Python 2.4.4 and devel/ncurses

[Kirill Ponomarew]

On Tue, Feb 27, 2007 at 03:58:57PM +0800, Rong-En Fan wrote:
> On Mon, Feb 26, 2007 at 04:19:04PM -0500, Thomas Dickey wrote:
> > On Mon, Feb 26, 2007 at 09:44:11PM +0100, Kirill Ponomarew wrote:
> > > On Mon, Feb 26, 2007 at 01:07:48PM -0700, Joshua Tinnin wrote:
> > > > > > It seems to me that rpath works as expected but /lib/libncurses.so.6 is
> > > > > > loaded first...
> > > > > 
> > > > > There was some discussion about a year ago, which pointed out a problem.
> > > > > The readline package loads termcap, which could be provided by ncurses.
> > > > > You might be hitting that problem (readline, or another package).
> > > > > The fix would be to ensure that "termcap" loads the same library as
> > > > > "curses".
> > > > 
> > > > Is there anything that can be done about this in the meantime? I have
> > > > two 6.2-R i386 machines which are encountering this problem.
> > > 
> > > BTW, it fails also on AMD64 CURRENT from yesterday with:
> > 
> > yes - someone has to provide a patch to make the python and ncurses
> > ports either work together, or ignore each other.  The top-level
> > setup.py appears to have all (or most) of the related information:  it
> > sets up a search list of library directories, walks through them to find
> > the given libraries.
> > 
> > I'm not sure how (reading the earlier email) python decides to get
> > libtinfo, though, since that string doesn't appear in the original
> > python sources.
> 
> Hmm... there must be something wrong  in devel/ncurses
> 
> /usr/local/lib/libncurses.so:
>         libtinfo.so.5.6 => /usr/local/lib/libtinfo.so.5.6 (0x800958000)
> 
> But on a current box which has ncurses 5.6 in base. libncurses.so
> does not depend on tinfo.
> 
> I will look into it.

It fails also with base ncurses on AMD64 current.

-Kirill