We now have unambiguous rules for separating the picture string of format-integer into a primary format token and modifiers; but we still have some ambiguity about how errors in the primary format token are handled.

In particular, for the first form of format token (consisting of digits, '#' signs, and grouping separators), we say "With this form, the primary format token must match the regular expression ^((\p{Nd}|#|[^\p{N}\p{L}])+?)$". The question is, what happens if it doesn't? There are two possible interpretations:

(a) it doesn't belong to this form of format token, it belongs to some other form, presumably the last form, which is implementatation defined, falling back to "1" if it's not recognized.

(b) it's an error.

I think the intent is (b), in which case we ought to make this clear, and should define an error code.

I propose to clarify that any format token that contains at least one Unicode digit belongs to the first form of format token and must conform to the rules for that form, otherwise a dynamic error occurs, for which I intend to re-use FODF1310.

A number of test cases in test set fn-format-integer are affected, in particular those that currently raise XTDE0030.