Ted, it appears this is some kind of bug with how the binaries were
built on the testing snapshot images. I had to reinstall the system
for other reasons and the
http://cdimage.debian.org/debian-cd/4.0_r0/i386/iso-cd/debian-40r0-i386-netinst.iso
image works perfectly on the Geode hardware with no floating point
exceptions. I'm not sure who would need to look into this. Do you know
who builds the binaries on the snapshot images? (or the configuration
the build uses) -Andrew
On 5/28/07, Theodore Tso <[EMAIL PROTECTED]> wrote:
On Mon, May 28, 2007 at 05:28:22PM -0400, Andrew Paprocki wrote:
> I was referring to the udeb package which is on the install DVD. I
> checked and these binaries are coming from:
> /cdrom/pool/main/e/e2fsprogs-udeb.1.39+1.40-WIP-2006.11.14+dfsg.2_i386.udeb
I just tried running those binaries on my system, and they work just
fine.
> It looks like there is some kind of problem with the way these
> binaries were built inside the .udeb package versus the actual
> installed .deb package? Who
Given that the mke2fs executable extracted from the above udeb works
just fine when I run it on my system So I'm not quite ready to blame
those binaries just yet.
Can you try booting the installation CD, and then ftping over
/sbin/mke2fs plus its necessary libraries:
27% ldd /sbin/mke2fs
linux-gate.so.1 => (0xffffe000)
libext2fs.so.2 => /lib/libext2fs.so.2 (0x4bfce000)
libcom_err.so.2 => /lib/libcom_err.so.2 (0x4d1a8000)
libblkid.so.1 => /lib/libblkid.so.1 (0x4e7c7000)
libuuid.so.1 => /lib/libuuid.so.1 (0x4c0f3000)
libe2p.so.2 => /lib/libe2p.so.2 (0x4bfa1000)
libc.so.6 => /lib/libc.so.6 (0x4be59000)
libdevmapper.so.1.02 => /lib/libdevmapper.so.1.02 (0x4bf8d000)
libselinux.so.1 => /lib/libselinux.so.1 (0x4ca7d000)
libsepol.so.1 => /lib/libsepol.so.1 (0x4c9ab000)
And then try to that version of mke2fs in the installation
environment? It could just as easily be something in the installation
environment. (Relatively few programs use floating point; it could
just as easily be the glibc udeb not being built with floating point
support, or something stupid like that.)
If that fails, then it would conclusively prove that there's some
problem in the installation CD/DVD environment.
This is working for everyone else, though, so it has to be some kind
of interaction between the installer CD/DVD programs (either in
e2fsprogs, or glibc, or other libraries used by mke2fs, or in the
kernel) and something about the Geode environment.
What the heck that might be I have no idea, but that's why I was
trying to investigate the lack of CONFIG_MATH_EMULATOIN theory: it
fits all of the known facs when nothing else I can think of explains
everything that you seem to be seeing.
Regards,
- Ted
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]