(random) isn't random

[alexdunn]

The following command always writes 92:

```
sbcl --non-interactive --no-sysinit --no-userinit --eval '(write-line (write-to-string (random 256)))'
```

Within a session, successive invocations of (random 256) will return the same series of numbers: 92, 246, 238, 121, 44, 223, 5...

Tested with SBCL 1.3.7 and with the git HEAD at 5d2d20fd81585b17197b68376ce3d0062e383908 on Mac OS 10.11.5; and on SBCL 1.2.7 on CentOS 7.0.

[Stas Boukarev (stassats)]

Well, you have to seed the random state with (setf *random-state* (make-random-state t))...

[alexdunn (dunn-alex)]

OK, but what's the reason for not doing that automatically? Every other implementation of (random) I tested comes pre-seeded.

[Stas Boukarev (stassats)]

>_Every_ other
That was determined to be incorrect by testing Clozure CL, Clisp, C, C++.

[alexdunn (dunn-alex)]

> I tested

Namely ABCL, CMUCL, ECL.

But I guess I just happened to test an unusual set of implementations.