[thunderer]

With a syntax whose opening tag overlaps itself ([[, {{, <<), RegularParser missed tags that follow a run of opening-tag characters. Its lexer cut the run into opening tags from the left, so with [[ it found a in [[[[a]] but not in [[[a]], where the odd [ stuck to the name. HybridParser and RegexParser find a at offset 1, the same way every parser finds [a] in [[a] with the default syntax.

• the lexer now reads a character as text when an opening tag starts on it and another one starts before it ends, so the opening tag token always sits right before the name,
• the lexer regex stays byte-identical for the default syntax and every one-character opening tag, and opening tags that cannot overlap themselves ([!, <!--, {%) keep the same output,
• a closing tag hidden in such a run is now found too, so [[a]]x[[[//a]] gives a with content x[, as HybridParser does,
• RegularParser and HybridParser now give the same output on 13 syntaxes in a differential test of 20000-40000 random inputs each, and the new data-provider test checks both parsers on 8 cases,
• verified PHPUnit on PHP 8.4, the new cases on PHP 5.5, 7.4 and 8.5, Mago and Infection; the 3 escaped mutants on the new lines do not change the output.