Re: gcc 3.4.4 cast warnings?

[James Youngman]

On Sat, Jul 30, 2005 at 10:38:28AM +0200, Oskar Liljeblad wrote:

> He *was* compiling it using g++. Is there a GNU coding standard
> statement for 'void *' casts? 

No, http://www.gnu.org/prep/standards/standards.html says nothing on
this subject.  While some code in there does show a cast to char* for
the return value of malloc(), that could just as easily be a hangover
from the days when the system headers didn't always include a
declaration.

> If you look in the libc documentation you'll find casts for malloc
> in some places, some places not.

My strong recommendation would be that for C code casts from void*
should be avoided.  The problem is that every time you make a cast to
(foo*) from some other type, you have set a trap for someone.  In five
years when someone modifies the function that is being called to
return some type not compatible with (foo*), the compiler will not be
able to warn of the problem because the explicit cast silences the
warning.  I much prefer software which is constructed in such a way as
to allow the compiler to help the programmer eliminate bugs.

Not having the cast will cause a problem for C++.  For such cases if a
lot of use of C++ is planned, it may be worthwhile to consider a
wrapper function:

static 
char* getbuffer(size_t siz)
{
#ifdef __cplusplus
        return new char[siz];   
        // or
        return (char*)malloc(siz);
#else
        return malloc(siz);
#endif
}

static 
char* discardbuffer(char *buf)
{
#ifdef __cplusplus
        delete[] buf;;  
        // or
        free(buf);
#else
        free(buf);
#endif
}

Regards,
James.