Need tests for \8 in templates and strict strings

[rkirsling]

Looks like tests for \8 were missed from #1063.

Everyone agrees that \08 is bad outside of sloppy non-template strings, but SM, V8, XS, and ChakraCore all accept `\8` and '\8' regardless of mode.

Templates should hit the NotEscapeSequence early errors in 12.2.9.1 while strict non-templates should fail to lex altogether per the definition of CharacterEscapeSequence in 11.8.4.

[anba]

\8 isn't allowed in any mode per spec (*). As non-standard extension, some engines allow it in strict and non-strict mode, and other engines only allow it in non-strict mode.

(*) B.1.1 Numeric Literals supports noctal digits through NonOctalDecimalIntegerLiteral, whereas B.1.2 String Literals doesn't have an equivalent production rule.

[rkirsling]

Oh hmm, you're right that B.1.2 doesn't actually call out 8 and 9 after all.

This feels really weird to have unspecified though, right? Perhaps we should add a NODIL-equivalent production to B.1.2, since everyone agrees on sloppy (non-template) strings?

[anba]

It's already planned to resolve this for ES7: https://tc39.es/archives/bugzilla/3477/ 😄

[rkirsling]

Hah, wow! Guess I'll open a 262 ticket tomorrow then. 😅

Edit: Opened tc39/ecma262#2039.