Re: Incorrect SELinux information in sed(1) man page

[Kian Kasad]

On 21/01/01 10:00AM, Jim Meyering wrote:
> On Wed, Dec 30, 2020 at 3:00 AM Kian Kasad <kian@kasad.com> wrote:
> > I noticed a slight quirk when building sed:
> > In the source tarball for sed, the doc/sed.1 file says that the program
> > is built with SELinux support. However, this isn't necessarily true. If
> > the program is build without SELinux support, the man page still says
> > that SELinux is enabled. Grepping in the resulting sed binary confirms
> > that the program was in fact built without SELinux support, despite what
> > the man page says.
> >
> > Is this a bug?
> 
> No bug here.
> 
> sed.1 is a generated file. It is generated from sed's own --help
> output, and that --help output tells whether the binary was built with
> SELinux.
> What matters is that the installed sed.1 file is consistent with the
> installed sed binary.

The installed sed.1 file is not consistent with the binary though.
I built and installed sed-4.8 and ran the following two commands:

    $ man ../sed-install/usr/local/share/man/man1/sed.1 | grep SELinux
               This sed program was built with SELinux support.  SELinux is 
enabled on this system.
        
        $ strings ../sed-install/usr/local/bin/sed | grep SELinux
        This sed program was built without SELinux support.

As you can see, the information conflicts. Just for clarity, here's what
I did to build sed:

    $ tar -xf ../sed-4.8.tar.xz
        $ ./configure
        $ make -j8
        $ make DESTDIR=../sed-install install

I'm using Arch Linux, and the official sed package is built directly
from upstream source with no patches. The information about SELinux
conflicts between the man page and the binary within the package, so I
know it's not just a quirk of the way I built it.

--
Kian Kasad