[Cscope-devel] [ cscope-Bugs-751104 ] curses search fails on Tru64

[SourceForge.net <no...@so...>]

Bugs item #751104, was opened at 2003-06-08 17:35
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=104664&aid=751104&group_id=4664

Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: Jason Duell (jcduell)
Assigned to: Nobody/Anonymous (nobody)
Summary: curses search fails on Tru64

Initial Comment:
Hans:

I just tried building cscope on an HP Alphaserver
(Compaq Tru64 UNIX V5.1A (Rev. 1885); Compaq
AlphaServer SC TS2.5).

It doesn't find any curses library, and so it winds up
failing at link time.  But if I just add '-lcurses' to
the link command, it seems fine.

I waded into our autoconf to see what's going on.  The
system doesn't seem to have any version of ncurses.h on
it, but it has /usr/include/curses.h and
/usr/lib/libcurses.a.

It looks like in the AC_EGREP_HEADER command we search
for 'init_color', but we don't find it, because it only
gets declared if _XOPEN_SOURCE_EXTENDED is defined, and
it isn't.  So no curses library is found, and
CURSES_LIBS doesn't get set to anything.

I'm attaching a patch that fixes this, by using a
manual AC_EGREP_CPP check that #defines
_XOPEN_SOURCE_EXTENDED
before it includes curses.h.

A note about the patch:  I was confused by the way we
run a AC_EGREP_CPP(USE_NCURSES) check after the
AC_EGREP_HEADER check, regardless of whether the
EGREP_HEADER check succeeded or not.  I stuck it into
it's own check on $search_ncurses, but perhaps that's
wrong thing to do.
I also didn't take any action to pass
-D_XOPEN_SOURCE_EXTENDED as a compiler flag, which
would probably be a good idea if we need it (it works
without it on Tru64, but it's probably getting defined
by some other header file)

I have a couple of other questions about our autoconf
setup:

1) We should probably fail if we can't find a usable
curses library--a helpful error message is better than
a build that fails with missing symbols.

2) how are we propagating the CURSES_LIBS to the
generated Makefile?  Makefile.am refers to
'$(CURSES_LIBS)' instead of @CURSES_LIBS@.  The latter
way is the standard way to grab the  values of
variables propagated with AC_SUBST.  I'm surprised that
the way we refer to them works at all.

3) Our curses checks are a mess.  Have we ever looked
at how vim (or some other standard curses program) does
this?  I'm not sure if our midnight commander inspired
approach is the best way to go.

Cheers,

Jason

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=104664&aid=751104&group_id=4664