Re: [SoX-users] Remove _fstati64 and _stati64 defs for mingw32

[Ulrich K. <ul...@ch...>]

Finally got around to read this not-too-recent thread ...

Motiejus Jak?tys <des...@gm...>:

> Reason: it did not compile on mingw32 4.2.1:

> [  1%] Building C object src/CMakeFiles/libsox.dir/formats_i.obj
> In file included from  
> /home/motiejus/code/stuff/sox/src/formats_i.c:23:  
> /usr/lib/gcc/i586-mingw32msvc/4.2.1-sjlj/../../../../i586-mingw32msvc/include/sys/stat.h:122: error: redefinition of ?struct  
> _stati64?
> make[2]: *** [src/CMakeFiles/libsox.dir/formats_i.obj] Error 1
> make[1]: *** [src/CMakeFiles/libsox.dir/all] Error 2
> make: *** [all] Error 2

> The way I see it is that on _certain_ systems (those that have _stati64
> defined), the following snippet of util.h takes effect
> 	#ifdef _stati64
> 	#define stat _stati64
> 	#else
> 	#define stat _stat
> 	#fi

I found a hint on the mythtv-dev mailing list how this may cause the  
double definition of struct _stati64: When the #define is in effect,  
the preprocessor turns a definition of "struct stat" into a (second)  
definition of "struct _stati64". It certainly looks from  
http://paste.ubuntu.com/1256655/ like this is the case. However ...

Jan Stary <ha...@st...>:

> On Oct 02 15:48:14, des...@gm... wrote:
> > So as far as I see it includes "util.h" _before_ <sys/stat.h>. At  
> least > in formats_i.c.
> Then again, "util.h" itself includes <sys/stat.h> at the top.

... this is true, so I didn't understand at first either how this  
could cause the problem. But then I noticed that the include in util.h  
is conditional: It only includes sys/stat.h if HAVE_SYS_STAT_H is set.  
And it seems we don't check for sys/stat.h at all when using cmake,  
but only when using the GNU build system.

Motiejus, could you verify that this is indeed the problem? I've  
prepared a small patch, but you could also simply remove the #ifdef  
around the include in util.h.

Ulrich