Should expect some errors for these new tests, right?

Yeah, we should. But AFAIK the errors in general are frequently wrong and I felt like giving up when I did this.

I found the errors that existed were very helpful when writing a parser, but I understand it's a pain to add and maintain them.

@nolanw I just don't have anything off-hand that's up-to-date to generate them from, which is what makes it a real pain. (HTMLReader doesn't have any easy-to-run test binary, does it? That, say, takes input from either a file or stdin and dumps the test expectation format?)

I kinda gave up on the errors once things mostly worked ok (surprise!). That's a good idea for a lil binary though, I'll tuck that under my hat and maybe one day the errors will rain down on us all.