Closes #2045. At the moment it breaks a lot of tests, but the couple I've spot checked so far look correct and the previous expectations were wildly wrong (mostly applying parskip more than once when nothing in the input document suggested that should be happening). I'll keep reviewing test but I think this is saner than what we had before even if more bugs may be lurking.
Closes #2045. At the moment it breaks a lot of tests, but the couple I've spot checked so far look correct and the previous expectations were wildly wrong (mostly applying parskip more than once when nothing in the input document suggested that should be happening). I'll keep reviewing test but I think this is saner than what we had before even if more bugs may be lurking.