msvcp90/msvcp90.h: right-hand operand of comma expression has no effect

[Gerald Pfeifer]

Building git head with GCC development sources (soon to be GCC 4.9.0)
gives me the following warnings

../msvcp90/msvcp90.h:529:75: warning: right-hand operand of comma expression has no effect [-Wunused-value]
../msvcp90/msvcp90.h:530:77: warning: right-hand operand of comma expression has no effect [-Wunused-value]

several times.

Digging into this in more detail I see

  ../msvcp90/ios.c: In function 'basic_streambuf_char__Xsgetn_s':
  ../msvcp90/msvcp90.h:529:75: warning: right-hand operand of comma expression
  has no effect [-Wunused-value]
  #define memcpy_s( dst, size, src, count ) (memcpy( (dst), (src), (count) ), 0)
                                                                            ^
  ../msvcp90/ios.c:1218:13: note: in expansion of macro 'memcpy_s'
             memcpy_s(ptr+copied, size, *this->prpos, chunk);
             ^

Someone must have written that code intentionally, so I am wondering
how to best address this.

Gerald

[Paul Chitescu]

On Wed Apr 16, 2014 00:48:00 Gerald Pfeifer wrote:
> Building git head with GCC development sources (soon to be GCC 4.9.0)
> gives me the following warnings
>
> ../msvcp90/msvcp90.h:529:75: warning: right-hand operand of comma
> expression has no effect [-Wunused-value] ../msvcp90/msvcp90.h:530:77:
> warning: right-hand operand of comma expression has no effect
> [-Wunused-value]
>
> several times.
>
> Digging into this in more detail I see
>
> ../msvcp90/ios.c: In function 'basic_streambuf_char__Xsgetn_s':
>   ../msvcp90/msvcp90.h:529:75: warning: right-hand operand of comma
> expression has no effect [-Wunused-value]
>   #define memcpy_s( dst, size, src, count ) (memcpy( (dst), (src), (count)
> ), 0) ^ ../msvcp90/ios.c:1218:13: note: in expansion of macro 'memcpy_s'
> memcpy_s(ptr+copied, size, *this->prpos, chunk);
>              ^
>
> Someone must have written that code intentionally, so I am wondering
> how to best address this.
>
> Gerald

memcpy_s() returns an errno_t so the author tried to force returning zero (no 
error).

A simple replacement (still without any checks memcpy_s does) would be:

#define memcpy_s( dst, size, src, count ) (!memcpy( (dst), (src), (count) ))

Paul

[Gerald Pfeifer]

On Wed, 16 Apr 2014, Paul Chitescu wrote:
> > Digging into this in more detail I see
> >
> > ../msvcp90/ios.c: In function 'basic_streambuf_char__Xsgetn_s':
> >   ../msvcp90/msvcp90.h:529:75: warning: right-hand operand of comma
> > expression has no effect [-Wunused-value]
> >   #define memcpy_s( dst, size, src, count ) (memcpy( (dst), (src), (count)
> > ), 0) ^ ../msvcp90/ios.c:1218:13: note: in expansion of macro 'memcpy_s'
> > memcpy_s(ptr+copied, size, *this->prpos, chunk);
> >              ^
> > memcpy_s() returns an errno_t so the author tried to force returning 
> > zero (no error).
>
> A simple replacement (still without any checks memcpy_s does) would be:
>
> #define memcpy_s( dst, size, src, count ) (!memcpy( (dst), (src), (count) ))

I tried to make this change, but that now gets:

  ../msvcp90/msvcp90.h:529:44: warning: value computed is not used [-Wunused-value]
  #define memcpy_s( dst, size, src, count ) (!memcpy( (dst), (src), (count) ))
                                             ^

Since, clearly, the result is never intended to be used, we can as
well simply cast the result away instead of using ", 0".

This is what the patch below does.  Tested successfully with GCC 4.9.0.

Gerald

diff --git a/dlls/msvcp90/msvcp90.h b/dlls/msvcp90/msvcp90.h
index 0d5e124..0691141 100644
--- a/dlls/msvcp90/msvcp90.h
+++ b/dlls/msvcp90/msvcp90.h
@@ -526,7 +526,7 @@ typedef struct {
 } complex_double;
 
 #if _MSVCP_VER < 80
-#define memcpy_s( dst, size, src, count ) (memcpy( (dst), (src), (count) ), 0)
+#define memcpy_s( dst, size, src, count ) ((void)memcpy( (dst), (src), (count) ))
 #define memmove_s( dst, size, src, count ) (memmove( (dst), (src), (count) ), 0)
 #define mbstowcs_s( ret, wcs, size, mbs, count ) (mbstowcs( (wcs), (mbs), (count) ), 0)
 #define hypotf( x, y ) ((float)hypot( (double)(x), (double)(y) ))