Le 2011-10-24 à 17:58:00, IOhannes m zmoelnig a écrit :
On 2011-10-24 16:24, Mathieu Bouchard wrote:
-maxsize between 4GB and 6GB will pretend that the size is 4GB less than what you specified.
[...]
-maxsize between 4GB and 6GB will pretend that the size is 8GB less than what you specified.
hein?
A big sorry... although I got the overall point right, I made FOUR mistakes in the mail :
- First, the obvious typo. I wanted to write « between 8GB and 10GB » in
the latter part. Basically, I just meant wrapping around at 2GB, and crossing zero at 4GB... a signed modulo pow(2,32).
- In 64-bit mode, the long type is 64-bit, which I also already knew ; so
in that case, there is no problem with crossing the 2GB limit.
- I didn't actually try it with a C compiler. When I did, I saw that it's
not at all the same as when casting from a big int type to a smaller int type. When I cast from float to a 32-bit int, any overflow gives -2147483648 instead of wrapping around.
- Pd's check for negative numbers does not replace them by 0, it posts
the «usage» message and refuses to do work. I had already verified this a few days ago and forgot it.
In any case, there is still the problem that if the user does -maxsize 1e+10 on a 32-bit platform, there's nothing that tells him what's wrong. The user just gets an uninformative error message that implies (without actually stating) that something in the syntax is wrong.
| Mathieu BOUCHARD ----- téléphone : +1.514.383.3801 ----- Montréal, QC