Re: Missing shared library for C++ bindings

[Thomas Dickey]

On Sun, Oct 12, 2014 at 04:32:08PM +0000, mestag_a wrote:
> Hello,
> 
> 
> I wanted to write an application using the C++ bindings of the ncurses 
> library.
> 
> I am using Archlinux and ncurses 5.9 which I got with the package manager.
> 
> 
> Unfortunately, I couldn't find any shared library with the C++ bindings (i.e 
> no libncurses++.so).
> 
> 
> I downloaded the sources and tried to compile it myself, I used the
> with-shared option, but no shared library is compiled for the C++ bindings.

The current (patched) source is here (also the ftp link noted in the report):

        http://invisible-island.net/datafiles/current/ncurses.tar.gz

It gets updated each time I make an update.
 
> Is it a "bug" or a "desired issue" ? Or did I miss something ?

You also need the
        --with-cxx-shared

option (as noted in the report).  I added that in

20130309
        + add configure option --with-cxx-shared to permit building
          libncurses++ as a shared library when using g++, e.g., the same
          limitations as libtool but better integrated with the usual build
          configuration (Redhat #911540).

The reason why it was not done originally (in the 1990s) was due to concerns
that building a C++ shared library would require different/special options.  It
took a while for someone to complain about this...

In a quick check, for instance, I don't see any reports in ncurses' NEWS
file which connect to an Arch bug-report.  (Perhaps someone can remind me
if there's an Arch developer listed in any of the reported-by or even
patch-by notes).

> I asked about this issue here
> <https://bbs.archlinux.org/viewtopic.php?id=188217>, and that's where someone
> pointed out the "no generation of libncurses++" issue.

well... I update the patch-date each time, and that's part of the version
number.
        tic -V
shows me
        ncurses 5.9.20141011

Barring regressions (which usually are reported/fixed within a week), the
whole sequence is supposed to be continually upward-compatible.

A comment on the bug-report indicates that someone would have preferred that
I had updated the minor number for what was an addition to a private function.
However, the ABI which Arch would have used is the same.  So... they can use
the patchdate if it's important.

Bumping 5 and/or 9 to something else indicates a major release, and that's when
I get to make incompatible changes :-)

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://invisible-island.net

signature.asc
Description: Digital signature