swig/CHANGES.current

Version 1.3.41 (in progress)
============================

2009-11-13: wsfulton
            Fix usage of nested template classes within templated classes so that compileable code 
            is generated.

2009-11-13: olly
	    [php] Fix place where class prefix (as specified with -prefix)
	    wasn't being used.  Patch from gverbruggen in SF#2892647.

2009-11-12: wsfulton
            Fix usage of nested template classes so that compileable code is generated - the nested
            template class is now treated like a normal nested classes, that is, as an opaque type
            unless the nestedworkaround feature is used.

2009-11-12: wsfulton
            Replace SWIGWARN_PARSE_NESTED_CLASS with SWIGWARN_PARSE_NAMED_NESTED_CLASS and
            SWIGWARN_PARSE_UNNAMED_NESTED_CLASS for named and unnamed nested classes respectively.

            Named nested class ignored warnings can now be suppressed by name using %warnfilter, eg:

              %warnfilter(SWIGWARN_PARSE_NAMED_NESTED_CLASS) Outer::Inner;

            but clearly unnamed nested classes cannot and the global suppression is still required, eg:

              #pragma SWIG nowarn=SWIGWARN_PARSE_UNNAMED_NESTED_CLASS

2009-11-11: wsfulton
            Added the nestedworkaround feature as a way to use the full functionality of a nested class 
            (C++ mode only). It removes the nested class from SWIG's type information so it is as if SWIG
            had never parsed the nested class. The documented nested class workarounds using a global
            fake class stopped working when SWIG treated the nested class as an opaque pointer, and
            this feature reverts this behaviour. The documentation has been updated with details of how
            to use and implement it, see the "Nested classes" section in SWIGPlus.html.

2009-11-11: wsfulton
            There were a number of C++ cases where nested classes/structs/unions were being handled
            as if C code was being parsed which would oftentimes lead to uncompileable code as an
            attempt was made to wrap the nested structs like it is documented for C code. Now all
            nested structs/classes/unions are ignored in C++ mode, as was always documented. However,
            there is an improvement as usage of nested structs/classes/unions is now always treated
            as an opaque type by default, resulting in generated code that should always compile.

            *** POTENTIAL INCOMPATIBILITY ***

2009-11-09: drjoe
	    Fix R for -fcompact and add std_map.i

2009-11-08: wsfulton
            Fix inconsistency for nested structs/unions/classes. Uncompileable code was being
            generated when inner struct and union declarations were used as types within the
            inner struct. The inner struct/union is now treated as a forward declaration making the
            behaviour the same as an inner class. (C++ code), eg:

              struct Outer {
                struct InnerStruct { int x; };
                InnerStruct* getInnerStruct();
              };

2009-11-08: wsfulton
	    Ignored nested class/struct warnings now display the name of the ignored class/struct.

2009-11-07: wsfulton
	    Bug #1514681 - Fix nested template classes within a namespace generated uncompileable
            code and introduced strange side effects to other wrapper code especially code
            after the nested template class. Note that nested template classes are still ignored.

2009-11-07: wsfulton
	    Add new debug options:
              -debug-symtabs    - Display symbol tables information
              -debug-symbols    - Display target language symbols in the symbol tables
              -debug-csymbols   - Display C symbols in the symbol tables

2009-11-03: wsfulton
	    Fix some usage of global scope operator, for example:

              namespace AA { /* ... */ }
              using namespace ::AA;

            and bug #1816802 - SwigValueWrapper should be used:

              struct CC {
                CC(int); // no default constructor
              };
              ::CC x();

            and in template parameter specializations:

              struct S {};
              template <typename T> struct X { void a() {}; };
              template <> struct X<S> { void b() {}; };
              %template(MyTConcrete) X< ::S >;

            plus probably some other corner case usage of ::.

2009-11-02: olly
	    [Python] Fix potential memory leak in initialisation code for the
	    generated module.

2009-10-23: wsfulton
	    Fix seg fault when using a named nested template instantiation using %template(name) 
            within a class. A warning that these are not supported is now issued plus processing 
            continues as if no name was given.

2009-10-20: wsfulton
	    [Python] Fix std::vector<const T*>. This would previously compile, but not run correctly.

2009-10-20: wsfulton
	    Fixed previously fairly poor template partial specialization and explicit 
            specialization support. Numerous bugs in this area have been fixed including:

            - Template argument deduction implemented for template type arguments, eg this now
              works:
                template<typename T> class X      {};
                template<typename T> class X<T *> {};
                %template(X1) X<const int *>; // Chooses T * specialization

              and more complex cases with multiple parameters and a mix of template argument
              deduction and explicitly specialised parameters, eg:
                template <typename T1, typename T2> struct TwoParm               { void a() {} };
                template <typename T1>              struct TwoParm<T1 *, int *>  { void e() {} };
                %template(E) TwoParm<int **, int *>;

            Note that the primary template must now be in scope, like in C++, when
            an explicit or partial specialization is instantiated with %template.

            *** POTENTIAL INCOMPATIBILITY ***

2009-09-14: wsfulton
	    [C#] Add %csattributes for adding C# attributes to enum values, see docs for example.

2009-09-11: wsfulton
	    Fix memmove regression in cdata.i as reported by Adriaan Renting.

2009-09-07: wsfulton
	    Fix constant expressions containing <= or >=.

2009-09-02: wsfulton
	    The following operators in constant expressions now result in type bool for C++ 
            wrappers and remain as type int for C wrappers, as per each standard: 

              && || == != < > <= >= (Actually the last 4 are still broken). For example:

              #define A 10
              #define B 10
              #define A_EQ_B  A == B // now wrapped as type bool for C++
              #define A_AND_B A && B // now wrapped as type bool for C++

2009-09-02: wsfulton
	    Fix #2845746. true and false are now recognised keywords (only when wrapping C++).
            Constants such as the following are now wrapped (as type bool):
              #define FOO true
              #define BAR FOO && false