[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: debian2.2r2 on powerbook G4

Peter Cordes wrote:

>  That's totally bogus.  Nothing depends on the hardware of the computer the
> kernel is compiled on.  With a cross-compiler, you should be able to use a
> fast PPC to compile a kernel on an old x86 laptop.

This was not meant to imply that you couldn't compile a kernel on the Pismo
for the Titanium, however, I haven't done any or seen any testing.  I've seen
mixed results of ppc linux running on the G4. So, I certainly wouldn't commit
to code being able to run on the Titanium, since it has a different chip.
Latest report is that it runs fine, but I don't know if that is for the 400mhz
or 500mhz or just what. They should be the same, but there could be some
timing issues arise, I don't know.

>  The .config files are the only confusing part.  I usually copy .config to
> /usr/src/machinename.conf when I'm done.

Yes, and the main reason I don't like to do that...because if you forget to
copy the .config and build a bad kernel for the wrong machine, it could cause
problems as the two PowerBooks will probably not be configured the exact same.
In this case they will be slightly different, but similar. Call this part of
me, paranoia. Providing you always remember the host is compiling for two
targets, it shouldn't be a problem.

> I use my P200MMX desktop machine
> to compile kernels for my slower x86 machines, a P75 and a 486 laptop.  The
> laptop didn't (until a recent factor-of-20 upgrade to disk space :) have
> enough room for the kernel source tree, and it would have taken all day to
> compile 2.4.2 on it with 20MB of RAM.  My PPC machine compiles its own
> kernel, because it's easier to do that than install a cross compiler to
> compile my x86 kernels on it.

You can do it, no question. But for myself, I would rather compile my kernel
on the native machine it is running on. I don't compile kernels all that often
on my development machines.

If there wasn't any problem moving kernels to other chipsets, we wouldn't need
the cross-compiler to begin with. I do use the cross-compiler on x86
architecture to produce a ppc 823 target, and compress images that can be
flashed over the net, but this will only work with the cross-compiler unless
you were using a 823 host (now that would be painful to build kernels). But I
*believe* compiling for an 823 target on a 750 host (Pismo) requires that you
use a cross compiler, that's how I do it.

As a side, do you know that a PPC host running the cross-compiler produces
different size binaries than an x86 host running the cross-compiler will
produce with the exact same code and compiler revs? Both seem to run fine on
the target, but size is different.

Due to my ignorance in differences between a G3 and G4, I would certainly opt
to compile the G4 on the native processor given the problems I've seen on
getting ppc linux to run on the G4 Titanium anyway, call me old fashioned...


Alan DuBoff
Software Orchestration, Inc.

Reply to: