[MPlayer-dev-eng] [PATCH 1/11] MPlayer <-> GeeXboX patches port :

[Aurelien Jacobs]

On Sat, 25 Feb 2006 09:53:13 +0100
Diego Biurrun <diego at biurrun.de> wrote:

> On Sat, Feb 25, 2006 at 02:39:38AM -0500, Alan Curry wrote:
> > Diego Biurrun writes the following:
> > >
> > >I don't see how #defining LINUX_MOD_DEVICETABLE_H prevents the inclusion
> > >of mod_devicetable.h if it only gets #included if __KERNEL__ is
> > >#defined...
> > 
> > It wouldn't, but it would make mod_devicetable.h behave as if it had already
> > been included - the second time, thewhole thing is #if'd out - so it'd be
> > just as good as not including it at all.
> > 
> > The big question is: who still has real kernel headers in /usr/include/linux?
> > That's been Wrong since what, Linux 2.2? 2.0?
> 
> Which makes me think that this whole workaround is just not worth it.
> We still haven't seen a failure scenario...

Ok, the senario which trigger this failure:
Cross-compiling a 2.6.15 kernel for ppc, and a uClibc for ppc using this
kernel header, and then cross-compiling mplayer for ppc using the
previously generated environement.
You can test this yourself very easily by grabbing geexbox sources
( http://www.geexbox.org/releases/0.98.7/geexbox-0.98.7-lite.tar.bz2 ),
setting TARGET_ARCH=ppc in config/options, and calling make.

Aurel