Re: testing glpk 4.62 on a 64-bit platform

Hello Andrew,

the code is still wrong in minisat.h:

typedef int  lit;

This assumes that int never has more bits then size_t. This is not
guaranteed. The x32 ABI uses 32bit addresses and supports 64bit integers
(https://lwn.net/Articles/456731/). It depends on the compiler if int
has 32 or 64bit.

Change the definition to either

typedef size_t lit;

or

typedef ssize_t lit;

depending on your need to use signed or unsigned integers.

The following line has a typo:
/* vector of 32-bit intergers (added for 64-bit portability) */

The following typedef is also wrong. The size of arrays should not be of
type int but of type size_t.
And int is not guaranteed to be 32bit. Use int32_t if you explicitly
need 32bits.

typedef struct /* veci_t */ {
    int    size; // this should be size_t
    int    cap;  // this should be size_t
    int*   ptr;  // this should be int32_t* if you need 32bit
} veci;

Same bug here:
typedef struct /* vecp_t */ {
    int    size; // this should be size_t
    int    cap;  // this should be size_t
    void** ptr;
} vecp;

Please, carefully review the types used in the whole module.
It should make no assumptions about the relative lengths of pointers and
integers.

Best regards

Heinrich Schuchardt

On 05/24/2017 10:25 AM, Andrew Makhorin wrote:
>> Please see a preliminary non-official release of glpk 4.62 at:
>> http://sourceforge.net/projects/noumenon/files/tmp/
>>
>>
> 
> I'd very appreciate if someone could test the --minisat option on a
> 64-bit platform. To test it just compile and install glpk as usual and
> then enter subdirectory glpk/examples/pbn and run glpsol as follows:
> 
> glpsol --minisat -m pbn.mod -d disney.dat
> 
> If glpsol crashes, please report to [email protected] . Thanks.
> 
> 
> Andrew Makhorin
> 
> 
> _______________________________________________
> Help-glpk mailing list
> [email protected]
> https://lists.gnu.org/mailman/listinfo/help-glpk
>