Limits of Random changed due the mersenne twister

Hi all,

I have noticed a regression with master in one of my codes, I try to generate
a random number like:

        (Random between: 100000000000 and: 999999999999) asString


It is working with 3.2, it fails with a primitiveError in master. Is this
something easy to fix?



Log:
Object: Random new: 625 "<0x403de1d0>" error: primitive operation failed
SystemExceptions.PrimitiveFailed(Exception)>>signal (ExcHandling.st:254)
SystemExceptions.PrimitiveFailed class(Exception class)>>signal
(ExcHandling.st:151)
Random(Object)>>primitiveFailed (Object.st:1349)
Random>>nextLimit: (Random.st:141)
Random>>between:and: (Random.st:124)
Random class>>between:and: (Random.st:85)
TOPDBPostgres class(TOPDB class)>>generateId (TOPDB.st:62)
TOPDBPostgres class(TOPDB class)>>newId (TOPDB.st:103)
[] in AddTopUp>>topUp: (AddTopUp.st:36)
BlockClosure>>on:do: (BlkClosure.st:193)
AddTopUp>>topUp: (AddTopUp.st:39)
[] in UndefinedObject>>executeStatements (AddTopUp.st:76)
BlockClosure>>ensure: (BlkClosure.st:269)
UndefinedObject>>executeStatements (AddTopUp.st:73)

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

On Tue, Jul 19, 2011 at 13:22, Holger Hans Peter Freyther
<[hidden email]> wrote:
Yes, it should be enough to add a handler for the failure of
#nextLimit:, and combine a few random integers together before scaling
the result down to the limit value.

Probably the numbers you were getting were not random at all, so it's
good that it now fails. :)

Paolo

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

On 07/19/2011 03:48 PM, Paolo Bonzini wrote:

> Yes, it should be enough to add a handler for the failure of
> #nextLimit:, and combine a few random integers together before scaling
> the result down to the limit value.
>
> Probably the numbers you were getting were not random at all, so it's
> good that it now fails. :)

One more Mersenne Twister question. So if someone knows a row of 624 random
numbers, he knows the internal state of the twister... does it make sense to
reseed more often?

not a crypto expert
        holger

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

Vue complète
Re : [Help-smalltalk] Limits of Random changed due the mersenne twister
De : Mathieu Suen <[hidden email]> Ajouter dans les contacts
À : Holger Hans Peter Freyther <[hidden email]>  
________________________________

----- Message d'origine ----

For me you don't know the state of the twister. This is because the operation:
  y = mt[mt_index];
  y ^= (y >> 11);
  y ^= (y << 7) & 0x9d2c5680;
  y ^= (y << 15) & 0xefc60000;
  y ^= (y >> 18);

is not reversible.


>
> not a crypto expert
>      holger
>
> _______________________________________________
> help-smalltalk  mailing list
> [hidden email]
> https://lists.gnu.org/mailman/listinfo/help-smalltalk
>

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

On Tue, Jul 19, 2011 at 18:05, Holger Hans Peter Freyther
<[hidden email]> wrote:
The Mersenne Twister is _absolutely_ not a cryptographic RNG.  It only
tries hard to produce numbers that look random, it does not prevent
you from reproducing or predicting the sequence.

Paolo

_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk