[cfe-commits] r103904 - in /cfe/trunk: lib/AST/ExprConstant.cpp test/CodeGenCXX/references.cpp

[Chandler Carruth]

On Sun, May 16, 2010 at 1:44 PM, Douglas Gregor <dgregor at apple.com> wrote:

> I am beginning to doubt our decision to inherit ParmVarDecl from VarDecl.
>

I can't tell whether that is the problem, or whether the decision to
represent default arguments as initializers is the problem. Fundamentally,
parameters   *are* variables within the function, so it does seem a
reasonable way to represent them.

A slight counter proposal: we re-use the storage of initializers for default
arguments, but the accessor methods for initializers should filter them out,
and ParmVarDecl should provide a dedicated API. Would that resolve this?
Does it introduce other problems?


>
> Sent from my iPhone
>
> On May 16, 2010, at 2:32 AM, Chandler Carruth <chandlerc at gmail.com> wrote:
>
> > Author: chandlerc
> > Date: Sun May 16 04:32:51 2010
> > New Revision: 103904
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=103904&view=rev
> > Log:
> > When constant folding reference variables with an initializer to the
> > initializer, don't fold paramters. Their initializers are just default
> > arguments which can be overridden. This fixes some spectacular
> regressions due
> > to more things making it into the constant folding.
> >
> > Modified:
> >    cfe/trunk/lib/AST/ExprConstant.cpp
> >    cfe/trunk/test/CodeGenCXX/references.cpp
> >
> > Modified: cfe/trunk/lib/AST/ExprConstant.cpp
> > URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/AST/ExprConstant.cpp?rev=103904&r1=103903&r2=103904&view=diff
> >
> ==============================================================================
> > --- cfe/trunk/lib/AST/ExprConstant.cpp (original)
> > +++ cfe/trunk/lib/AST/ExprConstant.cpp Sun May 16 04:32:51 2010
> > @@ -355,6 +355,10 @@
> >   } else if (VarDecl* VD = dyn_cast<VarDecl>(E->getDecl())) {
> >     if (!VD->getType()->isReferenceType())
> >       return Success(E);
> > +    // Reference parameters can refer to anything even if they have an
> > +    // "initializer" in the form of a default argument.
> > +    if (isa<ParmVarDecl>(VD))
> > +      return false;
> >     // FIXME: Check whether VD might be overridden!
> >     if (const Expr *Init = VD->getAnyInitializer())
> >       return Visit(const_cast<Expr *>(Init));
> >
> > Modified: cfe/trunk/test/CodeGenCXX/references.cpp
> > URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenCXX/references.cpp?rev=103904&r1=103903&r2=103904&view=diff
> >
> ==============================================================================
> > --- cfe/trunk/test/CodeGenCXX/references.cpp (original)
> > +++ cfe/trunk/test/CodeGenCXX/references.cpp Sun May 16 04:32:51 2010
> > @@ -155,3 +155,16 @@
> > // CHECK: load
> > // CHECK: ret
> > const int &f2() { return 0; }
> > +
> > +// Don't constant fold const reference parameters with default arguments
> to
> > +// their default arguments.
> > +namespace N1 {
> > +  const int foo = 1;
> > +  // CHECK: @_ZN2N14test
> > +  int test(const int& arg = foo) {
> > +    // Ensure this array is on the stack where we can set values instead
> of
> > +    // being a global constant.
> > +    // CHECK: %args_array = alloca
> > +    const int* const args_array[] = { &arg };
> > +  }
> > +}
> >
> >
> > _______________________________________________
> > cfe-commits mailing list
> > cfe-commits at cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20100516/686c0291/attachment.html>