Infinite loop when pattern matches nothing

[benogle]

Line to tokenize: <
This pattern

  {
    'begin': '(?<=\\A)'
    'end': '(?=<)'
    'name': 'meta.scope.outside-tag.html'
    'patterns': [
      {
        'include': '#entities'
      }
    ]
  }

In this line, match is an object with an empty nextTokens array:

This causes an infinite loop, and locks Atom. Here is an attempt to use these potentially-empty-matching patterns in language-html atom/language-html@bc7c4c9

cc @nathansobo

[nathansobo]

Not going to fix this for now. The reason is that the matches in question were zero width. This should probably be fixed to avoid user error, but for now grammars shouldn't be constructed with zero width matches.