Re: Kernel compilation failure
On Tue, Sep 27, 2005 at 06:47:43PM +0200, Gilles wrote:
> > That '2419 Segmentation fault gcc ....' looks bad. gcc only
> > segfaults on compiler bugs or hardware errors in my experience. Of
> > course being gcc 4.0.x compiler bugs are entirely possible.
> >
> > On the other hand the fact it is _not_ happening the same place each
> > time rather makes the hardware error much more likely.
> >
>
> :-(
Used to be you would see sig 11 or some such when gcc hit bad hardware.
It was common enough to see it during kernel compiles that it has an FAQ
entry in the kernel FAQs.
> > If you can do make clean and then build it and have it fail the same
> > place 3 or 4 times in a row, then submit a bugreport on gcc-4.0.
> > If it doesn't fail the same place each of those times, start swapping ram,
> > cpu, etc, until you find the faulty hardware.
> >
>
> The fault occurs at different places...
So if you continue the compile right away it will get further than it
was by continuing the compile?
> I already did a complete pass with "memtest86+"; it didn't report a
> single memory error.
memtest rarely finds flacky memory, only fully broken memory.
> What and how should I test next?
> Do I just unplug one of the processors?
Well if you have multiple, that might be an idea. You could remove half
the ram and half the cpus.
> Is there some other diagnostics tool which I can run?
Well gcc compiling kernels is amazingly good at finding unstable
hardware due to the high cpu/memory/io load it puts on the system.
Doing a make clean and starting a new compile after any change to the
hardware and see if it makes it all the way through. If it does, you
may have removed the problem hardware. Try swapping around hardware
until you find out either which cpu or which stick of memory is causing
problems, and if you are lucky, it isn't the mainboard.
If it was a gcc bug I would have expected more people to complain
already. Maybe they are just being quiet about it. :)
Which kernel are you running right now? I suppose there is also a small
chance that you are running a buggy kernel causing gcc to segfault
during the compile.
Len Sorensen
Reply to: