I have a system built with 2004.2 on a P4. When I boot it with "gentoo-sources-2.4.26-r6", CMU CL (both the Gentoo ebuild version 18e and the latest release 19a) segfault. However, when I boot with a kernel built from "gentoo-dev-sources 2.6.7", both versions of CMU CL work just fine. I have another machine that's still running 2004.1; on this machine, both versions of CMU CL work with "gentoo-sources-2.4.25-r1" but segfault with 2.4.26-r6.
This apparently got shuffled off to the "genkernel" team, and now it looks like it got sent to "ML". I don't see how this can be a "genkernel" or an "ML" problem. As far as I can tell, it's limited in its impact to CMU CL and is only happening on the 2.4.26-r6 kernel.
My best guess is that it has something to do with a kernel configuration flag that first appeared (in Gentoo on my machines) in 2.4.26-r6. If you look in the "Processor Type and Features" section of the ".config" file when you build a kernel, you will see:
# CONFIG_1GB is not set
# CONFIG_3GB is not set
# CONFIG_05GB is not set
I believe these are address space limits on a process memory map, and CMU CL is probably accessing a value beyond that, which is why it is segfaulting. I did try a kernel build with the 3 GB option and it didn't help. I don't know what happens if *none* of them are set; when I do the kernel build I use "menuconfig" and it doesn't appear to allow that.
I will try to reproduce this problem on gentoo-sources-2.4.26-r6 soon.
Are you using any GRSEC kernel options or "hardened" GCC options?
Nope ... straight "genkernel --menuconfig all" with only the processor type (Athlon XP, gcc>3.1) and a few other processor-specific options set. If you want, I can post the "config" file.
By the way, I'm going to check this out on 2.4.26-r7 in the next day or so. It seems to take a day for things to propagate to the mirrors. I'm now running 2.6.7 on all my machines except the Athlon XP laptop, which has a Synaptics touchpad that 2.6 isn't quite tracking just yet.
Closing this bug. Genkernel-3.2.2 is out and we are not really supporting 2.4
kernels anymore. Also there have been kernel bug fixes and enhancements since
this was initially posted. Also it may be a bug with your software as no one
else has reported this issue.