[PlanetCCRMA] no network in new kernel

Fernando Pablo Lopez-Lezcano nando@ccrma.Stanford.EDU
Wed Jan 29 11:47:01 2003

>     FPL> It could be a compiler compatibility problem as the
>     FPL> kernels I'm using are compiled with gcc2.96 (under 7.2)
>     FPL> and modules should be compiled using the same
>     FPL> compiler...
> I suspect ive overlooked or misunderstood some very basic concept
> here.  How would you proceed, step by step, to get a compatible
> driver to the new kernel?

The current problem is that Planet CCRMA uses one binary kernel for all
three distribution versions. That was perfectly fine for 7.2/7.3 as they
share the same compiler (minus bug fixes in 7.3). It made my life
simpler because I had to build the kernel only once :-) But 8.0 uses
gcc3.2 as its compiler. And kernel modules compiled with gcc3.2 will
most likely not run in a kernel compiled under 2.96. 

What I have to do is (regretfully) is to split the kernel between
distributions (at least between the 7.x and 8.x worlds - and 8.1 will
bring its own set of problems because they now use a newer version of
the pthread library that - I believe - needs a kernel patch). The split
will happen with the next Planet CCRMA kernel release. There are two
options at this point (or maybe three). One is to try to build the
kernel module with the old compiler (RedHat 7.2/7.3 uses gcc2.96, 8.0
uses gcc 3.2). Redhat 8 includes a backwards compatible 2.96 compiler.
You have to tell the build process of the driver to use (somehow) the
2.96 compiler (/usr/bin/gcc296) instead of just "gcc". Another option
would be for you to rebuild the planet ccrma kernel with the 3.2
compiler - not a lot of work but sort of wasted time. Yet another would
be for me to build the kernel module here with the proper compiler and
add it to the repository (I did that with the Nvidia proprietary

I definitely have to get the new kernel going and split the

-- Fernando