As in: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1703r1.html
Posting this mostly to get some feedback on the general path taken.
From reading the paper, I understand that the preprocessor should gain a "import-keyword" token, but that is already taken by the GNU #import extension.
The implementation here starts converting newlines to eod tokens when seeing a export or import token. However, that means ignoring the eod tokens when parsing an export block, an exported declaration or a non-header-unit module import.
Let me know what you think.
This confuses me. That's a stray 'export' token, followed by a well-formed import-directive