kernel and gcc issues -- giving it one more shot

Robert P. J. Day rpjday at
Tue Jul 29 18:59:51 UTC 2003

On Tue, 29 Jul 2003, Gerald Henriksen wrote:

> On Tue, 29 Jul 2003 13:12:05 -0400 (EDT), you wrote:
> >  first, there is the occasion where you have to deal with a kernel and
> >compiler mismatch, particularly with nvidia drivers.  if you try to
> >compile the latest nvidia driver normally, you get a "mismatch" error that
> >you can override by defining the "IGNORE_CC_MISMATCH" environment
> >variable when you build.  fair enough, but of course, you never know
> >when this issue will come up again elsewhere.
> Better might be CC=gcc32 so that the kernel and NVIDIA driver both use
> the same compiler.
> >  second issue is whether there is any problem simply recompiling the
> >installed kernel source with gcc (3.3, that is)?  is there anything about
> >that kernel that's necessarily gcc 3.2-related?
> See the following explanation from Alan Cox on why gcc32 is necessary
> for the kernel:

actually, that pretty much confirms what i had suspected -- that gcc 3.3
was simply more rigorous in rejecting "borderline bogus constructs".  so
i'm assuming that, as long as the compile succeeds, i should be all right.

and in fact, the compile just finished (after having deselected a couple
of those bogus modules).  so, lessee what happens ...


More information about the test mailing list