rand vs rand48 vs random

[Lawrence D¹Oliveiro]

glibc seems to have 3 different lots of pseudorandom number generator
calls. Which one would people recommend to use?

There's old rand/rand_r/srand, which is probably the most popular, but
only has 31 bits of state. Then there's the rand48 calls (drand48 et
al), which have 48 bits of state--much better. Finally there's
random/srandom/initstate/setstate, which allows for up to 496 bits of
internal state! Unfortunately, these only allow specifying a 32-bit
seed, which seems dumb.

All in all, I think the rand48 calls look like the best ones to use.
There is this note near the end of the man page:

These functions are declared obsolete by SVID 3, which states that
rand(3) should be used instead.

But then, SVID 3 itself is obsolete, so I expect this can be ignored.

Here's a function I came up with initialize the rand48 seed:

bool InitRand48
(
bool TrueRandom
)
/* reads 6 bytes from /dev/random (if TrueRandom) or /dev/urandom
(if not TrueRandom) and passes these to seed48. Returns true iff
successful. errno will be set on failure. */
{
bool Success;
int RandFile;
ushort Seed[3];
RandFile = NilFileFD;
Success = false; /* to begin with */
do /*once*/
{
RandFile = open(TrueRandom ? "/dev/random" : "/dev/urandom",
O_RDONLY);
if (RandFile < 0)
break;
const ssize_t BytesRead = read(RandFile, (void *)Seed, 6);
if (BytesRead < 6)
break;
(void)seed48(Seed);
/* All done */
Success = true;
}
while (false);
CloseFileVar(&RandFile);
return
Success;
} /*InitRand48*/

[David Schwartz]

Post by Lawrence D¹Oliveiro
glibc seems to have 3 different lots of pseudorandom number generator
calls. Which one would people recommend to use?

I wouldn't recommend using any of them. There's plenty of code out there
for PRNGs and RNGs. Choosing your own has two huge benefits. First, you can
make sure it meets your requirements. Second, you can make sure it works the
some on other platforms.

DS

[Lawrence D¹Oliveiro]

Post by David Schwartz

Post by Lawrence D¹Oliveiro
glibc seems to have 3 different lots of pseudorandom number generator
calls. Which one would people recommend to use?

I wouldn't recommend using any of them.

In that case, the simplest thing to do would be to leave a file
descriptor open on /dev/urandom, and read n bytes from it every time you
need them.