[Frontend] Errors at EOF are reported with an incorrect range

[rturrado]

Let's say we have a program such as:

version 3

qubit[5] q

h q[0, 4

Where, after the token 4, we just find EOF.

At the moment, libqasm is reporting the following error: Error at input.cq:5:9..14: missing ']' at '<EOF>'.

The starting column points to the position just after 4 (remember columns start at 1).

[rturrado]

I've seen today that: 1) EOF tokens contain an <EOF> text, i.e., a string of size 5. 2) We use the size of the token text for calculating the size of the token.

I've been valuing different options as possible fixes for the issue:

• Force the user to always include a newline before EOF. For the cases where the programmer forgets to write that newline, we would still get the same erroneous message.
• Make the lexer/parser add a newline token before each EOF. We should then differentiate between newlines added by the programmer or by the lexer/parser. And, for this latter case, we would still have an issue reporting the column number (as the column index for the newline added by the lexer/parser would be fictitious).
• Just check if the token is an EOF at the CustomErrorListener and, for that case, report a size of 0. The start and end column indexes would point right after the last character of the file. I have to check with the Frontend people if this is OK with them.