PHP mt_srand no longer seeds same number!!!!! aaarrrg

[A]

Hi all.

Is this a bug or what???

here is a simple code:

<?php
mt_srand(1);
echo mt_rand(0, 255)."<br />";
echo mt_rand(0, 255)."<br />";
echo mt_rand(0, 255)."<br />";
echo mt_rand(0, 255)."<br />";
?>

It should produce the SAME sequence every time when run.

however... it does not on PHP 5.2.6 on the linux box.

But on my test environment (windows) - php 5.2.3 does work properly!

So what am I missing? How do I get the same random sequences based on
mt_rand and mt_srand? I need same sequences based on starting seed!

I looked at the docs and it says:

The Mersenne Twister implementation in PHP now uses a new seeding
algorithm by Richard Wagner.
Identical seeds no longer produce the same sequence of values they did
in previous versions.
This behavior is not expected to change again, but it is considered
unsafe to rely upon it nonetheless.
but it says - AFTER 5.2.1 --- and I have both versions ABOVE 5.2.1...what to
do? ideas? pleeeeeeaseeeee eee! :)

[Jerry Stuckle]

A wrote:

Hi all.

Is this a bug or what???

here is a simple code:

<?php
mt_srand(1);
echo mt_rand(0, 255)."<br />";
echo mt_rand(0, 255)."<br />";
echo mt_rand(0, 255)."<br />";
echo mt_rand(0, 255)."<br />";
?>

It should produce the SAME sequence every time when run.

however... it does not on PHP 5.2.6 on the linux box.

But on my test environment (windows) - php 5.2.3 does work properly!

So what am I missing? How do I get the same random sequences based on
mt_rand and mt_srand? I need same sequences based on starting seed!

I looked at the docs and it says:

The Mersenne Twister implementation in PHP now uses a new seeding
algorithm by Richard Wagner.
Identical seeds no longer produce the same sequence of values they did
in previous versions.
This behavior is not expected to change again, but it is considered
unsafe to rely upon it nonetheless.
but it says - AFTER 5.2.1 --- and I have both versions ABOVE 5.2.1...what to
do? ideas? pleeeeeeaseeeee eee! :)

I don't know about 5.2.3, but it is working as designed in 5.2.6.

Random numbers are exactly that - random (actually, in programming they
can only be pseudo-random). Your problem is you are depending on the
ability to repeat a sequence. That's not a good assumption to make in
any language.

If you need to be able to repeat the sequence, I would suggest you write
your own pseudo-random number generator. That way you can control its
actions. They aren't too hard, and googling should get you several
examples.

--
=============== ===
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
js*******@attgl obal.net
=============== ===

[A]

just discovered that srand and rand behave the same way!

on 5.2.3 they produce identical seq...

on 5.2.6 always different doesn't care about the seed

dunno what to do. how to replace this? is this a php 5.2.6 bug? or a
feature?

[A]

Random numbers are exactly that - random (actually, in programming they

can only be pseudo-random). Your problem is you are depending on the
ability to repeat a sequence. That's not a good assumption to make in any
language.

If you need to be able to repeat the sequence, I would suggest you write
your own pseudo-random number generator. That way you can control its
actions. They aren't too hard, and googling should get you several
examples.

true but that just bloats the code for something that should work. seed
should start from the same point. yes, i need random numbers but with the
same starting point - that's the whole idea. thanks for the answer though.

any other ideas?

[Jerry Stuckle]

A wrote:

>Random numbers are exactly that - random (actually, in programming they
can only be pseudo-random). Your problem is you are depending on the
ability to repeat a sequence. That's not a good assumption to make in any
language.

If you need to be able to repeat the sequence, I would suggest you write
your own pseudo-random number generator. That way you can control its
actions. They aren't too hard, and googling should get you several
examples.

true but that just bloats the code for something that should work. seed
should start from the same point. yes, i need random numbers but with the
same starting point - that's the whole idea. thanks for the answer though.

any other ideas?

It does work. And no, a seed should not start from the same point. Such
an operation would not be even pseudo-random.

What you need is not random numbers - you need a sequence, which is
something entirely different.

Zend finally corrected the code, and you got caught because you depended
on something you shouldn't have.

--
=============== ===
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
js*******@attgl obal.net
=============== ===

[Michael Fesser]

..oO(Jerry Stuckle)

>I don't know about 5.2.3, but it is working as designed in 5.2.6.

Random numbers are exactly that - random (actually, in programming they
can only be pseudo-random). Your problem is you are depending on the
ability to repeat a sequence. That's not a good assumption to make in
any language.

Many languages allow to seed the random number generator, simply because
there are cases where you have to repeat the exact same sequence. You
still get random values, maybe even based on a random seed, but you're
able to reproduce them if necessary. Nothing wrong with that.

If there's such a seed function, then this is expected and documented
behaviour. With PHP's current MT implementation the seed function is
totally useless and the generator not backwards-compatible. I consider
this a bug or at least a really stupid design decision. If they would at
least have made this new behaviour optional.

rand()/srand() still seem to work correctly here (5.2.6):

<?php
srand(42);
for ($i = 3; $i--;) var_dump(rand() );
srand(42);
for ($i = 3; $i--;) var_dump(rand() );
?>

int(71876166)
int(708592740)
int(1483128881)
int(71876166)
int(708592740)
int(1483128881)

Micha

[A]

thanks for all your replies fellas.

i have an idea that's good enough for my purposes - I'll simply use a long
predefined random sequence and start from fixed "seed" point to get always
equal random sequence.

i also though i could use md5 for similar purpose as it also looks very much
random.

however, i do consider srand behaviour odd or a bug. as pseudo-random
numbers - they should behave within a certain pattern - the only thing is
that srand which should seed the random number with specified number should
start from the same point but as it seems it seeds it with random number. as
Michael Fesser says - pseudorandom function is called pseudorandom because
it generates same sequences for the given seed but this is not the case
here.

anyway, problem solved, thanks again but if you have some other ideas, i'll
be checking this post for a while as well.

[Michael Fesser]

..oO(A)

>however, i do consider srand behaviour odd or a bug. as pseudo-random
numbers - they should behave within a certain pattern - the only thing is
that srand which should seed the random number with specified number should
start from the same point but as it seems it seeds it with random number.

Can you test the little script I posted and post the result? Here on
5.2.6 CLI it works as expected.

Micha

[A]

Can you test the little script I posted and post the result? Here on

5.2.6 CLI it works as expected.

I did, first 3 numbers are random and second seq which should be identical
to the first one is again random (and on each reload of the page the first
and second seq are both random again). it's like srand would be commented
out of the code.

I notified server administrators to run the same script on other servers
running php 5.2.6 and will post their test results here.

[Jerry Stuckle]

Michael Fesser wrote:

.oO(Jerry Stuckle)

>I don't know about 5.2.3, but it is working as designed in 5.2.6.

Random numbers are exactly that - random (actually, in programming they
can only be pseudo-random). Your problem is you are depending on the
ability to repeat a sequence. That's not a good assumption to make in
any language.

Many languages allow to seed the random number generator, simply because
there are cases where you have to repeat the exact same sequence. You
still get random values, maybe even based on a random seed, but you're
able to reproduce them if necessary. Nothing wrong with that.

If there's such a seed function, then this is expected and documented
behaviour. With PHP's current MT implementation the seed function is
totally useless and the generator not backwards-compatible. I consider
this a bug or at least a really stupid design decision. If they would at
least have made this new behaviour optional.

rand()/srand() still seem to work correctly here (5.2.6):

<?php
srand(42);
for ($i = 3; $i--;) var_dump(rand() );
srand(42);
for ($i = 3; $i--;) var_dump(rand() );
?>

int(71876166)
int(708592740)
int(1483128881)
int(71876166)
int(708592740)
int(1483128881)

Micha

Micha,

Many languages have a seed function for compatibility only. Believe it
or not, there used to be a day when you didn't have a way to seed a
random number automatically. I still remember having to enter a number
when starting a program because there was no real time clock available.

And a "repeatable sequence of random numbers" is an oxymoron. Yes, many
people have used a random number generator as a quick way to generate a
repeatable sequence. I consider THAT a bug.

--
=============== ===
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
js*******@attgl obal.net
=============== ===