Both are valid, so there is no bug.

Euhr.. Nobody said it was a bug, I said "more readable and preferred". Call it an "improvement".

Revisiting this after a few years… A quick and dirty workaround is to preg_replace('/\-\n\s+/', '- ', Yaml::dump($array, 6)). I used 6 as the inline level, adjust accordingly. Also, can't guarantee it won't mess with more complicated structures.

Adding to the dirty workaround, using an indent of two spaces aligns the properties of the array item nicely: preg_replace('/\-\n\s+/', '- ', Yaml::dump($array, 6, 2))

Another version, which may be more robust in edge cases, e.g. if you have text that would cause false positives in the regex match.

  function ymlDump(mixed $data, $inline = 99, $indent = 2, $flags = Yaml::DUMP_MULTI_LINE_LITERAL_BLOCK): string {
    $marker = \uniqid('_', true);
    $dumper = new class($marker) extends Dumper {
      public function __construct(
        private string $marker,
      ) {
        parent::__construct(2);
      }
      public function dump($input, int $inline = 0, int $indent = 0, int $flags = 0): string {
        return $this->marker . parent::dump($input, $inline, $indent, $flags);
      }
    };
    $yml = $dumper->dump($data, $inline, $indent, $flags);
    $yml = \preg_replace(
      sprintf("@-\n%s +@", \preg_quote($marker, '@')),
      '- ',
      $yml,
    );
    $yml = \str_replace($marker, '', $yml);
    return $yml;
  }

Note that this only works well if indentation = 2.
Otherwise it would look inconsistent.

map:
    submap:
        a: A
        b: B
list:  # which of these is better?
    - a: A
      b: B
    -   a: A
        b: B