"Ian Sutton" <i...@openbsd.org> wrote:
>
> An important thing to ask yourself before suggesting things like this
> is "if this is such an obvious and trivial improvement, then why
> hasn't anyone already done it?". To put things in perspective, we had
> an entire release primarily predicated upon increasing the width of a
> similar type, time_t, from 32 -> 64 bits:
>
> https://www.openbsd.org/55.html
Obvious, perhaps. But it sure ain't trivial -- I'm very much aware of
that.
Okay, in theory, it's changing a couple of lines in a header file, and
everything should just magically work. But that's totally unrealistic
in practice, of course. In the case of size_t, to begin w/, there
appears to be lots of code that relies on the assumption that it's
really a simple unsigned long.
Which is true. Until the change is made.
On a related note, would you folks be interested in patches removing
said assumpting of equivalence from programs like dd(1)?
--schaafuit.
