On Fri, Aug 16, 2024 at 10:55:30AM -0700, Linus Torvalds wrote: > On Fri, 16 Aug 2024 at 10:19, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > > > > All it takes, and IMO it's simpler that way. > > Hey, if it' simpler and gives more natural semantics, I obviously > won't argue against it. > > That said, I do hate your "punch_hole" argument. At least make it a > 'struct' with start/end, not a random int pointer, ok? > > Oh, and can we please make 'dup_fd()' return an error pointer instead > of having that other int pointer argument for the error code? As in https://lore.kernel.org/all/20240812064427.240190-11-viro@xxxxxxxxxxxxxxxxxx/? Sure, I can separate it from dependency on alloc_fd() calling conventions (previous patch in that series) and fold it into this one...
