> Yes, having a high-level version of copy_file_range() that falls back to copyfileobj() should be okay.

I'm not sure about this. For the moment c_f_o() is available only if the syscall is there.


> I am wondering if it would be nice to rearrange the os.copy_file_range() signature and make more parameters optional, [...]
> 
> copy_file_range(in, out, count, offset_in=None, offset_out=None, flags=0)

I agree with this, most of the time you will want to just advance both offsets, and providing None all the time can be tiring.

I fixed this, modified a little the doc, but now I'll read about integer types and sizes.