Given the proposed list of tokens for tests, tests must be passable with only a subset since reporting parse errors is optional. Also, though the accurate reporting of errors is a good thing to test, when they are reported in relation to the emission of tokens is likely to differ between implementations, which would lead to different token lists. JKing 16:41, 21 December 2006 (UTC)
There are only ever two possible token streams: the full one, or the one that aborted at the first error. No other differing is allowed. -Hixie 22:55, 21 December 2006 (UTC)
Attributes in the parser test format
What's the order in which they have to be given? Preserving the attribute order does not seem to be needed or is it?