[SCIP] Errors in make test for x32 linux platform for scip 4.0.0

[David Rua]

Dear Stefan,

I can confirm that the double casting solved the problem.
Thanks (also to Micha W.) for the suggestion.

Best,
David

-----Original Message-----
From: Stefan Vigerske [mailto:stefan at math.hu-berlin.de] 
Sent: 20 de abril de 2017 16:30
To: David Rua <drua at inesctec.pt>; scip at zib.de
Subject: Re: [SCIP] Errors in make test for x32 linux platform for scip
4.0.0

Hi,

please try with the following change (adding a (double) cast to the
definition of SCIP_INVALID and SCIP_UNKNOWN):

diff --git a/src/scip/def.h b/src/scip/def.h index f58814939..17a943f32
100644
--- a/src/scip/def.h
+++ b/src/scip/def.h
@@ -153,8 +153,8 @@ extern "C" {
  #define SCIP_DEFAULT_HUGEVAL          1e+15  /**< values larger than 
this are considered huge and should be handled separately (e.g., in activity
computation) */
  #define SCIP_MAXEPSILON               1e-03  /**< maximum value for 
any numerical epsilon */
  #define SCIP_MINEPSILON               1e-20  /**< minimum value for 
any numerical epsilon */
-#define SCIP_INVALID                  1e+99  /**< floating point value 
is not valid */
-#define SCIP_UNKNOWN                  1e+98  /**< floating point value 
is not known (in primal solution) */
+#define SCIP_INVALID          (double)1e+99  /**< floating point value 
is not valid */
+#define SCIP_UNKNOWN          (double)1e+98  /**< floating point value 
is not known (in primal solution) */


  #define REALABS(x)        (fabs(x))



Thanks go to Micha W. to figure this out.
(On 32bit, with pure C99,
   double a=SCIP_INVALID; assert(a==SCIP_INVALID); fails before this
change.)

Best,
Stefan



On 04/18/2017 05:52 PM, David Rua wrote:
> Dear all,
> 
>   
> 
> I'm trying to use the new SCIP 4.0 on a linux x32 platform. For that 
> matter the scipoptsuite-4.0.0 code is being directly compiled to 
> generate the necessary *.so or *.a lib files to be used afterwards.
> 
> When building the lib files via "make" no error is produced and the 
> lib files are created but when running "make test" there are tests that
fail.
> 
> When trying the same approach for a linux x64 version the compiling 
> runs with no error and the make test returns success on all tests.
> 
>   
> 
> Digging into the output we can see that on the first test the failure 
> comes from the reading of the linking.cip file.
> 
> At the end there is a mix of failures for the failed tests
> 
>   
> 
> Anyone had the same issue or have an idea why the 32-bit compiling is 
> not working?
> 
>   
> 
> You can find attached the output of the "make test" command for both 
> 32- and 64-bit platform.
> 
>   
> 
> Thanks.
> 
> Best,
> 
> David
> 
>   
> 
> PS: On a side note to say that the compiling and execution of v3.2.1 
> run as expected
> 
>   
> 
> 
> 
> David Rua
> Centre for Power and Energy Systems
> 
> INESC TEC
> Campus da FEUP
> Rua Dr Roberto Frias
> 4200-465 Porto
> Portugal
> 
> T +351 22 209 4209
> F +351 22 209 4050
>   <mailto:drua at inesctec.pt> drua at inesctec.pt
>   <http://www.inesctec.pt/> www.inesctec.pt
> 
>   
> 
>   
> 
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
> 


--
http://www.gams.com/~stefan