[PlanetCCRMA] omlinux: line 41; 3980 Segmentation fault...

Fernando Pablo Lopez-Lezcano nando@ccrma.Stanford.EDU
Fri Mar 19 10:54:01 2004

> > Yuck, sorry about that, it works here on my two fc1 test systems. A few
> > questionss. Are you completely up to date in all packages? (apt-get
> > dist-ugprade).
> no, not completely up to date but my cd images are only about 3 weeks
> old and i have upgraded lots of bits and pieces since then -
> an apt-get dist-upgrade for me requires a 306MB download which I can't
> do on a connection which gives me a maximum of less than 28kbps ! 
> broadband hasn't yet reached my part of ireland...

I was wondering about glibc, one of the few global pieces that could (my
guess) affect cmucl. 

> >  Second, does it make a difference if you do a:
> >   LD_ASSUME_KERNEL=2.2.5 /usr/bin/lisp
> [root@localhost robcanning]# LD_ASSUME_KERNEL=2.2.5 /usr/bin/lisp
> /usr/bin/lisp: relocation error: /usr/bin/lisp: symbol dlopen, version
> GLIBC_2.1 not defined in file libdl.so.2 with link time reference

Aha, same message I was getting before I tried the "loader from rh9"
hack. Maybe the hack would work again. But why is this working on some
machines and not on others?

Most probably you still have the cmucl-2003.12-1 package lying around,
right? You could install that one with rpm and the "--oldpackage"
option. That should get lisp back, I guess. Then you could try applying
the same hack I used before to get the newer image running. 

After installing the -1 package save these files somewhere else:
The upgrade to the -2 package. 
Copy the libdl-2.3.2.so to its original place. 
Rename /usr/bin/lisp to /usr/bin/lisp.bin
Install the /usr/bin/lisp that you saved in /usr/bin

Try starting lisp...

> > Third, what kernel are you running?
> Fedora Core (2.4.22-9.23rc3.ll.rhfc1.ccrma)

Hmmm, a bit old, current is 2.4.25 (in planet edge). I was thinking it
might be the Fedora kernel, which has different stuff built into it like
stack checks and the like. 

-- Fernando