Re: [9front] [PATCH] libc: replace lrand's algorithm

[cinap_lenrek@felloff.net]

> I found (rand()<<16) + rand() in kernel's devsdp, which is
> strange considering lrand already has 32 bits of output. I
> think this should be considered a bug.

The manpage states:

          Rand returns a uniform pseudo-random number x, 0≤ x <2^15.

          Lrand returns a uniform long x, 0≤ x <2^31.

So lrand() only return a positive number (so really, is 31 bit
generator, not 32 bit).

Also, rand() returns only 15-bit integer. so doing rand()<<16
results in a bias as it will never set bit 15.

The lrand() implementation in the kernel is not complient
which i think is a mistake. (BUG!)

libc does not provide a function for unsigned 32-bit.
having one might be a good start.

say, lets define:

ulong ulrand(void)

in range 0<=ulrand()<2^32

and use that as the basis for all the other functions
instead of lrand().

Then as you say, constructs like rand()<<16 | rand()
should be revisited.

--
cinap