Fixes for DSN parser to use UTF8

lib/DSNParser.pm

@@ -379,6 +379,28 @@ sub get_dbh {
            . ": $EVAL_ERROR";
       }
    }
+   my ($mysql_version) = eval { $dbh->selectrow_array('SELECT VERSION()') };
+   if ($EVAL_ERROR) {
+       die "Cannot get MySQL version: $EVAL_ERROR";
+   }
+
+   my (undef, $character_set_server) = eval { $dbh->selectrow_array("SHOW VARIABLES LIKE 'character_set_server'") };
+   if ($EVAL_ERROR) {
+       die "Cannot get MySQL var character_set_server: $EVAL_ERROR";
+   }
+
+   if ($mysql_version =~ m/^(\d+)\.(\d)\.(\d+).*/) {
+       if ($1 >= 8 && $character_set_server =~ m/^utf8/) {
+           $dbh->{mysql_enable_utf8} = 1;
+           my $msg = "MySQL version $mysql_version >= 8 and character_set_server = $character_set_server\n".
+                     "Setting: SET NAMES $character_set_server";
+           PTDEBUG && _d($msg);
+           eval { $dbh->do("SET NAMES 'utf8mb4'") };
+           if ($EVAL_ERROR) {
+               die "Cannot SET NAMES $character_set_server: $EVAL_ERROR";
+           }
+       }
+   }
 
    PTDEBUG && _d('DBH info: ',
       $dbh,

lib/Sandbox.pm

@@ -133,7 +133,7 @@ sub create_dbs {
 sub get_dbh_for {
    my ( $self, $server, $cxn_ops, $user ) = @_;
    _check_server($server);
-   $cxn_ops ||= { AutoCommit => 1 };
+   $cxn_ops ||= { AutoCommit => 1, mysql_enable_utf8 => 1 };
    $user    ||= 'msandbox';
    PTDEBUG && _d('dbh for', $server, 'on port', $port_for{$server});
    my $dp = $self->{DSNParser};

lib/TableChunker.pm

@@ -44,6 +44,7 @@
 package TableChunker;
 
 use strict;
+use utf8;
 use warnings FATAL => 'all';
 use English qw(-no_match_vars);
 use constant PTDEBUG => $ENV{PTDEBUG} || 0;
@@ -591,7 +592,7 @@ sub _chunk_char {
       $dbh->do($sql);
       my $col_def = $args{tbl_struct}->{defs}->{$chunk_col};
       $sql        = "CREATE TEMPORARY TABLE $tmp_db_tbl ($col_def) "
-                  . "ENGINE=MEMORY";
+                  . "ENGINE=MEMORY DEFAULT CHARSET = utf8";
       PTDEBUG && _d($dbh, $sql);
       $dbh->do($sql);
 
@@ -601,7 +602,8 @@ sub _chunk_char {
       PTDEBUG && _d($dbh, $sql);
       my $ins_char_sth = $dbh->prepare($sql);  # avoid quoting issues
       for my $char_code ( $min_col_ord..$max_col_ord ) {
-         $ins_char_sth->execute($char_code);
+         # Ignore errors on invalid chars for UTF8
+         eval { $ins_char_sth->execute($char_code) };
       }
 
       # Select from the char-to-number map all characters between the

sandbox/servers/8.0/my.sandbox.cnf

@@ -29,7 +29,7 @@ report-host                = 127.0.0.1
 report-port                = PORT
 server-id                  = PORT
 #local_infile               = ON
-#default_authentication_plugin=mysql_native_password
+default_authentication_plugin=mysql_native_password
 
 # fkc test
 binlog_format              = STATEMENT

t/pt-online-schema-change/basics.t

@@ -315,6 +315,7 @@ sub test_alter_table {
       # still functiona: i.e. that they'll prevent us from delete
       # a parent row that's being referenced by a child.
       my $sql = "DELETE FROM $table WHERE $pk_col=1 LIMIT 1";
+      warn ">> SQL: $sql";
       eval {
          $master_dbh->do($sql);
       };
@@ -471,6 +472,7 @@ test_alter_table(
    ],
 );
 
+diag(">>>>>>>");
 test_alter_table(
    name       => "Basic FK drop_swap --execute",
    table      => "pt_osc.country",
@@ -487,6 +489,7 @@ test_alter_table(
    ],
 );
 
+diag("2 >>>>>>>");
 # Let the tool auto-determine the fk update method.  This should choose
 # the rebuild_constraints method because the tables are quite small.
 # This is tested by indicating the rebuild_constraints method, which
@@ -510,6 +513,7 @@ test_alter_table(
    ],
 );
 
+diag("3 >>>>>>>");
 # Specify --alter-foreign-keys-method for a table with no child tables.
 test_alter_table(
    name        => "Child table",
@@ -527,6 +531,7 @@ test_alter_table(
    ],
 );
 
+diag("4 >>>>>>>");
 # Use drop_swap to alter address, which no other table references,
 # so the tool should re-enable --swap-tables and --drop-old-table.
 test_alter_table(
@@ -545,6 +550,7 @@ test_alter_table(
    ],
 );
 
+diag("5 >>>>>>>");
 # Alter city and verify that its fk to country still exists.
 # (https://bugs.launchpad.net/percona-toolkit/+bug/969726)
 test_alter_table(

t/pt-online-schema-change/samples/basic_with_fks.sql

@@ -32,25 +32,25 @@ CREATE TABLE `address` (
   CONSTRAINT `fk_address_city` FOREIGN KEY (`city_id`) REFERENCES `city` (`city_id`) ON UPDATE CASCADE
 ) ENGINE=InnoDB;
 
-INSERT INTO pt_osc.country VALUES
-  (1, 'Canada', null),
-  (2, 'USA',    null),
-  (3, 'Mexico', null),
-  (4, 'France', null),
-  (5, 'Spain',  null);
+INSERT INTO pt_osc.country (`country_id`, `country`) VALUES
+  (1, 'Canada'),
+  (2, 'USA'),
+  (3, 'Mexico'),
+  (4, 'France'),
+  (5, 'Spain');
 
-INSERT INTO pt_osc.city VALUES
-  (1, 'Montréal', 1, null),
-  (2, 'New York', 2, null),
-  (3, 'Durango',  3, null),
-  (4, 'Paris',    4, null),
-  (5, 'Madrid',   5, null);
+INSERT INTO pt_osc.city (`city_id`, `city`, `country_id`) VALUES
+  (1, 'Montréal', 1),
+  (2, 'New York', 2),
+  (3, 'Durango',  3),
+  (4, 'Paris',    4),
+  (5, 'Madrid',   5);
 
-INSERT INTO pt_osc.address VALUES
-  (1, 'addy 1', 1, '10000', null),
-  (2, 'addy 2', 2, '20000', null),
-  (3, 'addy 3', 3, '30000', null),
-  (4, 'addy 4', 4, '40000', null),
-  (5, 'addy 5', 5, '50000', null);
+INSERT INTO pt_osc.address (`address_id`, `address`, `city_id`, `postal_code`) VALUES
+  (1, 'addy 1', 1, '10000'),
+  (2, 'addy 2', 2, '20000'),
+  (3, 'addy 3', 3, '30000'),
+  (4, 'addy 4', 4, '40000'),
+  (5, 'addy 5', 5, '50000');
 
 SET foreign_key_checks=1;