Re: Bug#134784: xfree86: Source package 4.1.0-14 doesn't build on woody/testing
On Wed, Feb 20, 2002 at 10:37:33AM -0500, Branden Robinson wrote:
> On Wed, Feb 20, 2002 at 10:28:02AM +0800, User wrote:
> > I'm unable to compile 4.1.0-14 on my testing system. The only test I have
> > done is to redirect stderr to a file. 'tail -n 20' of the file gives:
> >
> > > tail -n 20 stderr.txt
> > /usr/include/unistd.h:985: warning: redundant redeclaration of
> > thread_atfork' in same scope
> > /usr/include/pthread.h:673: warning: previous declaration of thread_atfork'
> > sis_alloc.c: In function is_alloc_fb':
> > sis_alloc.c:124: IS_IOCTL_FB_ALLOC' undeclared (first use in this function)
> > sis_alloc.c:124: (Each undeclared identifier is reported only once
> > sis_alloc.c:124: for each function it appears in.)
> > sis_alloc.c: In function is_free_fb':
> > sis_alloc.c:154: IS_IOCTL_FB_FREE' undeclared (first use in this function)
> > sis_alloc.c: In function is_alloc_agp':
> > sis_alloc.c:197: IS_IOCTL_AGP_ALLOC' undeclared (first use in this
> > function)
> > sis_alloc.c: In function is_free_agp':
> > sis_alloc.c:224: IS_IOCTL_AGP_FREE' undeclared (first use in this function)
> > make[7]: *** [sis_alloc.o] Error 1
> > make[6]: *** [all] Error 2
> > make[5]: *** [all] Error 2
> > make[4]: *** [all] Error 2
> > make[3]: *** [all] Error 2
> > make[2]: *** [World] Error 2
> > make[1]: *** [World] Error 2
> > make: *** [debian/stampdir/build] Error 2
> >
> > >From the above, I suspect a problem with the SIS drivers.
>
> You need to build against a glibc-dev which uses kernel headers that
> include sisfb information. Apparently, the only way to ensure this is
> to build glibc-dev using a kernel with sisfb support enabled. Why the
> kernels have different headers depending on whether a particular fbcon
> module is enabled or not is a damn good question, and one that should be
> asked of the kernel maintainers upstream.
>
> I've talked to Ben Collins (glibc maintainer) about this problem, and he
> says it's up to the Debian maintainers who build glibc for the
> respective architectures to take care of this problem. At least until
> the Linux kernel guys can be persuaded to have the ioctl()s defined in
> their headers unconditionally.
Assuming that a particular architecture supports sisfb, they should
configure it. If that particular architecture doesn't support sisfb,
then maybe the XFree86 build should enable that driver (perhaps some
sort of configure test).
It is odd that the sis ioctl's are only defined when the kernel is
configured to support sisfb (if that is the case). I don't think sis
showed up until later 2.4.x kernels though.
--
.----------=======-=-======-=========-----------=====------------=-=-----.
/ Ben Collins -- Debian GNU/Linux -- WatchGuard.com \
` bcollins@debian.org -- Ben.Collins@watchguard.com '
`---=========------=======-------------=-=-----=-===-======-------=--=---'
Reply to: