Re: [Vxl-users] [PATCH] fix for compiler errors with clang with C++11

[Sean McBride <sean@ro...>]

On Fri, 20 Apr 2012 17:47:55 +0100, Peter Vanroose said:

>> Peter, shall I update the patch as Amitha suggests?
>
>Certainly; sounds good to me.

See attached.  Also fixes a little warning I happened to notice.

-- 
____________________________________________________________
Sean McBride, B. Eng                 sean@...
Rogue Research                        http://www.rogue-research.com 
Mac Software Developer              Montréal, Québec, Canada

Re: [Vxl-users] [PATCH] fix for compiler errors with clang with C++11

[Peter Vanroose <peter_vanroose@ya...>]

> Peter, shall I update the patch as Amitha suggests?


Certainly; sounds good to me.

--      Peter.

Re: [Vxl-users] [PATCH] fix for compiler errors with clang with C++11

[Sean McBride <sean@ro...>]

On Fri, 20 Apr 2012 06:35:52 -0400, Amitha Perera said:

>On 4/19/2012 4:39 PM, Sean McBride wrote:
>> I guess the only part people might not like is the "#if __cplusplus > 
>> 199711L" check for C++11. If some compiler passes that test yet does 
>> not provide std::isnan, there will be an error. Some people prefer to 
>> do 'try compile'-type tests of the compiler, but I don't know how to 
>> do that, nor do I know what the vxl preference is. However you want to 
>> do it is fine with me. :) Cheers, 
>
>Sean, that macro test is somewhat inconsistent with the vxl way.

I know little of vxl, so I'll take your word. :)

Peter, shall I update the patch as Amitha suggests?

>seems to me that the check should be
>   #if VXL_CXX11
>      ...
>   #endif
>
>You should then add to vxl/vcl_compiler.h something like
>   #if __cplusplus > 199711L
>      #define VXL_CXX11 1
>   #else
>      #define VXL_CXX11 0
>   #endif
>if this check is expected to work across all compilers, including 
>non-CXX11 compilers. (I.e. if it a standard-defined way to check for a 
>C++11 compiler.)  If it is a check for a specific compiler, like clang, 
>then the stanzas in vcl_compiler.h should be more like
>
>     // Assume the compiler is not CXX 11 by default
>     #define VXL_CXX11 0
>     ...
>     #if <check for clang>
>         # if __cplusplus > 199711L
>             #undef VXL_CXX11
>             #define VXL_CXX11 1
>         #endif
>      #endif
>
>Since the intent seems to be to use this on  a CXX 11 compiler, as 
>opposed to on any compiler that happens to have std::nan, a test in 
>vcl_compiler.h is probably the better way over a try-compile.

FYI, any C++11-conforming compiler must pass the "#if __cplusplus > 199711L" test, see :
<http://www2.research.att.com/~bs/C++0xFAQ.html#0x>;

likewise std::isnan is part of the standard:
<http://www2.research.att.com/~bs/C++0xFAQ.html#C99>;

nothing about my patch is specific to clang, though that's the only compiler I tested with.

Cheers,

-- 
____________________________________________________________
Sean McBride, B. Eng                 sean@...
Rogue Research                        http://www.rogue-research.com 
Mac Software Developer              Montréal, Québec, Canada

Re: [Vxl-users] [Vxl-maintainers] image_view io for components

[Ian Scott <scottim@im...>]

On 19/04/2012 15:54, Antonio Garrido Carrillo wrote:
>
> //------------------------------ USING vil_rgb
> ------------------------------------------
>
> int main(int argc, char *argv[]) {
>
> vil_image_view<vil_rgb<vxl_byte> > img(100,100);
>
> vil_image_view<vxl_byte> img2(vil_plane(img,0));
>
> Print_Info(img,"img:");
> Print_Info(img2,"img2:");
>
> vsl_b_ofstream f("img2.dat");
> vsl_b_write(f,img2);
>
> vil_image_view<vxl_byte> img3;
> vsl_b_ifstream f2("img2.dat");
> vsl_b_read(f2,img3);
>
> return 0;
> }
>

This (vil_rgb) one really ought to work. Try explicitly closing the 
output file to force it to flush, before reading it in again.

In the complex case, try performing a deep copy of the real image before 
saving (and flushing the output file). I could put a fix in the 
vsl_b_write code, to detect this case and do the deep copy internally, 
but that would cause other problems if the user was expecting to save 
both the original and real-sliced image view while maintaining a single 
underlying data chunk.

Let me know how you get on, and I'll give some more thought to a better 
solution - possibly just documenting the behaviour of vil_view_real_part().


Ian.

Re: [Vxl-users] [PATCH] fix for compiler errors with clang with C++11

[Amitha Perera <amitha@th...>]

On 4/19/2012 4:39 PM, Sean McBride wrote:
> I guess the only part people might not like is the "#if __cplusplus > 
> 199711L" check for C++11. If some compiler passes that test yet does 
> not provide std::isnan, there will be an error. Some people prefer to 
> do 'try compile'-type tests of the compiler, but I don't know how to 
> do that, nor do I know what the vxl preference is. However you want to 
> do it is fine with me. :) Cheers, 

Sean, that macro test is somewhat inconsistent with the vxl way.  It 
seems to me that the check should be
   #if VXL_CXX11
      ...
   #endif

You should then add to vxl/vcl_compiler.h something like
   #if __cplusplus > 199711L
      #define VXL_CXX11 1
   #else
      #define VXL_CXX11 0
   #endif
if this check is expected to work across all compilers, including 
non-CXX11 compilers. (I.e. if it a standard-defined way to check for a 
C++11 compiler.)  If it is a check for a specific compiler, like clang, 
then the stanzas in vcl_compiler.h should be more like

     // Assume the compiler is not CXX 11 by default
     #define VXL_CXX11 0
     ...
     #if <check for clang>
         # if __cplusplus > 199711L
             #undef VXL_CXX11
             #define VXL_CXX11 1
         #endif
      #endif

Since the intent seems to be to use this on  a CXX 11 compiler, as 
opposed to on any compiler that happens to have std::nan, a test in 
vcl_compiler.h is probably the better way over a try-compile.

Thanks,
Amitha.