[cfe-commits] [PATCH] PR4111, PR5925, PR13210: Make tentative parsing smarter in the presence of undeclared identifiers

[Richard Smith]

On Fri, Jun 29, 2012 at 11:46 AM, Sean Silva <silvas at purdue.edu> wrote:

> +      case ANK_TentativeDecl:
> +      case ANK_TemplateName:
> +        // These two shouldn't happen.
> +      case ANK_Unresolved:
>         break;
>
> assert?
>

Sorry, I forgot to update that comment after some refactoring. The second
case is possible (but only for ill-formed code), and the first case is
completely harmless. I've updated and simplified this code to remove the
switch entirely:

  case tok::identifier: {
    Token Next = NextToken();
    if (Next.is(tok::colon)) { // C99 6.8.1: labeled-statement
      // identifier ':' statement
      return ParseLabeledStatement(Attrs);
    }

    // Look up the identifier, and typo-correct it to a keyword if it's not
    // found.
    if (Next.isNot(tok::coloncolon)) {
      // CorrectionCandidateCallback by default allows typo-correction to
      // anything which can appear at the start of a statement.
      CorrectionCandidateCallback DefaultValidator;
      if (TryAnnotateName(/*IsAddressOfOperand*/false, &DefaultValidator)
            == ANK_Error) {
        // Handle errors here by skipping up to the next semicolon or '}',
and
        // eat the semicolon if that's what stopped us.
        SkipUntil(tok::r_brace, /*StopAtSemi=*/true, /*DontConsume=*/true);
        if (Tok.is(tok::semi))
          ConsumeToken();
        return StmtError();
      }

      // If the identifier was typo-corrected, try again.
      if (Tok.isNot(tok::identifier))
        goto Retry;
    }

    // Fall through
  }
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20120629/2c324d6d/attachment.html>