"m-a" vs "make-kpkg"
I'm interested in finding out why my experience of "m-a" doesn't seem
to agree with yours.
> > I've just tried that; it's a bit less comfortable for me because
> > (as I indicated earlier) module-assistant doesn't seem to care
> > about EXTRAVERSION and APPEND_TO_VERSION, so that it auto-installs
> > the module in the *wrong* directory. So that it must be moved to
> > the correct place afterwards.
> m-a has always made the kernel match the sources I point it at excactly.
> It certainly seems to get the append_to_version and KD_REV correctly
> when I have used it. I have used it both with custom built kernels
> (made with make-kpkg) and debian built kernels (made with who knows
Maybe I didn't read the man page carefully enough; this is the invocation
m-a build -t -u $MA_DIR -k $KERNEL_DIR $MODULE
where $KERNEL_DIR is the kernel source directory used to build the
kernel image and "in-kernel" modules.
This directory contains a "conf.vars" file that contains:
VERSION = 2
PATCHLEVEL = 6
SUBLEVEL = 14
EXTRAVERSION = -vs2.1.0-rc5
APPEND_TO_VERSION = +g2
Debian Revision = 3:custom.1.0.0
do_parallel = -j3
fast_dep = -j3
And the debian package built by "m-a" did not include the "+g2".
> > I've been led to believe that "make-kpkg" was the debian way.
> > It has certainly been around longer than "m-a". There must be more
> > than one way to do it ;-)
> make-kpkg builds the kernel (and optionally any modules you ask it to
> build at the same time). For building modules later or using only the
> kernel-headers from a debian kernel, m-a is much nicer. Not that doing
> KD_REV=custom.1.0 KVERS=2.6.14-my-extra-stuff debian/rules binary-modules
> in the module source dir won't accomplish the same thing in most cases.
> m-a does take care of the KSRC, KVERS, KD_REV, CC, etc, which is rather
> handy, and it is much shorter to type. m-a a-i -t nvidia.
I certainly agree that if one does not have the actual kernel source
tree available, it is handier to use "m-a". Otherwise, "make-kpkg"
is equally fine I guess.
The worry is that the *result* is different, for me at least.