On Mon, Aug 30, 2021 at 10:15 PM Peter Maydell <peter.mayd...@linaro.org> wrote: > I think I would vote for following the type used by the ioctl() > function as declared in the headers for both Linux and the BSDs, > and using 'unsigned long'. > (We should change KVMState::irq_set_ioctl too, to match.)
I would agree to 'unsigned long', in that case I can generate a new patch. And in theory if all libc implementations specify the ioctl request as 'int' we could go back to 'int'. Johannes