[oi-dev] [review] 415 libreadline.so.5: symbol tgetent ... not found

[Gordon Ross]

On Wed, Jun 29, 2011 at 7:50 AM, Julian Wiesener <jw at vtoc.de> wrote:
> Hi,
>
> you implemented the wrong fix. That readline is not linked against
> curses, is a feature that we want to keep. Thus we can use readline with
> xcurses, ncurses and curses. What needs to be done is an -lcurses for
> smbclient.

Technically,  it links against -ltermcap, but on our platform that's
just a symlink to -lcurses.

Our linker (these days) does "direct binding" so in theory, this
library (libreadline) can link with -lcurses and applications can
link with ncurses or whatever and all should be well.

If you really want the consuming application to be able to
replace which termcap functions readline links in, that
requires some more work, i.e. NODIRECT mapfile entries
and/or dlopen+dlsym to find those functions at run time...

What makes you so sure it's incorrect to link readline
with the termcap/curses library?  Other platforms do.

Have you tried linking this readline with an application
that uses ncurses or xcurses?  Did it not work?

> Also i don't consider it as showstopper for 151 (dev) but we really
> should fix it for /stable. The reason for non-showstopping is, that the
> Bug is reportet against 147 (buggy sofware is already out) and not that
> critical (sure, sambausers need a workarround, but that few users need
> it for a long time now).

More likely, nobody noticed until now.
At present, you can not run smbclient.
Still seems a test stopper to me...
but that's not my call.

Gordon