[LLVMbugs] [Bug 11150] New: Pure virtual specifiers broken when Microsoft extensions are turned on

bugzilla-daemon at llvm.org bugzilla-daemon at llvm.org
Mon Oct 17 00:34:38 PDT 2011


           Summary: Pure virtual specifiers broken when Microsoft
                    extensions are turned on
           Product: clang
           Version: trunk
          Platform: PC
        OS/Version: Windows XP
            Status: NEW
          Severity: normal
          Priority: P
         Component: Frontend
        AssignedTo: unassignedclangbugs at nondot.org
        ReportedBy: johnw at boostpro.com
                CC: llvmbugs at cs.uiuc.edu

Doug Gregor's change r141539, which removes the Init argument from
Sema::ActOnCXXMemberDeclarator, causes a problem with Microsoft extensions and
pure virtual specifiers.

There is this code in ParseDeclCXX.cpp

    if (getLang().MicrosoftExt && Tok.is(tok::equal) &&
        DeclaratorInfo.isFunctionDeclarator() && 
        NextToken().is(tok::numeric_constant)) {
      Init = ParseInitializer();
      if (Init.isInvalid())
        SkipUntil(tok::comma, true, true);

This code gets executed for every kind of pure virtual function, causing Init
to be set.  Before r141539, Sema::ActOnCXXMemberDeclarator would see that Init
was set and call AddInitializerToDecl.  This code path is ultimately
responsible for setting Abstract on the class.

With the change, that code moved to Parser::ParseCXXClassMemberDeclaration --
but only if HasInitializer is true, which it never is if the above Microsoft
extension code has consumed the tok::equal that would normally trigger
HasInitializer getting set in Parser::ParseCXXClassMemberDeclaration:

    bool HasInitializer = false;
    bool HasDeferredInitializer = false;
    if (Tok.is(tok::equal) || Tok.is(tok::l_brace)) {
      if (BitfieldSize.get()) {
        Diag(Tok, diag::err_bitfield_member_init);
        SkipUntil(tok::comma, true, true);
      } else {
        HasInitializer = true;       // <---------
        HasDeferredInitializer = !DeclaratorInfo.isDeclarationOfFunction() &&
            != DeclSpec::SCS_static &&
            != DeclSpec::SCS_typedef;

I'm not sure what the best way to fix this is.  I have a feeling the Microsoft
extension code that's doing the eager parsing needs to be rewritten for the new

Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

More information about the llvm-bugs mailing list