Strange memory corruption

Hi,

I have a strange case of "memory corruption", if I can call it that way...

I have a struct:

Code:

---

typedef struct
{
        uint64_t localizacion;
        uint64_t tipo;
        float azimuth[64];
        float altura[64];
        float tilt[64];
} individuo;

---

And then, I have two arrays of that type:

Code:

---

individuo *padres, *hijos;

---

Which I initialize witch calloc:

Code:

---

padres = calloc ( 64, sizeof ( individuo ) );
hijos = calloc ( 64, sizeof ( individuo ) );

---

I initialize them with random values:

Code:

---

        for ( i = 0; i < 64; i++ )
        {
                // Localizaciones
                srand ( rand() );
                padres[i].localizacion = ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX ) << 32;
                srand ( rand() );
                padres[i].localizacion = padres[i].localizacion | ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX );

                // Tipo de antena
                srand ( rand() );
                padres[i].tipo = ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX ) << 32;
                srand ( rand() );
                padres[i].tipo = padres[i].tipo | ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX );

                //printf("Individuo %d: %llx - %llx\n",i,padres[i].localizacion,padres[i].tipo);

                for ( j = 0; j < 64;j++ )
                {
                        // Altura
                        srand ( rand() );
                        padres[i].altura[j] = 1 + 14 * ( ( float ) rand() / ( float ) RAND_MAX );

                        // Azimuth
                        srand ( rand() );
                        padres[i].azimuth[j] = 5 * ( ( float ) rand() / ( float ) RAND_MAX );

                        // Tilt
                        srand ( rand() );
                        padres[i].tilt[j] = 15 - 30 * ( ( float ) rand() / ( float ) RAND_MAX );
                }
        }

---

As you can see, altura, tilt and azimuth, have a range of valid values, which I check out every time I modified it (I add a smaller random value).

The problem is that, sometimes, some strange values appears, sometimes 0, sometimes big positive or negative values, or values very close to zero.

I tried to debug the binary, with no results, and using Valgrind, it throws this:

Code:

---

==7468== Invalid read of size 1
==7468==    at 0x4023EF1: memcpy (mc_replace_strmem.c:406)
==7468==    by 0x8049B01: main (dios2.c:415)
==7468==  Address 0x41B6977 is 1 bytes before a block of size 50,176 alloc'd
==7468==    at 0x402195F: calloc (vg_replace_malloc.c:279)
==7468==    by 0x804892A: main (dios2.c:100)

---

just before detecting a strange value (I wrote a function to check the integrity of the values in the arrays).

I don't know if I'm doing something wrong, as, these strange values, doesn't appear on all values, only from time to time, so, if you can point me where could be the problem... ;)

Thanks in advance.

Sorry, can't comment on your actual problem, but:
srand ( rand() );
is probably not a good construct.

--
Mats

1) Do not seed more than once.
2) Seed with the time, since it is guaranteed to be unique, thus giving you unique numbers.

Hi,

Thanks for the responses. Random numbers is not a priority, as I plan to move to dSFMT, and I don't think it could be the probleme of the strange behaviour.

Also, I tried to summarize the code, to make it readable, but there are more code, in fact, the Valgrind report, mentions memcpy, as I used this function to make copies of positions of the padres and hijos arrays.

Code:

---

        for ( i = 0; i < 64; i++ )
        {
                // Localizaciones
                srand ( rand() );
                padres[i].localizacion = ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX ) << 32;
                srand ( rand() );
                padres[i].localizacion = padres[i].localizacion | ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX );

                // Tipo de antena
                srand ( rand() );
                padres[i].tipo = ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX ) << 32;
                srand ( rand() );
                padres[i].tipo = padres[i].tipo | ( uint64_t ) ( ( pow ( 2, 32 ) * rand() ) / RAND_MAX );

                //printf("Individuo &#37;d: %llx - %llx\n",i,padres[i].localizacion,padres[i].tipo);

                for ( j = 0; j < 64;j++ )
                {
                        // Altura
                        srand ( rand() );
                        padres[i].altura[j] = 1 + 14 * ( ( float ) rand() / (RAND_MAX + 1.0));

                        // Azimuth
                        srand ( rand() );
                        padres[i].azimuth[j] = 5 * ( ( float ) rand() / (RAND_MAX + 1.0));

                        // Tilt
                        srand ( rand() );
                        padres[i].tilt[j] = 15 - 30 * ( ( float ) rand() / (RAND_MAX + 1.0));
                }
        }

---

"If srand() makes things more random, then more srand()'s must be SUPER random! Plus with random as its seed, it will be super ultra random!!!!"

Not exactly........ Surely it would help you spot the lack of randomality in the underlying algorithm if anything.

Quote:

---

Originally Posted by kwyjibo

Hi,

Thanks for the responses. Random numbers is not a priority, as I plan to move to dSFMT, and I don't think it could be the probleme of the strange behaviour.

Also, I tried to summarize the code, to make it readable, but there are more code, in fact, the Valgrind report, mentions memcpy, as I used this function to make copies of positions of the padres and hijos arrays.

---

Yes, it's not random that valgrind mentions memcpy - it's highly likely that you made some sort of mistake in that section. I don't see anything (besides from the random number generation) that is questionable in the code you posted. It is fairly common here that the posted code is not where the problem is, so don't feel bad about "posting the wrong bit of code".

--
Mats