Re: [PATCH] libsepol: use bool instead of int in validation functions
Stephen Smalley <[email protected]>
| Newsgroups | org.kernel.vger.selinux |
|---|---|
| Message-ID | <CAEjxPJ6NGxMic8PVrHZX6OtUMewp+WSfqsc2KDLso5=nKdUHng@mail.gmail.com> |
On Mon, Aug 17, 2026 at 9:35 AM Stephen Smalley <[email protected]> wrote: > > On Mon, Aug 17, 2026 at 8:43 AM Stephen Smalley > <[email protected]> wrote: > > > > On Fri, Aug 14, 2026 at 7:32 PM Kalevi Kolttonen <[email protected]> wrote: > > > > > > This is an effort to use boolean return values instead of > > > int in as many libsepol validation functions as possible. > > > The functions passed to hashtab_map() have been kept as is > > > to avoid excessive reworking of the codebase. > > > > > > The change from int to bool requires adjusting the callers > > > to invert their return value checking logic. The caller > > > conversion has been performed to keep the existing logic > > > intact. > > > > > > Signed-off-by: Kalevi Kolttonen <[email protected]> > > > > I am hesitant to apply this patch for the following reasons: > > 1. It changes the semantics of the return type of a function that is > > exposed to multiple source files, > > 2. It does not rename the functions to make the change in return type > > explicit and ensure all > > callers are updated. > > 3. It doesn't provide any runtime or memory benefit. > > > > Other SELinux userspace maintainers are free to disagree with me and > > merge it if they desire. > > Also, your patch breaks building the fuzzers due to missing #include > <stdbool.h>; > to reproduce, run ./scripts/oss-fuzz.sh to build the fuzzers. One last reason to not apply this patch: it makes the code less consistent, not more, since as you said, you don't convert all the functions to reduce the scope of the change. I'd rather keep it consistent.