mex should treat C files as C++

[Søren Hauberg]

Hi All,
  Once in a while I have to use mex files that people distribute. Or
rather, I have to compile the source code they distribute using the
'mex' command. A couple of times I have been struck be a small issue:
Some people seem to use non-C keywords (e.g. 'true' and 'false') in
their code, but they save the source code in a file that ends with
'.c' (like 'my_function.c'). When I run the 'mex' command I get errors
like these:

  kernel_function.c: In function ‘mexFunction’:
  kernel_function.c:37: fejl: ‘true’ undeclared (first use in this
function)
  kernel_function.c:37: fejl: (Each undeclared identifier is reported
only once
  kernel_function.c:37: fejl: for each function it appears in.)
  kernel_function.c:44: fejl: ‘false’ undeclared (first use in this
function)

The problem goes away if I rename the files to something that ends with
'.cc'. I guess this means 'mkoctfile' (which is called by 'mex') uses
'gcc' and not 'g++'. Could we change this, such that 'g++' is called
instead of 'gcc'?

Søren

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[dbateman]

Søren Hauberg wrote:

Hi All,
  Once in a while I have to use mex files that people distribute. Or
rather, I have to compile the source code they distribute using the
'mex' command. A couple of times I have been struck be a small issue:
Some people seem to use non-C keywords (e.g. 'true' and 'false') in
their code, but they save the source code in a file that ends with
'.c' (like 'my_function.c'). When I run the 'mex' command I get errors
like these:

  kernel_function.c: In function ‘mexFunction’:
  kernel_function.c:37: fejl: ‘true’ undeclared (first use in this
function)
  kernel_function.c:37: fejl: (Each undeclared identifier is reported
only once
  kernel_function.c:37: fejl: for each function it appears in.)
  kernel_function.c:44: fejl: ‘false’ undeclared (first use in this
function)

The problem goes away if I rename the files to something that ends with
'.cc'. I guess this means 'mkoctfile' (which is called by 'mex') uses
'gcc' and not 'g++'. Could we change this, such that 'g++' is called
instead of 'gcc'?

Søren

true and false are not valid C keywords. Adding

#define true 1
#define false 0

somewhere, but keep compiling the code with gcc would be a cleaner solution.

D.

[Søren Hauberg]

tor, 08 01 2009 kl. 05:54 -0800, skrev dbateman:
> true and false are not valid C keywords. Adding
>
> #define true 1
> #define false 0
>
> somewhere, but keep compiling the code with gcc would be a cleaner solution.

Okay, so I tried compiling the offending code in Matlab, and it breaks
there as well. I've been struck by this problem with many peoples mex
code so I thought Matlab did something weird like always compiling with
a C++ compiler, but it turns out they don't. I guess Visual Studio
allows this behaviour which just makes people sloppy.

Anyway, sorry about the noise...

Søren

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[Thomas Weber-8]

Am Donnerstag, den 08.01.2009, 16:00 +0100 schrieb Søren Hauberg:

> tor, 08 01 2009 kl. 05:54 -0800, skrev dbateman:
> > true and false are not valid C keywords. Adding
> >
> > #define true 1
> > #define false 0
> >
> > somewhere, but keep compiling the code with gcc would be a cleaner solution.
>
> Okay, so I tried compiling the offending code in Matlab, and it breaks
> there as well. I've been struck by this problem with many peoples mex
> code so I thought Matlab did something weird like always compiling with
> a C++ compiler, but it turns out they don't. I guess Visual Studio
> allows this behaviour which just makes people sloppy.

You can include stdlib.h.

        Thomas

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[Thomas Weber-8]

On Thu, Jan 08, 2009 at 04:25:55PM +0100, Thomas Weber wrote:

> Am Donnerstag, den 08.01.2009, 16:00 +0100 schrieb Søren Hauberg:
> > tor, 08 01 2009 kl. 05:54 -0800, skrev dbateman:
> > > true and false are not valid C keywords. Adding
> > >
> > > #define true 1
> > > #define false 0
> > >
> > > somewhere, but keep compiling the code with gcc would be a cleaner solution.
> >
> > Okay, so I tried compiling the offending code in Matlab, and it breaks
> > there as well. I've been struck by this problem with many peoples mex
> > code so I thought Matlab did something weird like always compiling with
> > a C++ compiler, but it turns out they don't. I guess Visual Studio
> > allows this behaviour which just makes people sloppy.
>
> You can include stdlib.h.

Urgs, that should be stdbool.h.

        Thomas
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave