On Fri, Mar 29, 2019 at 05:04:18PM -0700, Linus Torvalds wrote: > On Fri, Mar 29, 2019 at 4:25 PM Luc Van Oostenryck > <luc.vanoostenryck@xxxxxxxxx> wrote: > > > > Sparse will warn on casts removing the address space of a pointer > > if the destination type is not unsigned long. But the special role > > of 'unsigned long' is not explained in the man page. > > You might as well call it "uintptr_t". > > And yes, that is the same as "unsigned long" in all sane > implementations, of course, and yes, the Linux kernel tends to prefer > just using "unsigned long" rather than made-up abstract types > internally. Well, outside the kernel, uintptr_t is 'unsigned int' on ARM & i386 (and also riscv-32, MIPS with a 32-bit ABI, ppc32 & sparc32). > But when explaining it outside of the Linux kernel context, I think > "uintptr_t" might be a bit more appropriate (perhaps with an "aka > unsigned long", since sparse itself internally also has the same "long > and pointer are basically the same - it's the register size" model). > > The logic is that an "uintptr_t" cast isn't really a pointer cast, > it's a "get pointer value representation in an integer type" cast. > Which is address-space independent, which can be used as an > explanation of the logic. Yes, uintptr_t should be ideal for this kind of casts, very clear about the intention. It should be easy enough to accept both 'uintptr_t' & 'unsigned long'. -- Luc Van Oostenryck
