gcc - C++ - What is a "type transparent class"? -

using gcc compile program includes support decimal data types, encountered following error:

error: type transparent class 'std::decimal::decimal32' has base classes

a quick @ gcc's source tree shows error message found in gcc/cp/class.c.

what "type transparent class"? why error such class have "base classes"?

reading source code of gcc bit more, in semantics.c:

  if (tree_code (t) == record_type       && !processing_template_decl)     {       tree ns = type_context (t);       if (ns && tree_code (ns) == namespace_decl           && decl_context (ns) == std_node           && decl_name (ns)           && !strcmp (identifier_pointer (decl_name (ns)), "decimal"))         {           const char *n = type_name_string (t);           if ((strcmp (n, "decimal32") == 0)               || (strcmp (n, "decimal64") == 0)               || (strcmp (n, "decimal128") == 0))             type_transparent_aggr (t) = 1;         }     } 

this code means type marked transparent if:

• it struct, not template;
• and @ namespace level, , namespace std::decimal.
• and named decimal32, decimal64 or decimal128.

in class.c there error check encountered, , few more.

and in mangle.c:

      /* according c++ abi, library classes passed          same scalar type of single member , use same          mangling.  */       if (tree_code (type) == record_type && type_transparent_aggr (type))         type = tree_type (first_field (type)); 

the comment key here. think means transparent type replaced type of fist (and only) member, can used anywhere first member can. example, in include/decimal class std::decimal::decimal32 has single field of type __decfloat32 (from previous typedef float __decfloat32 __attribute__((mode(sd)));), function takes __decfloat32 can take std::decimal::decimal32 , vice-versa. function decoration done same. idea make classes abi compatible c types _decimal32, _decimal64 , _decimal128.

now, how getting class decimal32 base classes? guess including incompatible (maybe older) header files, totally different implementation.

update

after investigation, looks guess abi , function decoration right. following code:

#include <decimal/decimal> using namespace std::decimal;   //this synonym of c99 _decimal32, not directly available in c++ typedef float decimal32 __attribute__((mode(sd)));  void foo(decimal32 a) {} void foo(decimal32 a) {} 

gives curious error:

/tmp/ccr61gna.s: assembler messages: /tmp/ccr61gna.s:1291: error: symbol `_z3foodf' defined 

that is, compiler front-end sees no problem in overload , emits asm code, since both functions decorated same assembler fails.

now, non-conformance of gcc, ben voigt suggests in comments? don't know... should able write overloaded functions 2 different types want. otoh, impossible decimal32 type without using compiler extension, meaning of type implementation defined...