bug#81413: [PATCH] install: skip chown/chmod when -C finds nothing to change

Thanks a lot for merging!
Mateusz

On Thu, Jul 16, 2026 at 1:31=E2=80=AFPM P=C3=A1draig Brady <[email protected]=
m> wrote:
>
> On 15/07/2026 22:56, Mateusz Nosek wrote:
> > Hi P=C3=A1draig,
> >
> > Good point, thanks. v2 leaves setdefaultfilecon() inside
> > change_attributes(). Updated patch attached.
> >
> > I began to wonder if need_copy() should also do a more rigorous check
> > for the "selinux_enabled && use_default_selinux_context" case, to see
> > if the destination already has the default context. But that's
> > slightly out of scope of this patch, so just FYI.
> >
> > Thanks,
> > Mateusz
> >
> > On Wed, Jul 15, 2026 at 1:18=E2=80=AFPM P=C3=A1draig Brady <P@draigbrad=
y.com> wrote:
> >>
> >> On 14/07/2026 23:30, Mateusz Nosek wrote:
> >>> Hi,
> >>>
> >>> I attach a patch for the "install" utility: "install -C" was still
> >>> calling chown/chmod on the destination even when need_copy() had
> >>> already determined it matches (mode, owner, group, content), which
> >>> could needlessly fail, e.g. on an immutable file.
> >>>
> >>> This is my first patch to coreutils, so apologies for any mistakes in
> >>> the submission process -- happy to fix anything you point out.
> >> This makes sense I think.
> >>
> >> You moved setdefaultfilecon() out of this consideration.
> >> I'd be inclined to leave it where it was and treat the same way,
> >> because the comparison also compares selinux context,
> >> and also if the file is already present it probably has an appropriate=
 context anyway.
> I've pushed that with a NEWS entry.
> I also pushed the attached as a follow up
> to make cleanup of chattr +i files in tests more robust.
>
> Marking this as done.
>
> thanks!
> Padraig