2008/7/31 Szabolcs Nagy <[EMAIL PROTECTED]>: > On 7/31/08, Filippo Erik Negroni <[EMAIL PROTECTED]> wrote: > morale of the story: > the c integer type system is not strong enough to add hints about > which int is used in which context and how, or to write absolutely > architecture independent code. > > the good news is: > usually we don't want to do that (adding hints just makes the code > less readable and applications has a very restricted audience eg. dwm > doesn't want to be more portable than X) > >
My question was intentionally provocatory, and I agree with your answer. I have used size_t in many contexts with ptrdiff_t, and I have recently come to a similar conclusion. In 'kelp', I play around with off_t and size_t a lot, and it has started to give me headaches. My binary tree scan had to be ajdusted several times to allow for size_t wraparound. I have now decided to just check any 'input' arbitrary boundaries and make sure they fit within the boundaries I have chosen. I then make it possible for the user to redefine those boundaries via #defines or typedefs if they think they ever need more. It is an interesting problem because this inconsistency is part of the standard in c99 and I am struggling to come to a conclusion as to why this is the case. -- Cheers, Filippo
