Shebang error parsing: compatibility with tsc

[sandersn]

In tsc, #! anywhere except pos=0 in a file gives an error and increments past #. In tsgo, it was incrementing past #!. tsc's behaviour seems bad, but not bad enough to differ from, especially since the parse tree will be wrong either way.

[jakebailey]

The original code was:

case CharacterCodes.hash:
    if (pos !== 0 && text[pos + 1] === "!") {
        error(Diagnostics.can_only_be_used_at_the_start_of_a_file, pos, 2);
        pos++;
        return token = SyntaxKind.Unknown;
     }

What is all of that logic above the code you're modifying?

[sandersn]

The original code checked for valid #! separately, at the beginning of the function. Since shebang is rare, and # is rare, it makes more sense to move and inline the code for the valid case. So in tsgo it sits right next to the invalid case.

(This is speculation, maybe @ahejlsberg had some other reason for inlining it.)

[DanielRosenwasser]

It never made sense to check specifically at the beginning of the function if you want to give a good error message anyway. I did the same thing at microsoft/TypeScript#59244

[DanielRosenwasser]

Is it just me or does that inlined code not work for a shebang terminated by any newline character other than \n?

[sandersn]

Shebang only works in Unix doncha know 🙃
In JS, it's a regex, so whatever a regex considers a newline. I'll switch it to stringutils.IsLineBreak.