Description:
------------
When calling mt_srand with seed 0, the resulting sequence is the same as
with seed 1; when calling it with seed 2, the sequence is the same as
with seed 3, etc., generating the same sequences for even numbers as for
these numbers + 1.

The problem seems to come from this line:

register php_uint32 x = (seed | 1U) & 0xFFFFFFFFU, *s = BG(state);

The | 1U is apparently there to force the seed being odd, due to the
fact that the initialization uses a pure multiplicative
linear-congruential generator. Replacing the line:

  *s++ = (x *= 69069U) & 0xFFFFFFFFU);

with e.g.:

  *s++ = (x *= 69069U, ++x) & 0xFFFFFFFFU);

should eliminate the requirement that the seed be odd. The generator X
<- (X*69069+1) mod 2**32 is the 'VAX generator', has decent short-term
randomness properties and works fairly well for this purpose
(Wikipedia's article about MT uses it). The pure multiplicative X <-
(X*69069) mod 2**32 is not so well studied and does not work well with
all seeds. The seeding requirements of MT are just that not all elements
are zero, which is guaranteed in this case.

However, please consider using e.g. the algorithm in init_genrand() in
<http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/CODES/mt19937ar.
c> instead. See justification in
<http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/emt19937ar.html>
.

Reproduce code:
---------------
<?php
  for ($i = 0; $i < 10; $i++) {
    mt_srand($i);
    echo mt_rand(0, mt_getrandmax()), ", ",
         mt_rand(0, mt_getrandmax()), "\n";
  }
?>

Expected result:
----------------
All lines different.

Actual result:
--------------
Lines are equal by pairs.

This bug has been fixed in CVS.

Snapshots of the sources are packaged every three hours; this change
will be in the next snapshot. You can grab the snapshot at
http://snaps.php.net/.
 
Thank you for the report, and for helping us make PHP better.

This issues appears to have already been resolved. With latest 
CVS each line is different.