Commit Graph

361 Commits

Author SHA1 Message Date
Douglas Gregor 786ab2119f Tweak Sema::CheckReferenceInit so that it (optionally) computes an
ImplicitConversionSequence and, when doing so, following the specific
rules of [over.best.ics]. 

The computation of the implicit conversion sequences implements C++
[over.ics.ref], but we do not (yet) have ranking for implicit
conversion sequences that use reference binding.

llvm-svn: 58357
2008-10-29 02:00:59 +00:00
Douglas Gregor 8e1cf608dc Implement initialization of a reference (C++ [dcl.init.ref]) as part
of copy initialization. Other pieces of the puzzle:

  - Try/Perform-ImplicitConversion now handles implicit conversions
    that don't involve references.
  - Try/Perform-CopyInitialization uses
    CheckSingleAssignmentConstraints for C. PerformCopyInitialization
    is now used for all argument passing and returning values from a
    function.
  - Diagnose errors with declaring references and const values without
    an initializer. (Uses a new Action callback, ActOnUninitializedDecl).
  
We do not yet have implicit conversion sequences for reference
binding, which means that we don't have any overloading support for
reference parameters yet.

llvm-svn: 58353
2008-10-29 00:13:59 +00:00
Douglas Gregor cea4e74340 Some cleanups for the ambiguous derived-to-base conversion checks
llvm-svn: 58096
2008-10-24 16:17:19 +00:00
Douglas Gregor 39c16d445e First non-embarrassing cut at checking for ambiguous derived-to-base
conversions.

Added PerformImplicitConversion, which follows an implicit conversion sequence
computed by TryCopyInitialization and actually performs the implicit
conversions, including the extra check for ambiguity mentioned above.

llvm-svn: 58071
2008-10-24 04:54:22 +00:00
Douglas Gregor 5c407d9a9b Add support for conversions from a pointer-to-derived to a
pointer-to-base. Also, add overload ranking for pointer conversions
(for both pointer-to-void and derived-to-base pointer conversions).

Note that we do not yet diagnose derived-to-base pointer conversion
errors that stem from ambiguous or inacessible base classes. These
aren't handled during overload resolution; rather, when the conversion
is actually used we go ahead and diagnose the error.

llvm-svn: 58017
2008-10-23 00:40:37 +00:00
Douglas Gregor 29a9247ec2 Add representation of base classes in the AST, and verify that we
don't have duplicated direct base classes.

Seriliazation of base class specifiers is not yet implemented.

llvm-svn: 57991
2008-10-22 17:49:05 +00:00
Douglas Gregor 4b62ec6353 QualType::isMoreQualifiedThan and isAtLeastAsQualifiedAs assert that we
aren't trying to compare with address-space qualifiers (for now).

Clean up handing of DeclRefExprs in Expr::isLvalue and refactor part
of the check into a static DeclCanBeLvalue.

llvm-svn: 57980
2008-10-22 15:04:37 +00:00
Douglas Gregor e1eb9d8cc4 Implement ranking of standard conversion sequences by their qualification
conversions (e.g., comparing int* -> const int* against 
int* -> const volatile int*); see C++ 13.3.3.2p3 bullet 3.

Add Sema::UnwrapSimilarPointerTypes to simplify the control flow of
IsQualificationConversion and CompareQualificationConversion (and fix
the handling of the int* -> volatile int* conversion in the former).
 

llvm-svn: 57978
2008-10-22 14:17:15 +00:00
Douglas Gregor ea2d4211e5 Fix a thinko in the qualification-conversion check when the qualificaitons are disjoint, and add some overloading-based tests of qualification conversions
llvm-svn: 57942
2008-10-22 00:38:21 +00:00
Douglas Gregor 9a6579340f Initial step toward supporting qualification conversions (C++ 4.4).
Changes:
  - Sema::IsQualificationConversion determines whether we have a qualification
    conversion.
  - Sema::CheckSingleAssignment constraints now follows the C++ rules in C++,
    performing an implicit conversion from the right-hand side to the type of
    the left-hand side rather than checking based on the C notion of 
    "compatibility". We now rely on the implicit-conversion code to
    determine whether the conversion can happen or
    not. Sema::TryCopyInitialization has an ugly reference-related
    hack to cope with the initialization of references, for now.
  - When building DeclRefExprs, strip away the reference type, since
    there are no expressions whose type is a reference. We'll need to
    do this throughout Sema.
  - Expr::isLvalue now permits functions to be lvalues in C++ (but not
  in C).

llvm-svn: 57935
2008-10-21 23:43:52 +00:00
Douglas Gregor 5251f1b283 Preliminary support for function overloading
llvm-svn: 57909
2008-10-21 16:13:35 +00:00