On Fri, 2006-07-14 at 11:59 +0200, Jan Janssens wrote:
> I want my kernel modules to be stripped (using -g obviously) and they
> are. But the stripping process takes place after they are packaged in
> my deploy/ipk directory. So on my root_fs, the non-stripped modules
> appear.
>
> Does anybody recognise this problem? I've been tracing what happens
> and checking relevant bbclasses for modifications (I don't use a
> totally up-to-date checkout), but I can't seem to find where this goes
> wrong.
I think this is an unfortunate case of bit-rot. I guess it's been
broken for ages, actually: it's a bit odd that nobody has noticed
before. It looks like this is probably broken in the Familiar 0.8.4
tree as well.
Module stripping is handled by module_strip.bbclass, which adds itself
to the end of do_package(). When this was originally written, it meant
that the stripping was interposed between do_package and do_package_ipk,
so the modules would get stripped after having been installed into ${D}
but prior to being packed up into the .ipk files. Sadly, nowadays,
do_package_ipk() is called directly out of do_package(), so this doesn't
work anymore: do_module_strip actually runs after all packaging has been
completed.
The easiest fix would be to make a new variable PACKAGESTRIPFUNCS (cf
PACKAGEFUNCS) and arrange for populate_packages() to invoke all the
functions named therein at an appropriate point: that'd probably be just
before the "del localdata" at about line 230 of package.bbclass.
p.
_______________________________________________
Oe mailing list
[email protected]
https://www.handhelds.org/mailman/listinfo/oe
