| Re: [AD] Pending patches request |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
> srand(time(NULL)) at the start of the program?
Yes.
> Yes I tried it, and it didn't improve randomness in any way.
> The problem is the OSX libc rand() implementation places the most random
> bits in the higher word, while the lower word always gets values that
> repeat somewhat often. On Intel machines it seems to be the contrary.
Endianess, endianess...
> So (rand()+(rand()>>16)) should solve the issue...
But it's ugly. Try to devise an elegant macro to hide this ugliness.
> > This patch is only a last resort solution. We have first to understand
> > why we are not able to manipulate UTF16-BE, especially on big-endian
> > platforms.
>
> I agree; I'll investigate more on this.
I will too. The good news is that I can reproduce the failures on x86 by
switching everything (code and data) to UTF16-BE. And it's a bit weird
because if I swap back the first two bytes of message_it[], I get 'nuti' on
the screen!
--
Eric Botcazou