Re: mrand48 broken on 64-bit systems

[Matt Thomas <matt%3am-software.com@localhost>]

On Oct 22, 2013, at 2:13 AM, Alan Barrett <apb%cequrux.com@localhost> wrote:

>> lrand/nand returns change to:
>> 
>>      return __rand48_seed[2] * 32768 + (__rand48_seed[1] >> 1);
>> 
>> And all casts go away.  the multiply promotes everything to unsigned int.
> 
> Here, I think the multiply will be performed using signed int (in the common 
> case that int is larger than 16 bits), or signed long (in the case that both 
> int and short are 16 bits).  I can't see a case where unsigned int would be 
> used, but I could easily be wrong -- this stuff is confusing.

No, it's unsigned due to __rand48_seed being unsigned.  (confirmed by
checking at the resultant code).

> I think it's much better to use explicit fixed-width types.

Well, it's an ancient interface.  I had considered changing it
to use 64-bit types internally (at least for LP64) but given how
often it's used, I didn't think it was worth the effort.