<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 3.0//EN">
<!--
- $Id: announce.html.in,v 1.53 2004/01/18 01:27:26 tom Exp $
+ $Id: announce.html.in,v 1.57 2005/10/01 22:07:23 tom Exp $
-->
<HTML>
<HEAD>
and has all the other SYSV-curses enhancements over BSD curses.<P>
In mid-June 1995, the maintainer of 4.4BSD curses declared that he
-considered 4.4BSD curses obsolete, and is encouraging the keepers of
-Unix releases such as BSD/OS, freeBSD and netBSD to switch over to
+considered 4.4BSD curses obsolete, and encouraged the keepers of
+Unix releases such as BSD/OS, FreeBSD and NetBSD to switch over to
ncurses.<P>
The ncurses code was developed under GNU/Linux.
<H1>Release Notes</H1>
-This release is designed to be upward compatible from ncurses 5.0 and 5.3;
+This release is designed to be upward compatible from ncurses 5.0 through 5.4;
very few applications will require recompilation, depending on the platform.
-These are the highlights from the change-log since ncurses 5.3 release.
+These are the highlights from the change-log since ncurses 5.4 release.
<p>
Interface changes:
<ul>
- <li>add the remaining functions for X/Open curses wide-character
- support.
- <ul>
- <li>pecho_wchar()
- <li>slk_wset()
- </ul>
- These are only available if the library is configured using the
- <kbd>--enable-widec</kbd> option.
-
- <li>write <code>getyx()</code> and related 2-return macros in terms of
- <code>getcury()</code>,
- <code>getcurx()</code>, etc.
-
- <li>simplify ifdef for <code>bool</code> declaration in curses.h
-
- <li>modify ifdef's in curses.h that disabled use of
- <code>__attribute__()</code> for g++, since recent versions
- implement the cases which ncurses uses.
-
- <li>add <code>key_defined()</code> function, to tell which keycode a
- string is bound to.
+ <li>terminfo installs "xterm-new" as "xterm" entry rather than
+ "xterm-old" (aka xterm-r6).
+
+ <li>terminfo data is installed using the tic -x option (few systems
+ still use ncurses 4.2).
+
+ <li>modify C++ binding to work with newer C++ compilers by providing
+ initializers and using modern casts. Old-style header names are
+ still used in this release to allow compiling with not-so-old
+ compilers.
+
+ <li>modify parameter type in c++ binding for insch() and mvwinsch() to
+ be consistent with underlying ncurses library (was char, is chtype).
+
+ <li>change NCursesWindow::err_handler() to a virtual function.
+
+ <li>form and menu libraries now work with wide-character data.
+ Applications which bypassed the form library and manipulated the
+ FIELD.buf data directly will not work properly with libformw,
+ since that no longer points to an array of char. The
+ set_field_buffer() and field_buffer() functions translate to/from
+ the actual field data.
+
+
+ <li>add symbol to curses.h which can be used to suppress include of
+ stdbool.h, e.g.,
+ <pre>
+ #define NCURSES_ENABLE_STDBOOL_H 0
+ #include <curses.h>
+ </pre>
+
+ <li>change SP->_current_attr to a pointer, adjust ifdef's to ensure
+ that libtinfo.so and libtinfow.so have the same ABI. The reason
+ for this is that the corresponding data which belongs to the
+ upper-level ncurses library has a different size in each model.
+
+ <li>winnstr() now returns multibyte character strings for the
+ wide-character configuration.
+
+ <li>assume_default_colors() no longer requires that
+ use_default_colors() be called first.
+
+ <li>data_ahead() now works with wide-characters.
+
+ <li>slk_set() and slk_wset() now accept and store multibyte or
+ multicolumn characters.
+
+ <li>start_color() now returns OK if colors have
+ already been started.
+ start_color() also returns ERR if it cannot allocate memory.
+
+ <li>pair_content() now returns -1 for consistency with init_pair()
+ if it corresponds to the default-color.
+
+ <li>unctrl() now returns null if its parameter does not correspond
+ to an unsigned char.
+
</ul>
New features and improvements:
<ul>
<li>library
<ul>
- <li>implement logic in lib_mouse.c to handle position reports which
- are generated when XFree86 xterm is initialized with private
- modes 1002 or 1003. These are returned to the application as the
- REPORT_MOUSE_POSITION mask, which was not implemented.
- <li>modify soft-key initialization to use A_REVERSE if A_STANDOUT
- would not be shown when colors are used, i.e., if ncv#1 is set in
- the terminfo as is done in "screen".
- <li>add configure option for FreeBSD sysmouse, --with-sysmouse, and
- implement library support for that.
+ <li>environment variable NCURSES_NO_UTF8_ACS supports
+ miscellaneous terminal emulators which ignore alternate
+ character set escape sequences when in UTF-8 mode.
+
+ <li>modify initialization of key lookup table so that if an extended
+ capability (tic -x) string is defined, and its name begins with 'k',
+ ncurses will automatically treat it as a key.
+
+ <li>change GPM initialization, using dl library to load it
+ dynamically at runtime.
+
+ <li>form, menu and panel libraries support debug-tracing.
+
</ul>
- <li>programs:
+ <li>add NCURSES-Programming-HOWTO.html by Pradeep Padala
+ (see http://tldp.org/HOWTO/NCURSES-Programming-HOWTO/).
- <ul>
- <li>tack:
+ <li>programs:
+ <li>infocmp:
<ul>
- <li>allow it to run from fallback terminfo data.
- <li>reset colors after each color test, correct a place
- where <code>exit_standout_mode</code> was used instead of
- <code>exit_attribute_mode</code>.
- <li>improve <code>bce</code> test by making it set colors other
- than black on white.
+ <li>The -i option now matches 8-bit controls against its table
+ entries, e.g., so it can analyze the xterm-8bit entry.
+
+ <li>add "-x" option to infocmp like tic's "-x", for use in "-F"
+ comparisons. This modifies infocmp to only report extended
+ capabilities if the -x option is given, making this more
+ consistent with tic. Some scripts may break, since infocmp
+ previous gave this information without an option.
+
</ul>
<li>tic:
<ul>
- <li>handle a case where an entry has no
- description, and capabilities begin on the same line as the entry
- name.
- <li>allow a terminfo entry with a leading 2-character name.
- <li>improved warnings when suppressing items to fit in termcap's
- 1023-byte limit.
- <li>add check for multiple "tc=" clauses in a termcap.
- <li>correct logic for resolving "use=" clauses
- allow infocmp and tic to show cancelled capabilities.
- <li>check for incomplete line-drawing character mapping.
- <li>check for missing/empty/illegal terminfo name.
- </ul>
- <li>tput:
- <ul>
- <li>modify tput to use the same parameter analysis as tparm() does,
- to provide for user-defined strings, e.g., for xterm title, a
- corresponding capability might be
- title=\E]2;%p1%s^G,
+ <li>modify termcap-parsing to retain 2-character aliases at the
+ beginning of an entry if the "-x" option is used in tic.
+
+ <li>filter out long extended names when translating to termcap
+ format. Only two characters are permissible for termcap
+ capability names.
+
+ <li>correct translation of "%%" in terminfo format to termcap,
+ e.g., using "tic -C".
+
+ <li>modify the "-c -v" options to ignore delays when comparing
+ strings. Also modify it to ignore a canceled sgr string,
+ e.g., for terminals which cannot properly combine attributes
+ in one control sequence.
+
+ <li>add a check for improperly ended strings, i.e., where a
+ following line begins in column 1.
+
+ <li>add a check in tic for terminfo entries having an sgr0 but no
+ sgr string. This confuses Tru64 and HPUX curses when combined
+ with color, e.g., making them leave line-drawing characters in
+ odd places.
+
+ <li>add check (with debug configuration) that provides about the
+ runtime changes that would be made to sgr0 for termcap
+ applications.
+
</ul>
<li>tset:
<ul>
- <li>use the system's default values for CKILL and other default
- control characters.
- <li>correct interchanged defaults for kill and interrupt
- characters, which caused it to report unnecessarily.
+
+ <li>add -c and -w options to allow user to suppress ncurses'
+ resizing of the terminal emulator window in the special case
+ where it is not able to detect the true size.
+
</ul>
- </ul>
</ul>
Major bug fixes:
<ul>
- <li>prevent recursion in wgetch() via wgetnstr() if the connection
- cannot be switched between cooked/raw modes because it is not a TTY.
- <li>correct a case in _nc_remove_string(), used by
- <code>define_key()</code>, to avoid infinite loop if the given string
- happens to be a substring of other strings which are assigned to keys.
+ <li>improve logic in tgetent() which adjusts the termcap "me" string
+ to work with ISO-2022 string used in xterm-new.
+ This is a feature that was incompletely implemented in ncurses 5.3.
+ ncurses attempts to provide termcap clients with the portion of
+ the sgr0 (termcap "me") string that does not reset line-drawing.
- <li>modify logic of acsc to use the original character if no mapping is
- defined, rather than a null.
+ <li>cells in the WINDOW which are continuations of a multicolumn
+ character are encoded differently, making repainting more
+ reliable.
- <li>several improvements for handling multi-cell display of wide
- characters.
- <ul>
- <li>modify setcchar() to allow converting control characters to complex
- characters.
- <li>correct handling of multibyte characters in waddch_literal() which
- force wrapping because they are started too late on the line.
- <li>modify setcchar() to allow for wchar_t input strings that have
- more than one spacing character.
- </ul>
+ <li>amend change to setupterm() in ncurses 5.4 (20030405) which would
+ reuse the value of cur_term if the same output was selected. This
+ now reuses it only when setupterm() is called from tgetent(),
+ which has no notion of separate SCREENs. Note that tgetent() must
+ be called after initscr() or newterm() to use this feature.
- <li>other fixes for wide-character support:
- <ul>
- <li>rewrote lib_acs.c to allow PutAttrChar() to decide how to render
- alternate-characters, i.e., to work with Linux console and UTF-8
- locale.
- <li>implement a workaround so that line-drawing works with screen's
- crippled UTF-8 support (tested with 3.9.13). This only works with
- the wide-character support (--enable-widec); the normal library
- will simply suppress line-drawing when running in a UTF-8 locale in
- screen.
- <li>corrections to lib_get_wstr.c:
- <ul>
- <li>null-terminate buffer passed to setcchar(), which occasionally
- failed.
- <li>map special characters such as erase- and kill-characters into
- key-codes so those will work as expected even if they are not
- mentioned in the terminfo.
- </ul>
- </ul>
+ <li>make setcchar() now works when its wchar_t* parameter is
+ pointing to a string which contains more data than can be converted.
- <li>modify <code>setupterm()</code> to check if the terminfo and
- terminal-modes have already been read. This ensures that it does not
- reinvoke <code>def_prog_mode()</code> when an application calls more
- than one function, such as <code>tgetent()</code> and
- <code>initscr()</code>.
+ <li>win_wchnstr() now works for more than one cell.
- <li>fix form_driver() cases for REQ_CLR_EOF, REQ_CLR_EOL, REQ_DEL_CHAR,
- REQ_DEL_PREV and REQ_NEW_LINE, which did not ensure the cursor was at
- the editing position before making modifications.
+ <li>resizeterm() now processes all levels of window hierarchy.
- <li>correct <code>keybound()</code>, which reported definitions in the
- wrong table, i.e., the list of definitions which are disabled by
- <code>keyok()</code>.
+ <li>disable GPM mouse support when $TERM happens to be prefixed with
+ "xterm". Gpm_Open() would otherwise assert that it can deal with
+ mouse events in this case.
- <li>fixes related to safe_sprintf.c:
- <ul>
- <li>correct an allocation size in safe_sprintf.c for the "*" format
- code.
- <li>correct safe_sprintf.c to not return a null pointer if the format
- happens to be an empty string.
- <li>make return value from _nc_printf_string() consistent. Before,
- depending on whether --enable-safe-sprintf was used, it might not
- be cached for reallocating.
- </ul>
+ <li>add SP->_screen_acs_map[], used to ensure that mapping of missing
+ line-drawing characters is handled properly. For example, ACS_DARROW
+ is absent from xterm-new, and it was coincidentally displayed the
+ same as ACS_BTEE.
- <li>other low-level improvements to the optimization code include:
- <ul>
- <li>if the output is a socket or other non-tty device, use 1
- millisecond
- for the cost in mvcur; previously it was 9 milliseconds because the
- baudrate was not known.
- <li>modify lib_getch.c to avoid recursion via wgetnstr() when the
- input
- is not a tty and consequently mode-changes do not work.
- <li>fix several places in tack/pad.c which tested and used the
- parameter- and parameterless strings inconsistently.
- <li>change several tputs() calls in scrolling code to use putp(), to
- enable padding which may be needed for some terminals.
- <li>improve mvcur() by checking if it is safe to move when video
- attributes are set (msgr), and if not, reset/restore attributes
- within that function rather than doing it separately in the GoTo()
- function in tty_update.c.
- <li>use tputs() rather than putp() in a few cases in tty_update.c
- since the corresponding delays are proportional to the number of
- lines affected: repeat_char, clr_eos, change_scroll_region.
- </ul>
-
- <li>correct line/column reference in adjust_window(), needed to make
- special windows such as curscr track properly when resizing.
-
- <li>fix a potential recursion between napms() and _nc_timed_wait()
-
- <li>rewrote lib_insch.c, combining it with lib_insstr.c so both handle
- tab and other control characters consistently.
-
- <li>do not try to open gpm mouse driver if standard output is not a
- tty; the gpm library does not make this check.
</ul>
Portability:
<ul>
<li>new options:
<dl>
- <dt>--with-abi-version option.
- <dd>addresses platforms where packagers have diverged from
- ncurses ABI numbering.
- <dt>--with-manpage-format=catonly
- <dd>addresses
- behavior of BSDI, allow install of man+cat files on NetBSD,
- whose behavior has diverged by requiring both to be present.
- <dt>--with-manpage-aliases
- <dd>extends
- "--with-manpage-aliases" to provide the option of generating
- ".so" files rather than symbolic links for manpage aliases.
- <dt>--with-rel-version
- <dd>workaround to force
- libtool on Darwin generate the "same" library names as with
- the "--with-shared" option. The Darwin ld program does not
- work well with a zero as the minor-version value.
- <dt>--with-trace
- <dd>simplifies defining TRACE to incorporate trace() in libraries.
+ <dt>--enable-largefile
+ <dd>set compiler and linker flags to use largefile support.
+
+ <dt>--enable-ext-colors
+ <dd>Allow encoding of 256 foreground
+ and background colors, e.g., with the xterm-256color or
+ xterm-88color terminfo entries.
+ This requires ABI 6 because it changes the size of cchar_t.
+
+ <dt>--enable-ext-mouse
+ <dd>This defines NCURSES_MOUSE_VERSION 2, and modifies the encoding of
+ mouse events to support wheel mice, which may transmit buttons
+ 4 and 5. This works with xterm and similar terminal
+ emulators.
+ This requires ABI 6 because it changes the encoding of
+ mouse events.
+
+ <dt>--with-chtype
+ <dd>overriding of the non-LP64 model's use of chtype
+
+ <dt>--with-mmask-t
+ <dd>overriding of the non-LP64 model's use of mmask_t
+
+ <dt>--without-xterm-new
+ <dd>Installs "xterm-old" as the "xterm" entry of the terminfo
+ database.
+
</dl>
-
+
+ <li>The --with-termlib option now accepts a value which sets
+ the name of the terminfo library. This would allow a packager to
+ build libtinfow.so renamed to coincide with libtinfo.so
+
<li>fixes/improvements for cross-compiling:
-
- <ul>
- <li>allow BUILD_CC and related configure script variables to be
- overridden from the environment.
- <li>use AC_CHECK_TOOL to get proper values for AR and LD for cross
- compiling.
- <li>use <code>$cross_compiling</code> variable in configure script
- rather than comparing <code>$host_alias</code> and
- <code>$target</code> alias, since "host" is traditionally misused
- in autoconf to refer to the target platform.
- <li>modify run_tic.in to avoid using wrong shared libraries when
- cross-compiling.
- </ul>
-
- <li>fixes for Mac OS X:
- <ul>
- <li>fix a redefinition of <code>$RANLIB</code> in the configure
- script when libtool is used.
- <li>modify MKlib_gen.sh to avoid passing "#" tokens through the C
- preprocessor. This works around Mac OS X's preprocessor, which
- insists on adding a blank on each side of the token.
- </ul>
-
- <li>workarounds for broken tools:
+
<ul>
- <li>add configure check for wchar_t and wint_t types, rather than
- rely on preprocessor definitions. Also work around for gcc
- <code>fixinclude</code> bug which creates a shadow copy of
- curses.h if it sees these symbols apparently typedef'd.
- <li>modify configure script to omit -Winline for gcc 3.3, since
- that feature is broken.
- <li>several script fixes to work around the ironically named
- <em>POSIXLY_CORRECT</em> feature of GNU sed 4.0.
- <li>modify configure script to avoid using "head -1", which does
- not work if POSIXLY_CORRECT (sic) is set.
- <li>update configure script to reflect fix for
- AC_PROG_GCC_TRADITIONAL, which is broken in autoconf 2.5x for Mac
- OS X 10.2.3.
- <li>repair check for missing C++ compiler, which is broken in
- autoconf 2.5x by hardcoding it to g++.
+ <li>suppress $suffix in misc/run_tic.sh when cross-compiling.
+ This allows cross-compiles to use the host's tic program to
+ handle the "make install.data" step.
+
+ <li>correct BUILD_CPPFLAGS substitution in ncurses/Makefile.in, to allow
+ cross-compiling from a separate directory tree.
+
</ul>
-
- <li>corrected ifdef's relating to configure check for wchar_t, etc.
- <li>remove configure script check to allow -Wconversion for older
- versions of gcc
- <li>modify configure script to accommodate libtool 1.5, as well as
- add an parameter to the "--with-libtool" option which can specify
- the pathname of libtool.
- <li>change several sed scripts to avoid using "\+" since it is not a
- BRE (basic regular expression). One instance caused terminfo.5 to
- be misformatted on FreeBSD.
- <li>use '%' as sed substitute delimiter in run_tic script to avoid
- problems with pathname delimiters such as ':' and '@'.
- <li>add -D_XOPEN_SOURCE=500 if needed when configuring with
- "--enable-widec", to get <code>mbstate_t</code> declaration on HPUX
- 11.11.
+
</ul>
<li>library:
<ul>
- <li>adjust include-options in CF_ETIP_DEFINES to avoid missing
- ncurses_dll.h, fixing special definitions that may be needed for
- etip.h.
- <li>modify CF_LIB_SUFFIX for Itanium releases of HP-UX, which use a
- ".so" suffix.
- <li>improve ifdef's to guard against redefinition of wchar_t and wint_t
- in curses.h.
- <li>remove an #undef for KEY_EVENT from curses.tail used in the
- experimental NCURSES_WGETCH_EVENTS feature. The #undef confuses
- Debian <code>dpkg</code>'s build script.
+
+ <li>add ifdef's for _LP64 in curses.h to avoid using wasteful
+ 64-bits for chtype and mmask_t, but add configure option
+ --disable-lp64 in case anyone used that configuration.
+
+ <li>modify C++ binding to use some C internal functions to make it
+ compile properly on Solaris (and other platforms).
+
+ <li>remove check in newwin() that prevents allocating windows that
+ extend beyond the screen (Solaris does this).
+
+ <li>check for <code>nl_langinfo(CODESET)</code>, use it if available.
+ This replaces ad hoc tests of environment variables to
+ check if the terminal is setup for UTF-8 encoding.
+ Applications which do not call <code>setlocale()</code>
+ should be corrected, to make them work properly with
+ UTF-8 encoding.
+ <br>
+ <strong>In particular, applications which assume (and
+ do not call <code>setlocale()</code>) that
+ Latin-1 codes are printable will no longer work in
+ a UTF-8 locale since the ad hoc check of environment
+ variables to see if the locale was UTF-8 is not used
+ when <code>nl_langinfo(CODESET)</code> is available.</strong>
+
+ <li>use setlocale() to query the program's current locale rather
+ than using getenv(). This supports applications which rely
+ upon legacy treatment of 8-bit characters when the locale is
+ not initialized.
+
</ul>
</ul>