[csw-maintainers] [csw-users] python 2.6.1 now in testing

[Sebastian Kayser]

Mike Watters wrote:
> Philip Brown wrote:
>> On Thu, Apr 02, 2009 at 10:02:09AM -0500, Mike Watters wrote:
>>> ARRGGHHH!!
>>>
>>> after a little more searching,
>>> - -- wchgat symbol is in /opt/csw/lib/libncurses.so,
>>> but the new w32addch is NOT.  It lives in /usr/lib/libcurses.so
>> thatas for widecha support. if it isnt there it may be bug in our curses.
> 
>> or, you need to quit mixing curses headers with ncurses libraries.
> 
>> -I/opt/csw/include/ncurses
> 
>> if i remember correctly.
>> _______________________________________________
>> maintainers mailing list
>> maintainers at lists.opencsw.org
>> https://lists.opencsw.org/mailman/listinfo/maintainers
> 
> AH-HA!
> 
> it seems there are 2 sets of ncurses libs
> libncurses and libncursesw (wide support)
> My Headers are picking up the "Wide" support
> but my link is against the normal.
> 
> this is explained in
> http://www.opencsw.org/mantis/view.php?id=3457
> though I am still unclear as to why the headers
> are missing for the "regular" ncurses.
> 
> I will fix and get a new version in testing.

I didn't get my head around why your linkage went like it did quite yet.
Shouldn't it have bailed out with linking errors (for not finding
w32addch) during building already? James?? ;)

Dago recently built our ncurses with wide character support (thanks
btw.), so that we now also have w variants of the ncurses libraries in
addition to the regular ones (not binary compatible but as they have a w
suffix there's no harm done). Both in the same package.

Coming to the include files. According to the ncurses README:

If you configure using the --enable-widec option, a "w" is appended to
the library names (e.g., libncursesw.a), and the resulting libraries
support wide-characters, e.g., via a UTF-8 locale.  The _corresponding
header files are compatible with the non-wide-character configuration_;
wide-character features are provided by ifdef's in the header files.

I don't exactly know how to read the header bit WRT packages building
against these headers. Can non-wide and wide applications both use the
same headers and ./configure sets #defines in case wide-support is
requested by an app (that's how i would read it)?

At least we should have the original include/ncurses directory back in
the package - presumably filled with the header files of the non-wide
build (as a safe option). And then in addition have include/ncursesw
populated with the header files of the wide variant?

Sebastian