freebsd-nq/contrib/ncurses/INSTALL

487 lines
21 KiB
Plaintext

-- $Id: INSTALL,v 1.33 1999/09/18 23:04:36 tom Exp $
---------------------------------------------------------------------
How to install Ncurses/Terminfo on your system
---------------------------------------------------------------------
************************************************************
* READ ALL OF THIS FILE BEFORE YOU TRY TO INSTALL NCURSES. *
************************************************************
You should be reading the file INSTALL in a directory called ncurses-d.d, where
d.d is the current version number. There should be several subdirectories,
including `c++', `form', `man', `menu', 'misc', `ncurses', `panel', `progs',
and `test'. See the README file for a roadmap to the package.
If you are a Linux or FreeBSD or NetBSD distribution integrator or packager,
please read and act on the section titled IF YOU ARE A SYSTEM INTEGRATOR
below.
If you are converting from BSD curses and do not have root access, be sure
to read the BSD CONVERSION NOTES section below.
If you are using a version of XFree86 xterm older than 3.1.2F, see the section
on RECENT XTERM VERSIONS below.
If you are trying to build GNU Emacs using ncurses for terminal support,
read the USING NCURSES WITH EMACS section below.
If you are trying to build applications using gpm with ncurses,
read the USING NCURSES WITH GPM section below.
If you are trying to build Elvis using ncurses for terminal support,
read the USING NCURSES WITH ELVIS section below.
If you are running over the Andrew File System see the note below on
USING NCURSES WITH AFS.
If you want to build the Ada95 binding, go to the Ada95 directory and
follow the instructions there. The Ada95 binding is not covered below.
If you are using anything but (a) Linux, or (b) one of the 4.4BSD-based
i386 Unixes, go read the Portability section in the TO-DO file before you
do anything else.
REQUIREMENTS:
You will need the following in order to build and install ncurses under UNIX:
* ANSI C compiler (gcc is recommended)
* sh (bash will do)
* awk (mawk or gawk will do)
* sed
* BSD or System V style install (a script is enclosed)
Ncurses has been also built in the OS/2 EMX environment.
INSTALLATION PROCEDURE:
1. First, decide whether you want ncurses to replace your existing library (in
which case you'll need super-user privileges) or be installed in parallel
with it.
The --prefix option to configure changes the root directory for installing
ncurses. The default is in subdirectories of /usr/local. Use
--prefix=/usr to replace your default curses distribution. This is the
default for Linux and BSD/OS users.
The package gets installed beneath the --prefix directory as follows:
In $(prefix)/bin: tic, infocmp, captoinfo, tset,
reset, clear, tput, toe
In $(prefix)/lib: libncurses*.* libcurses.a
In $(prefix)/share/terminfo: compiled terminal descriptions
In $(prefix)/include: C header files
Under $(prefix)/man: the manual pages
Note however that the configure script attempts to locate previous
installation of ncurses, and will set the default prefix according to where
it finds the ncurses headers.
2. Type `./configure' in the top-level directory of the distribution to
configure ncurses for your operating system and create the Makefiles.
Besides --prefix, various configuration options are available to customize
the installation; use `./configure --help' to list the available options.
If your operating system is not supported, read the PORTABILITY section in
the file ncurses/README for information on how to create a configuration
file for your system.
The `configure' script generates makefile rules for one or more object
models and their associated libraries:
libncurses.a (normal)
libcurses.a (normal, a link to libncurses.a)
This gets left out if you configure with --disable-overwrite.
libncurses.so (shared)
libncurses_g.a (debug)
libncurses_p.a (profile)
If you do not specify any models, the normal and debug libraries will be
configured. Typing `configure' with no arguments is equivalent to:
./configure --with-normal --with-debug --enable-overwrite
Typing
./configure --with-shared
makes the shared libraries the default, resulting in
./configure --with-shared --with-normal --with-debug --enable-overwrite
If you want only shared libraries, type
./configure --with-shared --without-normal --without-debug
Rules for generating shared libraries are highly dependent upon the choice
of host system and compiler. We've been testing shared libraries on Linux
and SunOS with gcc, but more work needs to be done to make shared libraries
work on other systems.
You can make curses and terminfo fall back to an existing file of termcap
definitions by configuring with --enable-termcap. If you do this, the
library will search /etc/termcap before the terminfo database, and will
also interpret the contents of the TERM environment variable. See the
section BSD CONVERSION NOTES below.
3. Type `make'. Ignore any warnings, no error messages should be produced.
This should compile the ncurses library, the terminfo compiler tic(1),
captoinfo(1), infocmp(1), toe(1), clear(1) tset(1), reset(1), and tput(1)
programs (see the man pages for explanation of what they do), some test
programs, and the panels, menus, and forms libraries.
4. Run ncurses and several other test programs in the test directory to
verify that ncurses functions correctly before doing an install that
may overwrite system files. Read the file test/README for details on
the test programs.
NOTE: You must have installed the terminfo database, or set the
environment variable $TERMINFO to point to a SVr4-compatible terminfo
database before running the test programs. Not all vendors' terminfo
databases are SVr4-compatible, but most seem to be. Exceptions include
DEC's Digital Unix (formerly known as OSF/1).
The ncurses program is designed specifically to test the ncurses library.
You can use it to verify that the screen highlights work correctly, that
cursor addressing and window scrolling works OK, etc.
5. Once you've tested, you can type `make install' to install libraries,
the programs, the terminfo database and the man pages. Alternately, you
can type `make install' in each directory you want to install. In the
top-level directory, you can do a partial install using these commands:
'make install.progs' installs tic, infocmp, etc...
'make install.includes' installs the headers.
'make install.libs' installs the libraries (and the headers).
'make install.data' installs the terminfo data. (Note: `tic' must
be installed before the terminfo data can be
compiled).
'make install.man' installs the man pages.
############################################################################
# CAVEAT EMPTOR: `install.data' run as root will NUKE any existing #
# terminfo database. If you have any custom or unusual entries SAVE them #
# before you install ncurses. I have a file called terminfo.custom for #
# this purpose. Don't forget to run tic on the file once you're done. #
############################################################################
The terminfo(5) manual page wants to be preprocessed with tbl(1) before
being formatted by nroff(1). Modern man(1) implementations tend to do
this by default, but you may want to look at your version's man page
to be sure.
If the system already has a curses library that you need to keep using
for some bizarre binary-compatibility reason, you'll need to distinguish
between it and ncurses. If ncurses is installed outside the standard
directories (/usr/include and /usr/lib) then all your users will need
to use the -I option to compile programs and -L to link them.
If you have BSD curses installed in your system and you accidentally
compile using its curses.h you'll end up with a large number of
undefined symbols at link time. _waddbytes is one of them.
IF YOU DO NOT HAVE ROOT: Change directory to the `progs' subdirectory
and run the `capconvert' script. This script will deduce various things
about your environment and use them to build you a private terminfo tree,
so you can use ncurses applications.
If more than one user at your site does this, the space for the duplicate
trees is wasted. Try to get your site administrators to install a system-
wide terminfo tree instead.
See the BSD CONVERSION NOTES section below for a few more details.
6. The c++ directory has C++ classes that are built on top of ncurses and
panels. You must have c++ (and its libraries) installed before you can
compile and run the demo.
Use --without-cxx-binding to tell configure to not build the C++ bindings
and demo.
If you do not have C++, you must use the --without-cxx option to tell
the configure script to not attempt to determine the type of 'bool'
which may be supported by C++. IF YOU USE THIS OPTION, BE ADVISED THAT
YOU MAY NOT BE ABLE TO COMPILE (OR RUN) NCURSES APPLICATIONS WITH C++.
7. If you're running an older Linux, you must either (a) tell Linux that the
console terminal type is `linux' or (b) make a link to or copy of the
linux entry in the appropriate place under your terminfo directory, named
`console'. All 1.3 and many 1.2 distributions (including Yggdrasil and
Red Hat) already have the console type set to `linux'.
The way to change the wired-in console type depends on the configuration
of your system. This may involve editing /etc/inittab, /etc/ttytype,
/etc/profile and other such files.
Warning: this is not for the fainthearted, if you mess up your console
getty entries you can make your system unusable! However, if you are
a distribution maker, this is the right thing to do (see the note for
integrators near the end of this file).
The easier way is to link or copy l/linux to c/console under your terminfo
directory. Note: this will go away next time you do `make install.data'
and you'll have to redo it. There is no need to have entries for all
possible screen sizes, ncurses will figure out the size automatically.
IF YOU ARE A SYSTEM INTEGRATOR:
Beginning with 1.9.9, the ncurses distribution includes both a tset
utility and /usr/share/tabset directory. If you are installing ncurses,
it is no longer either necessary or desirable to install tset-jv.
Configuration and Installation:
Configure with --prefix=/usr to make the install productions put
libraries and headers in the correct locations (overwriting any
previous curses libraries and headers). This will put the terminfo
hierarchy under /usr/share/terminfo; you may want to override this with
--datadir=/usr/share/misc; terminfo and tabset are installed under the
data directory.
Please configure the ncurses library in a pure-terminfo mode; that
is, with the --disable-termcap option. This will make the ncurses
library smaller and faster. The ncurses library includes a termcap
emulation that queries the terminfo database, so even applications
that use raw termcap to query terminal characteristics will win
(providing you recompile and relink them!).
If you must configure with termcap fallback enabled, you may also
wish to use the --enable-getcap option. This option speeds up
termcap-based startups, at the expense of not allowing personal
termcap entries to reference the terminfo tree. See the code in
ncurses/tinfo/read_termcap.c for details.
Note that if you have $TERMCAP set, ncurses will use that value
to locate termcap data. In particular, running from xterm will
set $TERMCAP to the contents of the xterm's termcap entry.
If ncurses sees that, it will not examine /etc/termcap.
Keyboard Mapping:
The terminfo file assumes that Shift-Tab generates \E[Z (the ECMA-48
reverse-tabulation sequence) rather than ^I. Here are the loadkeys -d
mappings that will set this up:
keycode 15 = Tab Tab
alt keycode 15 = Meta_Tab
shift keycode 15 = F26
string F26 ="\033[Z"
Naming the Console Terminal
In various Linuxes (and possibly elsewhere) there has been a practice
of designating the system console driver type as `console'. Please
do not do this any more! It complicates peoples' lives, because it
can mean that several different terminfo entries from different
operating systems all logically want to be called `console'.
Please pick a name unique to your console driver and set that up
in the /etc/inittab table or local equivalent. Send the entry to the
terminfo maintainer (listed in the misc/terminfo file) to be included
in the terminfo file, if it's not already there. See the
term(7) manual page included with this distribution for more on
conventions for choosing type names.
Here are some recommended primary console names:
linux -- Linux console driver
freebsd -- FreeBSD
netbsd -- NetBSD
bsdos -- BSD/OS
If you are responsible for integrating ncurses for one of these
distribution, please either use the recommended name or get back
to us explaining why you don't want to, so we can work out nomenclature
that will make users' lives easier rather than harder.
RECENT XTERM VERSIONS
The terminfo database file included with this distribution assumes you
are running an XFree86 xterm based on X11R6 (i.e., xterm-r6). The
earlier X11R5 entry (xterm-r5) is provided as well.
If you are running XFree86 version 3.2 (actually 3.1.2F and up), you
should consider using the xterm-xf86-v32 (or later, the most recent
version is always named "xterm-xfree86") entry, which adds ANSI color
and the VT220 capabilities which have been added in XFree86. If you
are running a mixed network, however, where this terminal description
may be used on an older xterm, you may have problems, since
applications that assume these capabilities will produce incorrect
output on the older xterm (e.g., highlighting is not cleared).
CONFIGURING FALLBACK ENTRIES
In order to support operation of ncurses programs before the terminfo
tree is accessible (that is, in single-user mode or at OS installation
time) the ncurses library can be compiled to include an array of
pre-fetched fallback entries.
These entries are checked by setupterm() only when the conventional
fetches from the terminfo tree and the termcap fallback (if configured)
have been tried and failed. Thus, the presence of a fallback will not
shadow modifications to the on-disk entry for the same type, when that
entry is accessible.
By default, there are no entries on the fallback list. After you
have built the ncurses suite for the first time, you can change
the list (the process needs infocmp(1)). To do so, use the script
MKfallback.sh. A configure script option --with-fallbacks does this
(it accepts a comma-separated list of the names you wish, and does
not require a rebuild).
If you wanted (say) to have linux, vt100, and xterm fallbacks, you
would use the commands
cd ncurses;
MKfallback.sh linux vt100 xterm >fallback.c
Then just rebuild and reinstall the library as you would normally.
You can restore the default empty fallback list with
MKfallback.sh >fallback.c
The overhead for an empty fallback list is one trivial stub function.
Any non-empty fallback list is const-ed and therefore lives in sharable
text space. You can look at the comment trailing each initializer in
the generated ncurses/fallback.c file to see the core cost of the
fallbacks. A good rule of thumb for modern vt100-like entries is that
each one will cost about 2.5K of text space.
BSD CONVERSION NOTES:
If you need to support really ancient BSD programs, you probably
want to configure with the --enable-bsdpad option. What this does
is enable code in tputs() that recognizes a numeric prefix on a
capability as a request for that much trailing padding in milliseconds.
There are old BSD programs that do things like tputs("50").
(If you are distributing ncurses as a support-library component of
an application you probably want to put the remainder of this section
in the package README file.)
The following note applies only if you have configured ncurses with
--enable-termcap.
------------------------------- CUT HERE --------------------------------
If you are installing this application privately (either because you
have no root access or want to experiment with it before doing a root
installation), there are a couple of details you need to be aware of.
They have to do with the ncurses library, which uses terminfo rather
than termcap for describing terminal characteristics.
Though the ncurses library is terminfo-based, it will interpret your
TERMCAP variable (if present), any local termcap files you reference
through it, and the system termcap file. However, in order to avoid
slowing down your application startup, it will only do this once per
terminal type!
The first time you load a given terminal type from your termcap
database, the library initialization code will automatically write it
in terminfo format to a subdirectory under $HOME/.terminfo. After
that, the initialization code will find it there and do a (much
faster) terminfo fetch.
Usually, all this means is that your home directory will silently grow
an invisible .terminfo subdirectory which will get filled in with
terminfo descriptions of terminal types as you invoke them. If anyone
ever installs a global terminfo tree on your system, this will quietly
stop happening and your $HOME/.terminfo will become redundant.
The objective of all this logic is to make converting from BSD termcap
as painless as possible without slowing down your application (termcap
compilation is expensive).
If you don't have a TERMCAP variable or custom personal termcap file,
you can skip the rest of this dissertation.
If you *do* have a TERMCAP variable and/or a custom personal termcap file
that defines a terminal type, that definition will stop being visible
to this application after the first time you run it, because it will
instead see the terminfo entry that it wrote to $HOME/terminfo the
first time around.
Subsequently, editing the TERMCAP variable or personal TERMCAP file
will have no effect unless you explicitly remove the terminfo entry
under $HOME/terminfo. If you do that, the entry will be recompiled
from your termcap resources the next time it is invoked.
To avoid these complications, use infocmp(1) and tic(1) to edit the
terminfo directory directly.
------------------------------- CUT HERE --------------------------------
USING NCURSES WITH AFS:
AFS treats each directory as a separate logical filesystem, you
can't hard-link across them. The --enable-symlinks option copes
with this by making tic use symbolic links.
USING NCURSES WITH EMACS:
GNU Emacs has its own termcap support. By default, it uses a mixture
of those functions and code linked from the host system's libraries.
You need to foil this and shut out the GNU termcap library entirely.
In order to do this, hack the Linux config file (s/linux.h) to contain
a #define TERMINFO and set the symbol LIBS_TERMCAP to "-lncurses".
We have submitted such a change for the 19.30 release, so it may
already be applied in your sources -- check for the #define TERMINFO.
USING NCURSES WITH GPM:
Ncurses 4.1 and up can be configured to use GPM (General Purpose Mouse)
which is used on Linux console. Be aware that GPM is commonly
installed as a shared library which contains a wrapper for the curses
wgetch() function (libcurses.o). Some integrators have simplified
linking applications by combining all of libcurses.so (the BSD curses)
into the libgpm.so file, producing symbol conflicts with ncurses. You
may be able to work around this problem by linking as follows:
cc -o foo foo.o -lncurses -lgpm -lncurses
but the linker may not cooperate, producing mysterious errors.
A patched version of gpm is available:
ftp.clark.net:/pub/dickey/ncurses/gpm-1.10-970125.tgz
This patch is incorporated in gpm 1.12; however some integrators
are slow to update this library.
USING NCURSES WITH ELVIS:
To use ncurses as the screen-painting library for Elvis, apply the
following patch to the Elvis curses
*** curses.c.orig Sun Jun 26 05:48:23 1994
--- curses.c Sun Feb 11 16:50:41 1996
***************
*** 986,992 ****
{
if (has_IM)
do_IM();
! do_IC();
qaddch(ch);
if (has_EI)
do_EI();
--- 986,995 ----
{
if (has_IM)
do_IM();
!#ifdef NCURSES_VERSION
! else /* ncurses does insertion in a slightly nonstandard way */
!#endif
! do_IC();
qaddch(ch);
if (has_EI)
do_EI();
This patch is for elvis-1.8pl4 but it can even be used for elvis-1.8pl3 with
an offset of -11 lines.
BUGS:
Send any feedback to the ncurses mailing list at
bug-ncurses@gnu.org. To subscribe send mail to
bug-ncurses-request@gnu.org with body that reads:
subscribe ncurses <your-email-address-here>
The Hacker's Guide in the misc directory includes some guidelines
on how to report bugs in ways that will get them fixed most quickly.