[Cryptography] Other obvious issues being ignored?
Thierry Moreau
thierry.moreau at connotech.com
Sun Oct 25 11:01:59 EDT 2015
On 10/23/15 08:58, Peter Gutmann wrote:
>
> [The ggc compiler] claims that it doesn't know that adding 10 to
> INT_MAX on the same CPU that it knows every intimate detail of produces a
> negative value.
>
> This isn't "compliance to the C standard", this is just the gcc developers
> being idiots.
>
May I suggest that this language is not ideal for having a feature
request being handled with the best hope of client satisfaction.
As I understand it, a gcc release formally introduced an indefinite
precision arithmetic library for the compile-time arithmetic
computations. (Maybe a benefit is that CPU-architecture-independent
optimizations would be easier to implement/validate/debug with this
perfect integer arithmetic abstraction).
Your feature request would be compiler switches for compile-time integer
arithmetic aligned with the target CPU under a few defined contexts (to
be discussed between *polite* feature requestors and compiler writers).
Incidentally I have huge respect for the gcc developers as the most
significant piece of software for the whole information society.
- Thierry
More information about the cryptography
mailing list