Re: 2.6.0-test9: selinux compile error with "make O=..."

[Sam Ravnborg <sam@ravnborg.org>]

On Mon, Oct 27, 2003 at 08:40:42AM -0500, Stephen Smalley wrote:
> On Sun, 2003-10-26 at 04:49, Sam Ravnborg wrote:
> > Hi Adrian.
> > Known problem that has been reported back to the maintainers about
> > one month ago. But they do not seem to care enough to fix it.
> 
> I have no prior email regarding the issue. Who reported it, and to whom?
> Was it cc'd to any mailing list (e.g. lkml, lsm, or selinux)?
You and jmorris@redhat.com got an mail about it the 26th of September.
Probarly lost in usual mail noise.

> > The use of "-include" is a bad way to include files. The reader will
> > not see that global.h is included at all and will wonder how that
> > information get pulled in.
> 
> True, and the original reason for it is no longer valid, so we can
> change this.
Good.

> > Furhtermore the location of the header files under security/include
> > is considered bad practice. All headerfiles used from more than one
> > directory belongs to include/xxx, in this case include/security.
> > Then they can be included using
> > #include <security/secuity.h>
> 
> This was discussed when SELinux was originally submitted for merging,
> but these header files are private to the SELinux kernel module are
> never included into out-of-tree code, so it seemed unjustified to move
> them.  Now, if this breaks the build process, we can move them, but I
> would appreciate clarification as to whether this is truly a limitation
> of the build process for make O=.
The build system will handle the use of -I correct - I was only
referring to what is my understanding of the un-written rules for
location of .h files.
If the usage of -include is fixed then from a kbuild perspective there
is no problems.

	Sam