theora - Jun 2017 - Building theora 1.1.1 with mingw-w64-gcc 7.1 and msys

Hello,

I recently attempted to build theora 1.1.1 with mingw-w64-gcc 7.1 and 
msys and it fails to build the encoder_example.c example program. There 
are multiple declarations of the function 'rint'. The source file 
created its own version of the function that rounds AWAY from zero. 
MinGW-W64 has its own version of the 'rint' function, which does not 
round away from zero. Therefore, encoder_example.c's rint function is 
necessary, but needs to be renamed because it includes math.h, which 
defines "extern __cdecl double rint(double)".

Here is the relevant function from encoder_example.c lines 50-62 :

#ifdef _WIN32
/*supply missing headers and functions to Win32. going to hell, I know*/
#include <fcntl.h>
#include <io.h>

static double rint(double x)
{
   if (x < 0.0)
     return (double)(int)(x - 0.5);
   else
     return (double)(int)(x + 0.5);
}
#endif

Is this fixed in more recent versions? Should I use 1.2 alpha? Should I 
build from the latest repository source? Ideas?

I'm willing to submit a patch, but I don't know what the resulting 
function should be renamed.

Edgar

Hi people,

I fixed the problem by renaming the function and every instance of 
'rint' to 'rint_az' (Round Int Away from Zero) and then 
encoder_example.c built successfully.

Don't know if this needs to be patched upstream or not.

Edgar



On 6/1/2017 2:53 PM, Edgar Reynaldo wrote:
> Hello,
>
> I recently attempted to build theora 1.1.1 with mingw-w64-gcc 7.1 and 
> msys and it fails to build the encoder_example.c example program. 
> There are multiple declarations of the function 'rint'. The source 
> file created its own version of the function that rounds AWAY from 
> zero. MinGW-W64 has its own version of the 'rint' function, which
does
> not round away from zero. Therefore, encoder_example.c's rint function 
> is necessary, but needs to be renamed because it includes math.h, 
> which defines "extern __cdecl double rint(double)".
>
> Here is the relevant function from encoder_example.c lines 50-62 :
>
> #ifdef _WIN32
> /*supply missing headers and functions to Win32. going to hell, I know*/
> #include <fcntl.h>
> #include <io.h>
>
> static double rint(double x)
> {
>   if (x < 0.0)
>     return (double)(int)(x - 0.5);
>   else
>     return (double)(int)(x + 0.5);
> }
> #endif
>
> Is this fixed in more recent versions? Should I use 1.2 alpha? Should 
> I build from the latest repository source? Ideas?
>
> I'm willing to submit a patch, but I don't know what the resulting 
> function should be renamed.
>
> Edgar
>
>
>

On 2017-06-01 3:40 PM, Edgar Reynaldo wrote:
> I fixed the problem by renaming the function and every instance of
> 'rint' to 'rint_az' (Round Int Away from Zero) and then
> encoder_example.c built successfully.
> 
> Don't know if this needs to be patched upstream or not.
Upstream has the same code, so I expect the fix is needed there too.

Renaming it means always using the replacement, while a libc which
provides it might have a better implementation. Is there a preprocessor
define we could use to distinguish it? e.g.

#if defined(_WIN32) && !defined(__MINGW__)

or something similar. Or you could add a configure check for HAVE_RINT.
Note that quite a few of the examples use it.

Alternatively, it looks like Visual Studio provides rint these days, so
just removing the replacement code might be a viable option.

 -r