L. David Baron wrote: > I'm not sure what the reason to prefer size_t/ptrdiff_t vs. > intptr_t/uintptr_t are (and I hope there aren't any cases where > they're different). I suppose size_t and ptrdiff_t have been around > longer, but intptr_t and uintptr_t have clearer names. But I also > tend to think we should only use one set of them.
[u]intptr_t is an integer type that holds pointers. ptrdiff_t is the integer type that holds the result of pointer arithmetic. Because they have different meanings, the distinction between the two types to document the code and to (in theory) help static analysis. It may be helpful to ensure that the range of these types (and ssize_t/size_t) is the same (via static_assert) but we should still use each distinct type as appropriate. uintptr_t x; ptrdiff_t y; It is expected that I can convert x contains the value of some pointer and that casting it to a pointer is meaningful. It does not make sense to add x to any pointer or any other value of type uintptr_t. It is not expected that y contains the value of a pointer and so it is not expected that casting it to a pointer is meaningful. It may or may not make sense to add y to some particular pointer (or value of uintptr_t) or to some other value of type ptrdiff_t. The compiler should (but probably doesn't) warn you when you cast the result of pointer arithmetic to unintptr_t, or when you cast a pointer type to ptrdiff_t, even if it is a reinterpret_cast. Cheers, Brian _______________________________________________ dev-platform mailing list [email protected] https://lists.mozilla.org/listinfo/dev-platform
