[Python-Dev] readline not getting built when .so symlink doesn't exist

Michael Hudson mwh@python.net
04 Jun 2001 22:55:34 +0100


Skip Montanaro <skip@pobox.com> writes:

>     [my readline woes snipped]
> 
>     Michael> Hmm.  Does compiling a proggie
> 
>     Michael> $ gcc foo.c -lreadline 
> 
>     Michael> work?  It doesn't here if I move libreadline.so & libreadline.a
>     Michael> out of the way.
> 
> Yup, it does:
> 
>     beluga:tmp% cc -o foo foo.c -lreadline -ltermcap
>     beluga:tmp% ./foo
>     >>sdfsdfsdf
>     sdfsdfsdf
> 
> (This after deleting both /lib/libreadline.so and /lib/libhistory.so.)

Odd.  What does the output of 

$ gcc -o foo foo.c -lreadline -ltermcap -Wl,--verbose

look like?  In particular the bit at the end where you get things
like:

attempt to open /usr/lib/gcc-lib/i386-redhat-linux/2.95.1/libreadline.so failed
attempt to open /usr/lib/gcc-lib/i386-redhat-linux/2.95.1/libreadline.a failed
attempt to open /usr/i386-redhat-linux/lib/libreadline.so failed
attempt to open /usr/i386-redhat-linux/lib/libreadline.a failed
attempt to open /usr/bin/../lib/libreadline.so succeeded
-lreadline (/usr/bin/../lib/libreadline.so)

(this is more for my personal curiosity than any important reason).

> Got that.  I just noticed that "rpm -q --whatprovides /lib/libreadline.so"
> does list readline-devel as the provider.  I just reinstalled it using
> --force.  Now the .so symlinks are there.  Go figure...

No :-)

> Oh well, probably ought to drop it unless another Mandrake user complains.

Sounds reasonable.

Cheers,
M.

-- 
  After a heavy night I travelled on, my face toward home - the comma
  being by no means guaranteed.           -- paraphrased from cam.misc