ksh's ${RANDOM}

[Paul de Weerd <weerd@weirdnet.nl>]

List,

After some discussion with a friend, it turns out that the default ksh
as shipped with OpenBSD has a ${RANDOM} which isn't very random :

$ for d in `jot 10`; do echo -n ${RANDOM}\ ; done; echo
24831 11980 17685 42 3611 27576 7569 29942 28919 2916 

If you're not that fast on the numbers, check this out :

$ for d in `jot 30`; do echo -n $((${RANDOM} % 2))\ ; done; echo
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 

This appears to be coming from /usr/src/bin/ksh/var.c - it uses the
rand() function. From rand(3) :
     rand, srand - bad random number generator

and

DESCRIPTION
     These interfaces are obsoleted by random(3).


So this might not be the best choice. I patched up ksh's var.c to use
random() in stead of rand(). Then I noticed revision 1.17 in CVS, this
has a similar (better, it has the option to revert to the rand()
behaviour by seeding the RNG) patch so I figured I should be safe.

However, on my newly installed machine (snapshot from 20040601), I get
the same behaviour with ksh's $RANDOM producing alternating odd and
even numbers.

...

OK .. after a lot of debugging (newbie programmer - sorry) I found the
problem. How about the following patch :


Index: main.c
===================================================================
RCS file: /cvs/src/bin/ksh/main.c,v
retrieving revision 1.26
diff -u -r1.26 main.c
--- main.c	8 Jan 2004 05:43:14 -0000	1.26
+++ main.c	7 Jun 2004 22:52:12 -0000
@@ -254,9 +254,6 @@
 	}
 	ppid = getppid();
 	setint(global("PPID"), (long) ppid);
-#ifdef KSH
-	setint(global("RANDOM"), (long) (time((time_t *)0) * kshpid * ppid));
-#endif /* KSH */
 	/* setstr can't fail here */
 	setstr(global(version_param), ksh_version, KSH_RETURN_ERROR);
 

Revision 1.17 of var.c introduces the option to restore the old
behaviour by seeding the RNG. Unfortunately, this seeding is done in
main.c. The patch above removes this standard seeding (I don't think
it's required with arc4random). No changes to the manpage are
required.

Comments ?

Cheers,

Paul 'WEiRD' de Weerd

-- 
>++++++++[<++++++++++>-]<+++++++.>+++[<------>-]<.>+++[<+
+++++++++++>-]<.>++[<------------>-]<+.--------------.[-]
                 http://www.weirdnet.nl/