randomize/0 doesn't work (further comments)

[mskala]

After sending the message below I took another look at the code and
realized that even with the absolute instead of the relative time, line
790 of machine.c re-seeds the random number generator with an integer in
the range 0..255; so there are only 256 possible seeds for a generator
that ought to be able to do much better than that.

I don't understand why line 790 exists at all; it looks to me like it
would be safe, and sensible, to delete that line entirely.  Another
possibility would be to replace the yucky hardcoded constant 256 with
RAND_MAX, so we don't lose more information than necessary by re-seeding
the generator.
-- 
Matthew Skala
address@hidden                    Embrace and defend.
http://ansuz.sooke.bc.ca/

---------- Forwarded message ----------
Date: Sat, 17 Aug 2002 17:55:15 -0400 (EDT)
From: address@hidden
To: address@hidden
Subject: randomize/0 doesn't work (fix included)

I have a program that calls randomize/0 in its initialization.  It usually
gets the same random seed, because although randomize/0 is documented as
using "a random value depending on the absolute time", it actually gets
its seed by calling M_Real_Time(), which is the number of milliseconds
since the process started.

If I compile my code and put randomize/0 in the initialization, then as
long as my system load stays reasonably low, the number of milliseconds
between process start and the randomize/0 call will be the same every
time, and so I get the same seed value every time.  That's not what I
want, nor what the documentation leads me to expect.

I suggest changing line 789 of src/EnginePL/machine.c from:

  M_Set_Seed(M_Real_Time());

to

  M_Set_Seed(M_Real_Time() + start_real_time);

That way it uses the current time, as documented.
-- 
Matthew Skala
address@hidden                    Embrace and defend.
http://ansuz.sooke.bc.ca/