
+select pg_get_viewdef('v1');
+        pg_get_viewdef         
+-------------------------------
+ SELECT '(1)'::type2 AS type1;
+(1 row)
+
+drop view v1 cascade;
+create table atest12 as select x as a, 10001-x as b from generate_series(1, 10000) x;
+create function func1(integer, integer) returns boolean as $$begin return $1 < $2; end $$ language plpgsql immutable;
+create operator <<< (procedure=func1, leftarg=integer, rightarg=integer, restrict=scalarltsel);
+create view atest12v as select * from atest12 where b <<< 5;
+select pg_get_viewdef('atest12v');
+                 pg_get_viewdef                  
+-------------------------------------------------
+ SELECT  * FROM atest12 WHERE (atest12.b <<< 5);
+(1 row)
+
+drop schema if exists deserialize_func1 cascade;
+NOTICE:  schema "deserialize_func1" does not exist, skipping
+drop schema if exists deserialize_func2 cascade;
+NOTICE:  drop cascades to 9 other objects
+DETAIL:  drop cascades to sequence test_seq
+drop cascades to function test_func()
+drop cascades to table t1
+drop cascades to table t2
+drop cascades to type type2
+drop cascades to table atest12
+drop cascades to function func1(integer,integer)
+drop cascades to operator <<<(integer,integer)
+drop cascades to view atest12v

src/test/regress/expected/drop_if_exists.out

@@ -132,11 +132,14 @@ DROP EXTENSION test_extension_exists;
 ERROR:  extension "test_extension_exists" does not exist
 DROP EXTENSION IF EXISTS test_extension_exists;
 NOTICE:  extension "test_extension_exists" does not exist, skipping
--- exists but doesn't support drop
-DROP EXTENSION plpgsql;
-ERROR:  EXTENSION is not yet supported.
-DROP EXTENSION IF EXISTS plpgsql;
-ERROR:  EXTENSION is not yet supported.
+-- only extension in white list support drop, but the guc support_extended_features is true in pg_regress mode, so hdfs_fdw can be droped
+CREATE DATABASE ext_test_db;
+\c ext_test_db
+DROP EXTENSION hdfs_fdw;
+DROP EXTENSION IF EXISTS hdfs_fdw;
+NOTICE:  extension "hdfs_fdw" does not exist, skipping
+\c regression
+DROP DATABASE ext_test_db;
 -- functions
 DROP FUNCTION test_function_exists();
 ERROR:  function test_function_exists does not exist

src/test/regress/expected/hw_partition_interval_exchange.out

@@ -215,3 +215,107 @@ select * from interval_normal_exchange where logdate > '2020-06-01' order by log
 (0 rows)
 
 drop table interval_normal_exchange;
+drop table table_001;
+ERROR:  table "table_001" does not exist
+create table table_001(
+    COL_1 smallint,
+    COL_2 char(5),
+    COL_3 int,
+    COL_4 date,
+    COL_5 boolean,
+    COL_6 nchar(5),
+    COL_7 float
+    );
+drop table partition_table_001;
+ERROR:  table "partition_table_001" does not exist
+create table partition_table_001(
+    COL_1 smallint,
+    COL_2 char(5),
+    COL_3 int,
+    COL_4 date,
+    COL_5 boolean,
+    COL_6 nchar(5),
+    COL_7 float
+    )
+    PARTITION BY RANGE (COL_4)
+    INTERVAL ('1 month')
+    (
+        PARTITION partition_table_001_p1 VALUES LESS THAN ('2020-03-01'),
+        PARTITION partition_table_001_p2 VALUES LESS THAN ('2020-04-01'),
+        PARTITION partition_table_001_p3 VALUES LESS THAN ('2020-05-01')
+    );
+insert into partition_table_001 values (1,'aaa',1,'2020-02-23',true,'aaa',1.1);
+insert into partition_table_001 values (2,'bbb',2,'2020-03-23',false,'bbb',2.2);
+insert into partition_table_001 values (3,'ccc',3,'2020-04-23',true,'ccc',3.3);
+insert into partition_table_001 values (4,'ddd',4,'2020-05-23',false,'ddd',4.4);
+insert into partition_table_001 values (5,'eee',5,'2020-06-23',true,'eee',5.5);
+insert into partition_table_001 values (6,'fff',6,'2020-07-23',false,'fff',6.6);
+ALTER TABLE partition_table_001 EXCHANGE PARTITION (sys_p1) WITH TABLE table_001;
+select * from table_001 order by 1;
+ col_1 | col_2 | col_3 |          col_4           | col_5 | col_6 | col_7 
+-------+-------+-------+--------------------------+-------+-------+-------
+     4 | ddd   |     4 | Sat May 23 00:00:00 2020 | f     | ddd   |   4.4
+(1 row)
+
+select * from partition_table_001 order by 1;
+ col_1 | col_2 | col_3 |          col_4           | col_5 | col_6 | col_7 
+-------+-------+-------+--------------------------+-------+-------+-------
+     1 | aaa   |     1 | Sun Feb 23 00:00:00 2020 | t     | aaa   |   1.1
+     2 | bbb   |     2 | Mon Mar 23 00:00:00 2020 | f     | bbb   |   2.2
+     3 | ccc   |     3 | Thu Apr 23 00:00:00 2020 | t     | ccc   |   3.3
+     5 | eee   |     5 | Tue Jun 23 00:00:00 2020 | t     | eee   |   5.5
+     6 | fff   |     6 | Thu Jul 23 00:00:00 2020 | f     | fff   |   6.6
+(5 rows)
+
+select relname, parttype, partstrategy, boundaries from pg_partition
+                where parentid = (select oid from pg_class where relname = 'partition_table_001')
+                order by relname;
+        relname         | parttype | partstrategy |          boundaries          
+------------------------+----------+--------------+------------------------------
+ partition_table_001    | r        | i            | 
+ partition_table_001_p1 | p        | r            | {2020-03-01}
+ partition_table_001_p2 | p        | r            | {2020-04-01}
+ partition_table_001_p3 | p        | r            | {2020-05-01}
+ sys_p1                 | p        | i            | {"Mon Jun 01 00:00:00 2020"}
+ sys_p2                 | p        | i            | {"Wed Jul 01 00:00:00 2020"}
+ sys_p3                 | p        | i            | {"Sat Aug 01 00:00:00 2020"}
+(7 rows)
+
+ALTER TABLE partition_table_001 EXCHANGE PARTITION (sys_p1) WITH TABLE table_001;
+select * from table_001 order by 1;
+ col_1 | col_2 | col_3 | col_4 | col_5 | col_6 | col_7 
+-------+-------+-------+-------+-------+-------+-------
+(0 rows)
+
+select * from partition_table_001 order by 1;
+ col_1 | col_2 | col_3 |          col_4           | col_5 | col_6 | col_7 
+-------+-------+-------+--------------------------+-------+-------+-------
+     1 | aaa   |     1 | Sun Feb 23 00:00:00 2020 | t     | aaa   |   1.1
+     2 | bbb   |     2 | Mon Mar 23 00:00:00 2020 | f     | bbb   |   2.2
+     3 | ccc   |     3 | Thu Apr 23 00:00:00 2020 | t     | ccc   |   3.3
+     4 | ddd   |     4 | Sat May 23 00:00:00 2020 | f     | ddd   |   4.4
+     5 | eee   |     5 | Tue Jun 23 00:00:00 2020 | t     | eee   |   5.5
+     6 | fff   |     6 | Thu Jul 23 00:00:00 2020 | f     | fff   |   6.6
+(6 rows)
+
+insert into table_001 values (7,'eee',7,'2020-08-23',true,'eee',7.7);
+ALTER TABLE partition_table_001 EXCHANGE PARTITION (sys_p1) WITH TABLE table_001 with validation verbose;
+select * from table_001 order by 1;
+ col_1 | col_2 | col_3 |          col_4           | col_5 | col_6 | col_7 
+-------+-------+-------+--------------------------+-------+-------+-------
+     4 | ddd   |     4 | Sat May 23 00:00:00 2020 | f     | ddd   |   4.4
+(1 row)
+
+select * from partition_table_001 order by 1;
+ col_1 | col_2 | col_3 |          col_4           | col_5 | col_6 | col_7 
+-------+-------+-------+--------------------------+-------+-------+-------
+     1 | aaa   |     1 | Sun Feb 23 00:00:00 2020 | t     | aaa   |   1.1
+     2 | bbb   |     2 | Mon Mar 23 00:00:00 2020 | f     | bbb   |   2.2
+     3 | ccc   |     3 | Thu Apr 23 00:00:00 2020 | t     | ccc   |   3.3
+     5 | eee   |     5 | Tue Jun 23 00:00:00 2020 | t     | eee   |   5.5
+     6 | fff   |     6 | Thu Jul 23 00:00:00 2020 | f     | fff   |   6.6
+     7 | eee   |     7 | Sun Aug 23 00:00:00 2020 | t     | eee   |   7.7
+(6 rows)
+
+drop table table_001;
+drop table partition_table_001;

src/test/regress/expected/join_test_alias.out

@@ -0,0 +1,55 @@
+CREATE TABLE t11 (
+sec_code character(6) NOT NULL,
+issue_type character(3) NOT NULL
+);
+CREATE TABLE t22 (
+company_code character(6) NOT NULL,
+list_profit_flag character(1) NOT NULL
+);
+CREATE TABLE t33 (
+company_code character(6) NOT NULL,
+issue_flag character(6) NOT NULL
+);
+insert into t11 values(1,'S04');
+insert into t22 values(1,'Y');
+insert into t33 values(1,'1');
+insert into t33 values(2,'2');
+select
+T5.issue_type
+,T6.is_type
+from
+(
+select T1.sec_code
+,issue_type
+--,case when list_profit_flag='Y' then '是' else '否' end as list_profit_flag
+from
+(
+select sec_code
+,case when issue_type in ('S04','S05','S06','S09') then '增发'
+when issue_type in ('S01','S02') then '首发'
+else '其他' end as issue_type
+from t11
+)T1
+left join
+(
+select list_profit_flag
+,company_code
+from t22
+)T2
+on T1.sec_code=T2.company_code
+)T5
+full join
+(
+SELECT CASE
+WHEN issue_flag = ANY (ARRAY['1'::bpchar, '3'::bpchar, '11'::bpchar]) THEN '首发'::text
+ELSE '增发'::text
+END AS is_type from t33
+)T6
+on T5.issue_type=T6.is_type
+;
+ issue_type | is_type 
+------------+---------
+ 增发       | 增发
+            | 首发
+(2 rows)
+

src/test/regress/expected/row_compression/unsupported_feature.out

@@ -89,5 +89,8 @@ ERROR:  Can not use compress option in ustore index.
 -- segment
 CREATE TABLE unspported_feature.segment_table(id int, c1 text) WITH(compresstype=2, segment=on);  --failed
 ERROR:  only row orientation table support compresstype.
-CREATE INDEX on unspported_feature.index_test(c1) WITH(compresstype=2, segment=on); --faled
+CREATE INDEX on unspported_feature.index_test(c1) WITH(compresstype=2, segment=on); --failed
 ERROR:  Can not use compress option in segment storage.
+-- set compress_diff_convert
+create table unspported_feature.compress_byte_test(id int) with (compresstype=2, compress_byte_convert=false, compress_diff_convert = true); -- failed
+ERROR:  compress_diff_convert should be used with compress_byte_convert.

src/test/regress/expected/truncate_gpi.out

@@ -1062,6 +1062,73 @@ select /*+ indexscan(tg_hash i_tg_hash_local_a) */ * from tg_hash where a < 9;
 (81 rows)
 
 drop table tg_hash;
+drop table if exists test_ugi_045;
+NOTICE:  table "test_ugi_045" does not exist, skipping
+create table test_ugi_045
+(
+ c_id integer not null,
+ c_date DATE,
+ c_info varchar(20) not null
+)
+partition by range(c_date)
+interval('2 day')
+(
+ partition p1 values less than ('2021-06-01 00:00:00'),
+ partition p2 values less than ('2021-06-03 00:00:00'),
+ partition p3 values less than ('2021-06-05 00:00:00'),
+ partition p4 values less than ('2021-06-07 00:00:00'),
+ partition p5 values less than ('2021-06-09 00:00:00')
+);
+insert into test_ugi_045(c_id, c_date, c_info) values(1, '2021-05-30 00:00:00', '1-1');
+insert into test_ugi_045(c_id, c_date, c_info) values(2, '2021-05-31 00:00:00', '1-1');
+insert into test_ugi_045(c_id, c_date, c_info) values(3, '2021-06-01 00:00:00', '1-1');
+insert into test_ugi_045(c_id, c_date, c_info) values(4, '2021-06-02 00:00:00', '1-2');
+insert into test_ugi_045(c_id, c_date, c_info) values(5, '2021-06-03 00:00:00', '1-2');
+insert into test_ugi_045(c_id, c_date, c_info) values(6, '2021-06-04 00:00:00', '1-3');
+insert into test_ugi_045(c_id, c_date, c_info) values(7, '2021-06-05 00:00:00', '1-3');
+insert into test_ugi_045(c_id, c_date, c_info) values(8, '2021-06-06 00:00:00', '1-4');
+insert into test_ugi_045(c_id, c_date, c_info) values(9, '2021-06-07 00:00:00', '1-4');
+insert into test_ugi_045(c_id, c_date, c_info) values(10, '2021-06-08 00:00:00', '1-5');
+create index global_index_id_045 on test_ugi_045(c_date) global;
+create index global_index_info_045 on test_ugi_045(c_info) global;
+select *from test_ugi_045;
+ c_id |       c_date        | c_info 
+------+---------------------+--------
+    1 | 2021-05-30 00:00:00 | 1-1
+    2 | 2021-05-31 00:00:00 | 1-1
+    3 | 2021-06-01 00:00:00 | 1-1
+    4 | 2021-06-02 00:00:00 | 1-2
+    5 | 2021-06-03 00:00:00 | 1-2
+    6 | 2021-06-04 00:00:00 | 1-3
+    7 | 2021-06-05 00:00:00 | 1-3
+    8 | 2021-06-06 00:00:00 | 1-4
+    9 | 2021-06-07 00:00:00 | 1-4
+   10 | 2021-06-08 00:00:00 | 1-5
+(10 rows)
+
+alter table test_ugi_045 truncate partition p3 update global index;
+set enable_seqscan = off;
+set enable_bitmapscan = off;
+select *from test_ugi_045;
+ c_id |       c_date        | c_info 
+------+---------------------+--------
+    1 | 2021-05-30 00:00:00 | 1-1
+    2 | 2021-05-31 00:00:00 | 1-1
+    3 | 2021-06-01 00:00:00 | 1-1
+    4 | 2021-06-02 00:00:00 | 1-2
+    7 | 2021-06-05 00:00:00 | 1-3
+    8 | 2021-06-06 00:00:00 | 1-4
+    9 | 2021-06-07 00:00:00 | 1-4
+   10 | 2021-06-08 00:00:00 | 1-5
+(8 rows)
+
+select *from test_ugi_045 where c_info = '1-2';
+ c_id |       c_date        | c_info 
+------+---------------------+--------
+    4 | 2021-06-02 00:00:00 | 1-2
+(1 row)
+
+drop table if exists test_ugi_045;
 -- 清理过程
 drop view check_truncate_results;
 drop materialized view pg_partition_before_truncate;

src/test/regress/input/subscription.source

@@ -24,6 +24,7 @@ CREATE SUBSCRIPTION testsub CONNECTION 'foo';
 CREATE SUBSCRIPTION testsub PUBLICATION foo;
 -- fail - could not connect to the publisher
 create subscription testsub2 connection 'host=abc' publication pub;
+create subscription testsub2 connection 'host=abc port=12345' publication pub;
 set client_min_messages to error;
 -- fail - syntax error, invalid connection string syntax: missing "="
 CREATE SUBSCRIPTION testsub CONNECTION 'testconn' PUBLICATION testpub;
@@ -32,12 +33,15 @@ CREATE SUBSCRIPTION testsub CONNECTION 'dbname=doesnotexist' PUBLICATION testpub
 CREATE SUBSCRIPTION testsub CONNECTION 'dbname=doesnotexist' PUBLICATION testpub WITH (ENABLED=false, slot_name='testsub', synchronous_commit=off);
 -- create SUBSCRIPTION with conninfo in two single quote, used to check mask string bug
 CREATE SUBSCRIPTION testsub_maskconninfo CONNECTION 'host=''1.2.3.4'' port=''12345'' user=''username'' dbname=''postgres'' password=''password_1234''' PUBLICATION testpub WITH (ENABLED=false, slot_name='testsub', synchronous_commit=off);
-
+-- fail - The number of host and port are inconsistent
+create subscription sub1 connection 'dbname=postgres user=pubusr password=Huawei@123 host=192.168.0.38,192.168.0.38,192.168.0.38 port=14001,14501' publication pub1;
+-- fail - a maximum of 9 servers are supported
+create subscription sub1 connection 'dbname=postgres user=pubusr password=Huawei@123 host=192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38 port=14001,14501' publication pub1;
 -- alter connection
 ALTER SUBSCRIPTION testsub CONNECTION 'host=''1.2.3.4'' port=''12345'' user=''username'' dbname=''postgres'' password=''password_1234''';
 ALTER SUBSCRIPTION testsub CONNECTION 'dbname=does_not_exist';
 reset client_min_messages;
-select subname, pg_get_userbyid(subowner) as Owner, subenabled, subconninfo, subpublications from pg_subscription where subname='testsub';
+select subname, pg_get_userbyid(subowner) as Owner, subenabled, subconninfo, subpublications, subbinary from pg_subscription where subname='testsub';
 --- alter subscription
 ------ set publication
 ALTER SUBSCRIPTION testsub SET PUBLICATION testpub2, testpub3;
@@ -57,6 +61,9 @@ select subname, subenabled, subsynccommit from pg_subscription  where subname='t
 ALTER SUBSCRIPTION testsub SET (slot_name='testsub');
 -- alter owner
 ALTER SUBSCRIPTION testsub owner to regress_subscription_user2;
+-- alter subbinary to true
+ALTER SUBSCRIPTION testsub SET (binary=true);
+select subname, subbinary from pg_subscription where subname='testsub';
 --rename
 ALTER SUBSCRIPTION testsub rename to testsub_rename;
 --- inside a transaction block

src/test/regress/output/subscription.source

@@ -60,8 +60,10 @@ LINE 1: CREATE SUBSCRIPTION testsub PUBLICATION foo;
                                     ^
 -- fail - could not connect to the publisher
 create subscription testsub2 connection 'host=abc' publication pub;
+ERROR:  The number of host and port are inconsistent.
+create subscription testsub2 connection 'host=abc port=12345' publication pub;
 WARNING:  apply worker could not connect to the remote server
-ERROR:  could not connect to the publisher
+ERROR:  Failed to connect to publisher.
 set client_min_messages to error;
 -- fail - syntax error, invalid connection string syntax: missing "="
 CREATE SUBSCRIPTION testsub CONNECTION 'testconn' PUBLICATION testpub;
@@ -72,14 +74,20 @@ ERROR:  unrecognized subscription parameter: create_slot
 CREATE SUBSCRIPTION testsub CONNECTION 'dbname=doesnotexist' PUBLICATION testpub WITH (ENABLED=false, slot_name='testsub', synchronous_commit=off);
 -- create SUBSCRIPTION with conninfo in two single quote, used to check mask string bug
 CREATE SUBSCRIPTION testsub_maskconninfo CONNECTION 'host=''1.2.3.4'' port=''12345'' user=''username'' dbname=''postgres'' password=''password_1234''' PUBLICATION testpub WITH (ENABLED=false, slot_name='testsub', synchronous_commit=off);
+-- fail - The number of host and port are inconsistent
+create subscription sub1 connection 'dbname=postgres user=pubusr password=Huawei@123 host=192.168.0.38,192.168.0.38,192.168.0.38 port=14001,14501' publication pub1;
+ERROR:  The number of host and port are inconsistent.
+-- fail - a maximum of 9 servers are supported
+create subscription sub1 connection 'dbname=postgres user=pubusr password=Huawei@123 host=192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38,192.168.0.38 port=14001,14501' publication pub1;
+ERROR:  Currently, a maximum of 9 servers are supported.
 -- alter connection
 ALTER SUBSCRIPTION testsub CONNECTION 'host=''1.2.3.4'' port=''12345'' user=''username'' dbname=''postgres'' password=''password_1234''';
 ALTER SUBSCRIPTION testsub CONNECTION 'dbname=does_not_exist';
 reset client_min_messages;
-select subname, pg_get_userbyid(subowner) as Owner, subenabled, subconninfo, subpublications from pg_subscription where subname='testsub';
- subname |           owner           | subenabled |      subconninfo       | subpublications 
----------+---------------------------+------------+------------------------+-----------------
- testsub | regress_subscription_user | f          | dbname=does_not_exist  | {testpub}
+select subname, pg_get_userbyid(subowner) as Owner, subenabled, subconninfo, subpublications, subbinary from pg_subscription where subname='testsub';
+ subname |           owner           | subenabled |      subconninfo       | subpublications | subbinary 
+---------+---------------------------+------------+------------------------+-----------------+-----------
+ testsub | regress_subscription_user | f          | dbname=does_not_exist  | {testpub}       | f
 (1 row)
 
 --- alter subscription
@@ -122,6 +130,14 @@ ALTER SUBSCRIPTION testsub SET (slot_name='testsub');
 ERROR:  Currently enabled=false, cannot change slot_name to a non-null value.
 -- alter owner
 ALTER SUBSCRIPTION testsub owner to regress_subscription_user2;
+-- alter subbinary to true
+ALTER SUBSCRIPTION testsub SET (binary=true);
+select subname, subbinary from pg_subscription where subname='testsub';
+ subname | subbinary 
+---------+-----------
+ testsub | t
+(1 row)
+
 --rename
 ALTER SUBSCRIPTION testsub rename to testsub_rename;
 --- inside a transaction block
@@ -134,8 +150,7 @@ COMMIT;
 -- -- active SUBSCRIPTION
 BEGIN;
 ALTER SUBSCRIPTION testsub_rename ENABLE;
-WARNING:  apply worker could not connect to the remote server
-ERROR:  could not connect to the publisher
+ERROR:  invalid connection string syntax, missing host and port
 select subname, subenabled from pg_subscription  where subname='testsub_rename';
 ERROR:  current transaction is aborted, commands ignored until end of transaction block, firstChar[Q]
 ALTER SUBSCRIPTION testsub_rename SET (ENABLED=false);
@@ -281,10 +296,11 @@ SELECT object_name,detail_info FROM pg_query_audit('2022-01-13 9:30:00', '2031-1
  testsub_maskconninfo | ALTER SUBSCRIPTION testsub_maskconninfo SET (conninfo='*************************************************************************************************************************;
  testsub              | ALTER SUBSCRIPTION testsub SET (synchronous_commit=on);
  testsub              | ALTER SUBSCRIPTION testsub owner to regress_subscription_user2;
+ testsub              | ALTER SUBSCRIPTION testsub SET (binary=true);
  testsub              | ALTER SUBSCRIPTION testsub rename to testsub_rename;
  testsub_rename       | DROP SUBSCRIPTION IF EXISTS testsub_rename;
  testsub_maskconninfo | DROP SUBSCRIPTION IF EXISTS testsub_maskconninfo;
-(14 rows)
+(15 rows)
 
 --clear audit log
 SELECT pg_delete_audit('1012-11-10', '3012-11-11');

src/test/regress/parallel_schedule0

@@ -25,7 +25,7 @@ test: pldeveloper_gs_source
 test: index_advisor
 test: pl_debugger_server pl_debugger_client
 test: update_for_wait_s1 update_for_wait_s2
-test: plan_hint plan_hint_set plan_hint_no_expand plan_hint_iud null_test_opt
+test: plan_hint plan_hint_set plan_hint_no_expand plan_hint_iud null_test_opt deserialize_func
 test: large_sequence int16 gs_dump_sequence
 test: gs_dump_tableof
 test: analyze_commands
@@ -916,3 +916,5 @@ test: fdw_audit
 test: gs_global_config_audit
 test: detail
 test: gs_dump_encrypt
+test: composite_datum_record
+test: join_test_alias

src/test/regress/pg_regress.cpp

@@ -5238,7 +5238,7 @@ static void check_global_variables()
     }
 }
 
-#define BASE_PGXC_LIKE_MACRO_NUM 1401
+#define BASE_PGXC_LIKE_MACRO_NUM 1402
 static void check_pgxc_like_macros()
 {
 #ifdef BUILD_BY_CMAKE 

src/test/regress/sql/composite_datum_record.sql

@@ -0,0 +1,11 @@
+--
+-- check for over-optimization of whole-row Var referencing an Append plan
+--
+create table int4_test(f1 int4);
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6 where f1<=0) q) from int4_test;
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6.0 where f1<=0) q) from int4_test;
+insert into int4_test(f1) values(' 0 ');
+insert into int4_test(f1) values('123456 ');
+select * from int4_test;
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6 where f1<=0) q) from int4_test;
+select (select q from (select 1,2,3 where f1>0 union all select 4,5,6.0 where f1<=0) q) from int4_test;

src/test/regress/sql/deserialize_func.sql

@@ -0,0 +1,63 @@
+create schema deserialize_func1;
+set current_schema = deserialize_func1;
+
+create sequence test_seq;
+
+create or replace function test_func() return int
+as
+begin
+return 1;
+end;
+/
+
+create table t1(key int, id int default nextval('test_seq'));
+
+create table t2(key int, id int default test_func());
+
+select * from pg_get_tabledef('deserialize_func1.t1');
+
+select * from pg_get_tabledef('deserialize_func1.t2');
+
+alter schema deserialize_func1 rename to deserialize_func2;
+
+select * from pg_get_tabledef('deserialize_func2.t1');
+
+select * from pg_get_tabledef('deserialize_func2.t2');
+
+alter table deserialize_func2.t1 alter column id drop default;
+
+alter table deserialize_func2.t1 alter column id set default nextval('deserialize_func2.test_seq');
+
+select * from pg_get_tabledef('deserialize_func2.t1');
+
+alter table deserialize_func2.t2 alter column id drop default;
+
+alter table deserialize_func2.t2 alter column id set default deserialize_func2.test_func();
+
+select * from pg_get_tabledef('deserialize_func2.t2');
+
+set current_schema = deserialize_func2;
+
+create type type1 as (a int);
+
+create view v1 as select '(1)'::type1;
+
+alter type type1 rename to type2;
+
+select pg_get_viewdef('v1');
+
+drop view v1 cascade;
+
+create table atest12 as select x as a, 10001-x as b from generate_series(1, 10000) x;
+
+create function func1(integer, integer) returns boolean as $$begin return $1 < $2; end $$ language plpgsql immutable;
+
+create operator <<< (procedure=func1, leftarg=integer, rightarg=integer, restrict=scalarltsel);
+
+create view atest12v as select * from atest12 where b <<< 5;
+
+select pg_get_viewdef('atest12v');
+
+drop schema if exists deserialize_func1 cascade;
+
+drop schema if exists deserialize_func2 cascade;

src/test/regress/sql/drop_if_exists.sql

@@ -145,9 +145,13 @@ DROP TEXT SEARCH CONFIGURATION test_tsconfig_exists;
 -- doesn't exists
 DROP EXTENSION test_extension_exists;
 DROP EXTENSION IF EXISTS test_extension_exists;
--- exists but doesn't support drop
-DROP EXTENSION plpgsql;
-DROP EXTENSION IF EXISTS plpgsql;
+-- only extension in white list support drop, but the guc support_extended_features is true in pg_regress mode, so hdfs_fdw can be droped
+CREATE DATABASE ext_test_db;
+\c ext_test_db
+DROP EXTENSION hdfs_fdw;
+DROP EXTENSION IF EXISTS hdfs_fdw;
+\c regression
+DROP DATABASE ext_test_db;
 
 -- functions
 DROP FUNCTION test_function_exists();

src/test/regress/sql/hw_partition_interval_exchange.sql

@@ -97,3 +97,61 @@ select * from interval_exchange_test order by logdate;
 select * from interval_normal_exchange order by logdate;
 select * from interval_normal_exchange where logdate > '2020-06-01' order by logdate;
 drop table interval_normal_exchange;
+
+drop table table_001;
+create table table_001(
+    COL_1 smallint,
+    COL_2 char(5),
+    COL_3 int,
+    COL_4 date,
+    COL_5 boolean,
+    COL_6 nchar(5),
+    COL_7 float
+    );
+
+drop table partition_table_001;
+create table partition_table_001(
+    COL_1 smallint,
+    COL_2 char(5),
+    COL_3 int,
+    COL_4 date,
+    COL_5 boolean,
+    COL_6 nchar(5),
+    COL_7 float
+    )
+    PARTITION BY RANGE (COL_4)
+    INTERVAL ('1 month')
+    (
+        PARTITION partition_table_001_p1 VALUES LESS THAN ('2020-03-01'),
+        PARTITION partition_table_001_p2 VALUES LESS THAN ('2020-04-01'),
+        PARTITION partition_table_001_p3 VALUES LESS THAN ('2020-05-01')
+    );
+
+insert into partition_table_001 values (1,'aaa',1,'2020-02-23',true,'aaa',1.1);
+insert into partition_table_001 values (2,'bbb',2,'2020-03-23',false,'bbb',2.2);
+insert into partition_table_001 values (3,'ccc',3,'2020-04-23',true,'ccc',3.3);
+insert into partition_table_001 values (4,'ddd',4,'2020-05-23',false,'ddd',4.4);
+insert into partition_table_001 values (5,'eee',5,'2020-06-23',true,'eee',5.5);
+insert into partition_table_001 values (6,'fff',6,'2020-07-23',false,'fff',6.6);
+
+ALTER TABLE partition_table_001 EXCHANGE PARTITION (sys_p1) WITH TABLE table_001;
+select * from table_001 order by 1;
+select * from partition_table_001 order by 1;
+
+select relname, parttype, partstrategy, boundaries from pg_partition
+                where parentid = (select oid from pg_class where relname = 'partition_table_001')
+                order by relname;
+
+ALTER TABLE partition_table_001 EXCHANGE PARTITION (sys_p1) WITH TABLE table_001;
+select * from table_001 order by 1;
+select * from partition_table_001 order by 1;
+
+insert into table_001 values (7,'eee',7,'2020-08-23',true,'eee',7.7);
+
+ALTER TABLE partition_table_001 EXCHANGE PARTITION (sys_p1) WITH TABLE table_001 with validation verbose;
+
+select * from table_001 order by 1;
+select * from partition_table_001 order by 1;
+
+drop table table_001;
+drop table partition_table_001;

src/test/regress/sql/join_test_alias.sql

@@ -0,0 +1,53 @@
+CREATE TABLE t11 (
+sec_code character(6) NOT NULL,
+issue_type character(3) NOT NULL
+);
+
+CREATE TABLE t22 (
+company_code character(6) NOT NULL,
+list_profit_flag character(1) NOT NULL
+);
+
+CREATE TABLE t33 (
+company_code character(6) NOT NULL,
+issue_flag character(6) NOT NULL
+);
+
+insert into t11 values(1,'S04');
+insert into t22 values(1,'Y');
+insert into t33 values(1,'1');
+insert into t33 values(2,'2');
+
+select
+T5.issue_type
+,T6.is_type
+from
+(
+select T1.sec_code
+,issue_type
+--,case when list_profit_flag='Y' then '是' else '否' end as list_profit_flag
+from
+(
+select sec_code
+,case when issue_type in ('S04','S05','S06','S09') then '增发'
+when issue_type in ('S01','S02') then '首发'
+else '其他' end as issue_type
+from t11
+)T1
+left join
+(
+select list_profit_flag
+,company_code
+from t22
+)T2
+on T1.sec_code=T2.company_code
+)T5
+full join
+(
+SELECT CASE
+WHEN issue_flag = ANY (ARRAY['1'::bpchar, '3'::bpchar, '11'::bpchar]) THEN '首发'::text
+ELSE '增发'::text
+END AS is_type from t33
+)T6
+on T5.issue_type=T6.is_type
+;

src/test/regress/sql/row_compression/unsupported_feature.sql

@@ -57,4 +57,6 @@ CREATE TABLE unspported_feature.ustore_table(id int, c1 text) WITH(compresstype=
 CREATE INDEX tbl_pc_idx1 on unspported_feature.index_test(c1) WITH(compresstype=2, storage_type=ustore); --failed
 -- segment
 CREATE TABLE unspported_feature.segment_table(id int, c1 text) WITH(compresstype=2, segment=on);  --failed
-CREATE INDEX on unspported_feature.index_test(c1) WITH(compresstype=2, segment=on); --faled
+CREATE INDEX on unspported_feature.index_test(c1) WITH(compresstype=2, segment=on); --failed
+-- set compress_diff_convert
+create table unspported_feature.compress_byte_test(id int) with (compresstype=2, compress_byte_convert=false, compress_diff_convert = true); -- failed

src/test/regress/sql/truncate_gpi.sql

@@ -192,6 +192,46 @@ explain(costs off) select /*+ indexscan(tg_hash i_tg_hash_local_a) */ * from tg_
 select /*+ indexscan(tg_hash i_tg_hash_local_a) */ * from tg_hash where a < 9;
 
 drop table tg_hash;
+
+drop table if exists test_ugi_045;
+create table test_ugi_045
+(
+ c_id integer not null,
+ c_date DATE,
+ c_info varchar(20) not null
+)
+partition by range(c_date)
+interval('2 day')
+(
+ partition p1 values less than ('2021-06-01 00:00:00'),
+ partition p2 values less than ('2021-06-03 00:00:00'),
+ partition p3 values less than ('2021-06-05 00:00:00'),
+ partition p4 values less than ('2021-06-07 00:00:00'),
+ partition p5 values less than ('2021-06-09 00:00:00')
+);
+
+insert into test_ugi_045(c_id, c_date, c_info) values(1, '2021-05-30 00:00:00', '1-1');
+insert into test_ugi_045(c_id, c_date, c_info) values(2, '2021-05-31 00:00:00', '1-1');
+insert into test_ugi_045(c_id, c_date, c_info) values(3, '2021-06-01 00:00:00', '1-1');
+insert into test_ugi_045(c_id, c_date, c_info) values(4, '2021-06-02 00:00:00', '1-2');
+insert into test_ugi_045(c_id, c_date, c_info) values(5, '2021-06-03 00:00:00', '1-2');
+insert into test_ugi_045(c_id, c_date, c_info) values(6, '2021-06-04 00:00:00', '1-3');
+insert into test_ugi_045(c_id, c_date, c_info) values(7, '2021-06-05 00:00:00', '1-3');
+insert into test_ugi_045(c_id, c_date, c_info) values(8, '2021-06-06 00:00:00', '1-4');
+insert into test_ugi_045(c_id, c_date, c_info) values(9, '2021-06-07 00:00:00', '1-4');
+insert into test_ugi_045(c_id, c_date, c_info) values(10, '2021-06-08 00:00:00', '1-5');
+
+create index global_index_id_045 on test_ugi_045(c_date) global;
+create index global_index_info_045 on test_ugi_045(c_info) global;
+select *from test_ugi_045;
+
+alter table test_ugi_045 truncate partition p3 update global index;
+set enable_seqscan = off;
+set enable_bitmapscan = off;
+select *from test_ugi_045;
+select *from test_ugi_045 where c_info = '1-2';
+drop table if exists test_ugi_045;
+
 -- 清理过程
 drop view check_truncate_results;
-drop materialized view pg_partition_before_truncate;
+drop materialized view pg_partition_before_truncate;