Could not compile BASH with --with-curses

[biffcool@…]

On two boxes (Mandrake 8.0 and Slackware 8.0) there was no entry for libcurses.a, a symlink to libncurses.a fixed the problem but this link did not exist by default.

[gerard@…]

Did you try to compile bash-2.05?

Bash-2.05's configure script was fixed: it is supposed to detect either libcurses.a or libncurses.a so you wouldn't have to create the symlink. I have confirmed this myself, but it seems there are the odd few people who still report that problem with the missing symlink.

Could you remove your symlink, remove the bash source tree, unpack it again, and run configure on it like the lfs-book tells you to. Then i'd like to see the output of ./configure so I can have a look at it. I'd also like to see the config.log and config.cache files after ./configure is done.

If you could send those files and configure's output to gerard@… I'd appreciate it.

Thanks,

[gerard@…]

[Kevan Shea (the reporter) sent that output per my request.]

I think I found the problem: ./configure found libtermcap on your system, but it's probably not completely installed (ie: the static library that be missing but it did find the shared library). So if you have termcap installed, make sure you installed the termcap development package as well.

I've double checked the configure script and it checks in this order: libtermcap libcurses libncurses

Configure scripts compile a test program dynamic. So perhaps your system has the dynamic (shared) termcap library but not the static version. So when you actually run make it will fail because it's missing something.

I'm not sure yet why the symlink fixed it for you, it just doesn't compute when I look at the configure script.

I am planning to install some different distributions soon (i'm going to download the latest redhat, slackware, suse, etc, distributions to test them out) and see what's up with them.

I'm going to set this bug to INVALID for now, since I still am convinced this is an isolated incident and not a problem with bash. But I may be proven wrong, so it'll be addressed then anyways.