From: david.brown@hesbynett.no   
      
   On 04/01/2026 01:39, Keith Thompson wrote:   
   > David Brown  writes:   
   >> On 02/01/2026 23:18, Chris M. Thomasson wrote:   
   > [...]   
   >>> typedef void (*generic_func_ptr)(void)   
   >>   
   >> There is no generic function pointer type equivalent to "void *" - you   
   >> are always going to need casts when converting between different   
   >> pointer types.  And you /really/ need to make sure you convert to   
   >> exactly the correct type before calling the pointed-to function.  But   
   >> the definition you gave here is commonly used when people want to have   
   >> generic function pointers.   
   >   
   > Yes, but ...   
   >   
   > There is no generic function pointer type equivalent to void* in the   
   > sense that values can be converted to and from the type implicitly.   
   > For example:   
   >   
   >      int n = 42;   
   >      void *vp = &n;   // implicit conversion   
   >      double *dp = vp; // implicit conversion, potentially dangerous   
   >   
   > There are no such implicit conversions for function pointer types.   
   >   
      
   Yes.   
      
   > However, in the sense of a round-trip conversion yielding the   
   > original result *all* function pointer types can be treated as   
   > generic function pointer types.  You just have to use casts to do   
   > any conversions.   
   >   
      
   Yes.  This also means there is no generic function pointer type   
   equivalent to "void *" for object pointers, since there is no one   
   function pointer type that stands out, no special case, and no common   
   agreement on the type.  Some people will use "void (*)(void)", some will   
   use "int (*)(int)", some "int (*)()", and so on.  That is another sense   
   in which there is no function pointer type equivalent to "void *".   
      
   > It's very common for all pointer types in a given implementation   
   > to have the same representation, and for round-trip conversions   
   > to safely yield the original value, but it's not guaranteed by   
   > the language, and there are implementations where, for example,   
   > function pointer types are bigger than object pointer types, or   
   > void* is bigger than int*.  Certain collections of pointer types   
   > are guaranteed by the language to have the same representation:   
   >   
   > - void*, char*, signed char*, unsigned char*   
   > - All struct pointer types   
   > - All union pointer types   
   > - All function pointer types   
   >   
   > Incidentally, if I want a generic function pointer type, I might   
   > define it so that it can't be used accidentally without a cast.   
   > For example:   
   >   
   >      typedef struct dummy__ (generic_function)(void);   
   >   
   > where struct dummy__ is an incomplete type.  (Note that I've   
   > typedef'ed the function type, not the pointer type.)  But that   
   > might be considered overkill; treating a void function with   
   > no parameters as generic is probably good enough.   
   >   
      
   The dummy struct idea is nice - it could help avoid accidents.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)