[Buildroot] [PATCH 1/1] elfutils: Add a --{enable, disable}-backends option

[Thomas Petazzoni <thomas.petazzoni@free-electrons.com>]

Dear Jan Heylen,

On Thu, 2 Apr 2015 09:12:53 +0200, Jan Heylen wrote:

> > Patches should have a description and a Signed-off-by line.
> Ah, I didn`t realise that on that level it was required, I'll fix it

Yes, we also require a description + SoB *inside* patches for packages.
See
http://buildroot.org/downloads/manual/manual.html#_format_and_licensing_of_the_package_patches.

> > Also, could you submit this patch upstream? I'd like to see if upstream
> > is willing to accept such a patch, since it is more or less a "feature"
> > patch, and we prefer to not have to carry such patches forever.
> This is based on 0002-disable-progs.patch, so it belongs in buildroot?
> I'll try to upstream it to the elfutils package.

Agreed that 0002-disable-progs.patch is basically the same sauce. It
would be great to upstream these patches.

> > This explanation seems weird: why would you link against libelf and not
> > use it?
> 
> The backends are plaform specific implementations, for every supported
> platform. And, e.g. perf package which sources come with the Linux
> kernel has a compile time dependency on it.On older kernels (<= 3,7),
> the perf package requires libelf, from the eflutils package. But for
> plenty of the features of perf, elfutils backends (libebl) are not
> required at runtime. As backporting patches for not depending on
> elfutils for all former kernel version is obviously not the best
> method, I believe making it possible to not compile/install the
> backends of libebl in your rootfs is a legitimate option (saves >300K
> (stripped)).

Ok, understood.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com