Public SourceMage Discussion List

[Arwed von Merkatz <v.merkatz AT gmx.net>]

On Thu, Feb 12, 2004 at 04:04:57PM +1100, Hamish Greig wrote:
> On Thu, 12 Feb 2004 09:25, Arwed von Merkatz wrote:
> > Ok, after having tested most of those spells with 2.4 kernel headers by
> > accident, i now ran all tests again and fixed some compile problems.
> > parted still doesn't compile, no idea why, the rest works.
> > lilo has one caveat: it won't be able to boot from lvm devices if
> > compiled with 2.6 kernel headers, as the 2.6 kernel doesn't have a lvm.h
> > header.
> > All of this is in test and devel and should be safe to use (this time
> > really :)
> 
> xfree86-devel fails due to use of linux/config.h during compile. I removed 
> (locally) it's (config.h's) include of the linux/err_kernel_only.h file so 
> it 
> didn't choke. It has continued passed the previous failure point (still 
> casting ), but I am not sure if this is a valid fix for this problem or not.
> Hamish

I enountered the same problem with xine-lib and just fixed it there, and
i think it's useful to have an indication when some program includes
config.h, as that's definitely not the intended use of it, and may point
to some wrong assumption in the code. xine-lib uses config.h to
determine if devfs is compiled into the kernel and changes the default
audio device to /dev/sound/dsp if it is. Since it only does it for the
oss audio device and not for any other devices it uses it's basically
useless so i removed it (especially because devfs in the kernel doesn't
mean it's actually used on /dev).
For the sake of not breaking any more spells with this, we could replace
that error include in config.h by a #warning. That way there would still
be some feedback when a program includes it, but it won't break the
whole compile.

-- 
Arwed v. Merkatz
Grimoire Guru for video
Grimoire Guru for xfce
Sourcemage GNU/Linux
http://www.sourcemage.org