TYPO3 Core

the issue is not DatabaseConnection class. charset is read properly from configuration there

it seems that
TYPO3\CMS\Core\Database\Schema\ConnectionMigrator
or TYPO3\CMS\Core\Database\Schema\SchemaMigrator
does not read that configuration parameter

I was able to work arround by adding the following
in TYPO3\CMS\Core\Database\Schema\ConnectionMigrator line 1211

$tableOptions = $table->getOptions();
            $connectionParams = $connection->getParams();
            if (isset($connectionParams['charset'])) {
                $tableOptions['charset'] = $connectionParams['charset'];
            }
            if (isset($connectionParams['collate'])) {
                $tableOptions['collate'] = $connectionParams['collate'];
            }

Hi Morton

we need to store 4 Byte Unicode characters like emoji 'http://apps.timwhitlock.info/emoji/tables/unicode'
the default utf8 does only allow storing 3 byte unicode characters. most of emoji characters cannot be stored into utf8.

thats why i need the connection to be utf8mb4 and the database to create tables with charset utf8mb4 and collate utf8mb4_unicode_ci

the first part does indeed work. Typo3 does connect with charset utf8mb4 if i set it in LocalConfiguration.php

but how can I ensure that tables are created with correct charset and collate during setup?

Symfony has separate config parameter for table schemes http://symfony.com/doc/current/doctrine.html

doctrine:
    dbal:
        charset: utf8mb4
        default_table_options:
            charset: utf8mb4
            collate: utf8mb4_unicode_ci

as a suggestion

'DB' => [
        'Connections' => [
            'Default' => [
                'charset' => 'utf8mb4',
                'dbname' => '--dbname--',
                'driver' => 'mysqli',
                'host' => '127.0.0.1',
                'password' => '--mypassword--',
                'port' => 3306,
                'user' => '--myuser--',      
                'tableoptions' => [
                     'charset' => 'utf8mb4',
                     'collate' => 'utf8mb4_unicode_ci'
                ]     
            ],
        ],
    ],

please keep in mind that utf8mb4 uses 4 bytes per char, while "standard" utf8 collation uses 3 bytes per char, which means that indices created might exceed maximum key length limit in mysql.
E.g. by default the key size is 767 which is lower than varchar(255) in utf8, but exceeded with varchar(255) with utf8mb4 (255*4 = 1020)

FYI, MySQL 8 will come with utf8mb4 as default charset

I have pushed a change to gerrit which implements the config suggestion by Marco von Arx. I'm not entirely sure I found all relevant places to change, but in my tests this worked for the database analyzer in the install tool. Newly created tables are generated with utf8mb4.

Patch set 2 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 3 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 4 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 5 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 6 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 7 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 8 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440

Patch set 9 for branch master of project Packages/TYPO3.CMS has been pushed to the review server.
It is available at https://review.typo3.org/56440