Re: [Boost-testing] [math] How to disable tests which use NTL, GMP, E_FLOAT packages?

Subject: Re: [Boost-testing] [math] How to disable tests which use NTL, GMP, E_FLOAT packages?
From: Sergey Sprogis (sergey.sprogis_at_[hidden])
Date: 2014-12-30 12:06:07

On 12/30/14, 5:03 AM, John Maddock wrote:
>> I'd like to run math regression unit tests on a system that does not
>> have any of 3 extra packages defined inside libs/math/test/Jamfile.v2 
>> as:
>>
>> local ntl-path = [ modules.peek : NTL_PATH ] ;
>> local gmp_path = [ modules.peek : GMP_PATH ] ;
>> local e_float_path = [ modules.peek : E_FLOAT_PATH ] ;
>>
>> By default, my bjam's logfile contains a bunch of error messages like:
>>
>> Error: Could not open include file<gmpfrxx.h>
>> Error: Could not open include file<NTL/RR.h>
>> Error: Could not open include file<gmpxx.h>
>> ....
>
> You shouldn't ever see those messages - they're produced as part of 
> the Jamfile's configuration process.  The only time I'm aware of when 
> they show up is if you invoke bjam from inside the Visual Studio IDE - 
> in that case the IDE captures cl compiler output before bjam gets to 
> see it.  I think there may be an environment variable that controls 
> that, but I'm not sure.
>
No John,  I do not use Visual Studio IDE. I just do 'cd math/test' and 
launch simple bjam command there:

  bjam --disable-icu --toolset=sun_cpp11 -d+2 -j4 link=shared 
runtime-link=shared variant=release threading=multi

After that I see that bjam's logfile contains configuration checks with 
wrong "yes" labels.

Performing configuration checks
     ......
     - has_ntl_rr builds        : yes
     - has_mpfr_class builds    : yes
     - has_mpreal builds        : yes
     - has_e_float builds       : yes
     - has_gmpxx builds         : yes
     - 128-bit floatmax_t       : yes

which I guess, triggers "Error: Could not open include file ..." messages.

Probably all those wrong labels have been generated inside 
build/src/build/configure.jam, but how exactly  configuration checks 
work  in this case is beyond my understanding.

Thanks,

Sergey