The code for nextToken() only increments LINENO if EOL is classified as a
whitespace character. However, if eolIsSignificant() then arguably
EOL is *not* whitespace, even though the code is treating it that way.
The test for EOL is significant comes after the check for whitespace.
If the syntax table has been reset, then it's possible that a client
may not classify EOL as whitespace, because it is considered significant.
So either the documentation should reflect the need for EOL to be whitespace
even if significant, or the code ought to be changed to fix the bug.