"do{;}while(false)false" prohibited in spec but allowed in consensus reality

[littledan]

This bug is following up on an earlier bug in the old tracker. I just checked, and the syntax is still allowed in Chrome and Firefox. Any thoughts on whether this should be standardized? I don't have any data on how often it's used, but it seems useful for minimizers. in V8, the syntax has been allowed since inception, with the following comment in the code from 2008:

// Allow do-statements to be terminated with and without
// semi-colons. This allows code such as 'do;while(0)return' to
// parse, which would not be the case if we had used the
// ExpectSemicolon() functionality here.

I think the grammar change would be as simple as leaving off the semicolon at the end of IterationStatement. Thoughts? cc @erights who filed the original bug.

[bterlson]

FWIW it also parses in Chakra and JSC. But isn't this just ASI semantics? ASI calls out this case specifically in bullet 1.3 (https://tc39.github.io/ecma262/#sec-rules-of-automatic-semicolon-insertion).

[bterlson]

FWIW it also parses in Chakra and JSC. But isn't this just ASI semantics? ASI calls out this case specifically in bullet 1.3 (https://tc39.github.io/ecma262/#sec-rules-of-automatic-semicolon-insertion).

[littledan]

Oh, my mistake, it's right there.

[littledan]

Oh, my mistake, it's right there.

[erights]

Yup. I just checked https://es5.github.io/#x7.9.1 and this third bullet is new; presumably to close exactly this bug.

[erights]

Yup. I just checked https://es5.github.io/#x7.9.1 and this third bullet is new; presumably to close exactly this bug.

[allenwb]

Also see Annex E entry for 11.9.1.

[allenwb]

Also see Annex E entry for 11.9.1.