Re: [Freeipmi-users] fish crashes with "Segmentation fault" at startup

[robm_campbell]

Hi, Albert.

Thanks for the quick response ... see in-line replies below

>Just to make sure, this is the Freeipmi 0.1.3 version?

Yes, I just downloaded version 0.1.3 from the freeipmi.gnu.org site the other
day.

>1) Since this is a chrooted environment, is it possible you don't have
>all the same file system mounts in your chrooted environment as you do
>in your normal environment?  Off the top of my head, the libfreeipmi
>library stores/uses a few things in /var/lib.  So if there isn't a /var,
>some things may get messed up.   Fish also reads from some files in a
>few other directories.  This is just a guess of course.

Yes, I spotted that. I did a 'make -n install > ~/install.log' and then grep'ed
for 'mkinstalldirs' to see where things were being stored and then manually
created the bits that were missed when I copied the contents of '/usr/local'
over to the image file. I'll double-check that I've not missed something
here. I'm assuming from the Troubleshooting section of the documentation
that 'ipckey' doesn't contain any data, it's just the inode number that gets
used.

Just to be sure, when I built I used GUILE=/usr/local/bin/guile 
GUILE_CONFIG=/usr/local/bin/guile-config
GUILE_TOOLS=/usr/local/bin/guile-tools in the 'configure' command-line. This
should mean that the build uses only what I built and installed under 
/usr/local,
right?
Just thinking as I type, fish should use the version under what's specified
by --prefix & --exec-prefix, /usr/local by default, and not what's in $PATH.
The standard RedHat 9-1 distro. of guile/readline and what I built are slightly
different versions.

>2) Is the 2.4.31 IPMI kernel driver turned on?  FreeIPMI doesn't require
>it, and wouldn't think that the kernel driver would interfere with
>libfreeipmi, but who knows.

Hmmm, I can't remember if the kernel driver is enabled. Again, I'll double-check
the .config file. However, since I don't get the poblem with the build-host
unless I 'chroot' to the mounted image file (i.e. the RedHat 2.4.20-8 kernel
without IPMI is still running, it's just the file-system that's changed)
I don't think that's the cause either.

There isn't a connection with the shell used is there? Both the target and
the chroot'ed build-host run ash whereas the normal build-host environment
is bash.

I forgot to mention, the 2.4.31 kernel was built with patches from i2c-2.9.2
and lm_sensors-2.9.2 though the kernel modules aren't (yet) loaded. Just
a thought but since I get the same (apparent) problem with two different
kernels it's maybe not relevant.

Looks like I need to rebuild with debug on and try to work with the core-dump
in gdb.

Thanks,
Rob



___________________________________________________________

Unlimited Tiscali Broadband from 14.99!
http://www.tiscali.co.uk/products/broadband/