this is html5
? expect ?
�x3f; should fail
&#}; should fail
? expect ?
? should fail
&#-10; should fail
� should fail
 should fail
 should fail
” error but workaround to "
𿿾 should fail
􃀀 should be something
� should fail
� should fail
Ж cyrillic Zh

& ampersand
link check attrs
(w)≫⃒(w) expands to 6 bytes; would've broken under the old in-place substitution
⟨ ⟩ are different in html4 and html5 (look the same)
& should fail
&---; I left the strchr(":_.-", *s) alone. It's probably not exactly correct for html5.
& & & bare ampersands
' expect '
&asdfghjkl; unrecognized