circumstances make doesn't like the conditional I wrote.
(also, what the previous log message says about a bump is wrong; I bumped
only ncursesw. but there shouldn't be any need to bump ncurses for this.)
completely; rectify that. Now setting USE_NCURSES=yes won't install
pkgsrc ncurses on e.g. Linux where it's really built in.
Not sure if a version bump is wanted, but since I just did one we'll
ride that.
not only with the different directory structure under Mac OS X (due to
the case-insensitive file-system) but also with the files which don't
get installed under Solaris. Bump the package revision of the "ncurses"
package as the binary package changed under Mac OS X.
This fixes PR pkg/43446 by Edgar Fuss.
/usr/pkg/include in the header search path shadows some of the system's
curses headers (happens when building or developing curses application
outside of the pkgsrc build system).
This changes the buildlink3.mk files to use an include guard for the
recursive include. The use of BUILDLINK_DEPTH, BUILDLINK_DEPENDS,
BUILDLINK_PACKAGES and BUILDLINK_ORDER is handled by a single new
variable BUILDLINK_TREE. Each buildlink3.mk file adds a pair of
enter/exit marker, which can be used to reconstruct the tree and
to determine first level includes. Avoiding := for large variables
(BUILDLINK_ORDER) speeds up parse time as += has linear complexity.
The include guard reduces system time by avoiding reading files over and
over again. For complex packages this reduces both %user and %sys time to
half of the former time.
New features and improvements:
* library
o new flavor of the ncurses library provides rudimentary
support for POSIX threads. Several functions are reentrant,
but most require either a window-level or screen-level
mutex.
(This is API-compatible, but not ABI-compatible with
the normal library).
o add NCURSES_OPAQUE symbol to curses.h, will use to make
structs opaque in selected configurations.
o add NCURSES_EXT_FUNCS and NCURSES_EXT_COLORS symbols
to curses.h to make it simpler to tell if the extended
functions and/or colors are declared.
o add wresize to C++ binding
o eliminate fixed-buffer vsprintf calls in C++ binding.
o add several functions to C++ binding which wrap C
functions that pass a WINDOW* parameter.
o adapt mouse-handling code from menu library in form-library
o improve tracing for form library, showing created forms, fields, etc.
o make $NCURSES_NO_PADDING feature work for termcap interface .
o add check to trace-file open, if the given name is a
directory, add ".log" to the name and try again.
o several new manpages: curs_legacy.3x, curs_memleaks.3x,
curs_opaque.3x and curs_threads.3x
* programs:
o modified three test-programs to demonstrate the threading
support in this version: ditto, rain, worm.
o several new test-programs: demo_panels, dots_mvcur,
inch_wide, inchs, key_name, key_names, savescreen,
savescreen.sh test_arrays, test_get_wstr, test_getstr,
test_instr, test_inwstr and test_opaque.
o add adacurses-config to the Ada95 install.
o modify tic -f option to format spaces as \s to prevent
them from being lost when that is read back in unformatted
strings.
o The tack program is now distributed separately from
ncurses.
* terminal database
o added entries:
+ Eterm-256color, Eterm-88color and rxvt-88color
+ aterm
+ konsole-256color
+ mrxvt
+ screen.mlterm
+ screen.rxvt
+ teraterm4.59 is now the primary primary teraterm
entry, renamed original to teraterm2.3
+ 9term terminal
+ Newbury Data entries
o updated/improved entries:
+ gnome to version 2.22.3
+ h19, z100
+ konsole to version 1.6.6
+ mlterm, mlterm+pcfkeys
+ xterm, and building-blocks for function-keys to
xterm patch #230.
Major bug fixes:
* add logic to tic for cancelling strings in user-defined
capabilities (this is needed for current konsole terminfo
entry).
* modify mk-1st.awk so the generated makefile rules for linking
or installing shared libraries do not first remove the library,
in case it is in use, e.g., libncurses.so by /bin/sh.
* correct check for notimeout in wgetch.
* fix a sign-extension bug in infocmp's repair_acsc function.
* change winnstr to stop at the end of the line.
* make Ada95 demo_panels example work.
* fix for adding a non-spacing character at the beginning of a line.
* fill in extended-color pair to make colors work for
wide-characters using extended-colors.
* improve refresh of window on top of multi-column characters,
taking into account split characters on left/right window
boundaries.
* modify win_wchnstr to ensure that only a base cell is returned
for each multi-column character.
* improve waddch and winsch handling of EILSEQ from mbrtowc by
using unctrl to display illegal bytes rather than trying to
append further bytes to make up a valid sequence.
* restore curs_set state after endwin/refresh
* modify keyname to use "^X" form only if meta has been called,
or if keyname is called without initializing curses, e.g., via
initscr or newterm.
* modify unctrl to check codes in 128-255 range versus isprint.
If they are not printable, and locale was set, use a "M-" or
"~" sequence.
* improve resizeterm by moving ripped-off lines, and repainting
the soft-keys.
* modify form library to accept control characters such as
newline in set_field_buffer, which is compatible with Solaris.
* use NCURSES_MOUSE_MASK in definition of BUTTON_RELEASE, etc.,
to make those work properly with the --enable-ext-mouse
configuration
* correct some functions in Ada95 binding which were using
return value from C where none was returned.
* reviewed/fixed issues reported by Coverity and Klocwork tools.
Portability:
* configure script:
o new options:
--disable-big-strings
control whether static string tables are generated
as single large strings (to improve startup
performance), or as array of individual strings.
--disable-relink
control whether shared libraries are relinked
(during install) when rpath is enabled.
--disable-tic-depends
make explicit whether tic library depends on
ncurses/ncursesw library.
--enable-mixed-case
override the configure script's check if the
filesystem supports mixed-case filenames. This
allows one to control how the terminal database
maps to the filesystem. For filesystems that do
not support mixed-case, the library uses generate
2-character (hexadecimal) codes for the lower-level
of the filesystem terminfo database
--enable-reentrant
builds a different flavor of the ncurses library
(ncursest) which improves reentrant use of the
library by reducing global and static variables
(see the "--with-pthread" option for the threaded
support).
--enable-weak-symbols
use weak-symbols for linking to the POSIX thread
library, and use the same soname for the ncurses
shared library as the normal library (caveat: the
ABI is for the threaded library, which makes global
data accessed via functions).
--with-pthread
build with the POSIX thread library (tested with
AIX, GNU/Linux, FreeBSD, OpenBSD, HPUX, IRIX64,
Solaris, Tru64).
--with-ticlib
build/install the tic-support functions in a separate
library
o improved options:
--enable-ext-colors
requires the wide-character configuration.
--with-chtype
ignore option value "unsigned" is always added to
the type in curses.h; do the same for --with-mmask-t.
--with-dmalloc
build-fix for redefinition of strndup.
--with-hashed-db
+ accepts a parameter which is the install-prefix
of a given Berkeley Database.
+ the $LIBS environment variable overrides the
search for the db library.
--without-hashed-db
assumed when "--disable-database" is used.
* other configure/build issues:
o build-fixes for LynxOS
o modify shared-library rules to allow FreeBSD 3.x to use rpath.
o build-fix for FreeBSD "contemporary" TTY interface.
o build-fixes for AIX with libtool.
o build-fixes for Darwin and libtool.
o modify BeOS-specific ifdef's to build on Haiku.
o corrected gcc options for building shared libraries on
Solaris and IRIX64.
o change shared-library configuration for OpenBSD, make rpath work.
o build-fixes for using libutf8, e.g., on OpenBSD 3.7
o add "-e" option in ncurses/Makefile.in when generating
source-files to force earlier exit if the build environment
fails unexpectedly.
o add support for shared libraries for QNX.
o change delimiter in MKlib_gen.sh from '%' to '@', to
avoid substitution by IBM xlc to '#' as part of its
extensions to digraphs.
* library:
o rewrite wrapper for wcrtomb, making it work on Solaris.
This is used in the form library to determine the length
of the buffer needed by field_buffer.
o add/use configure script macro CF_SIG_ATOMIC_T, use
the corresponding type for data manipulated by signal
handlers.
o set locale in misc/ncurses-config.in since it uses a
range
o disable GPM mouse support when $TERM does not happen
to contain "linux", since Gpm_Open no longer limits its
assertion to terminals that it might handle, e.g., within
"screen" in xterm.
o reset mouse file-descriptor when unloading GPM library.
* test programs:
o update test programs to build/work with various UNIX
curses for comparisons.
been the case before, but wasn't due to a missing default vaule.
+ Move the BUILDLINK_TRANSFORM remapping of -lncurses into the section
which only applies if we're using the built-in ncurses, and map it to
-l${BUILTIN_LIBNAME.ncurses}.
This should fix the issue reported for Mac OS X where "-lncurses" was
being mapped to nothing by the wrapper scripts.
always be used in other builtin.mk files.
+ In the USE_BUILTIN.* == "yes" case, set BUILDLINK_LIBNAME.* to the
corresponding BUILTIN_LIBNAME.* value so that BUILDLINK_LIBNAME.*
can always be used in other buildlink3.mk files.
that need basic termlib functionality, i.e. tgetent(), tgoto(),
tputs(), etc. Together with the termlib.builtin.mk file, they will
use either a built-in termcap library, a built-in X/Open "enhanced"
curses library, or ncurses to provide these functions.
+ Add BUILDLINK_LIBNAME.* definitions to the various curses buildlink3.mk
and builtin.mk files that give the "base" library name of the curses
library, e.g. curses, ncurses, etc. These are used by the termlib
files to set BUILDLINK_LIBNAME.termlib.
-lcurses so that GNU configure scripts that check for those ahead of
<ncurses.h> and -lncurses will still use ncurses instead of picking up
a system curses.
This was a regression introduced by my recent ncurses commits.
used by packages that need curses. From curses.buildlink3.mk:
CURSES_DEFAULT
This value represents the type of curses we wish to use on the
system. Setting this to "curses" means that the system curses
implementation is fine.
Possible: curses, ncurses, pdcurses
+ Move all code to detect a built-in version of curses into a
curses.builtin.mk file.
+ Add code to {n,pd}curses/buildlink3.mk so make the headers and
libraries usable as <curses.h> and -lcurses if _PKG_USE_CURSES is
defined. _PKG_USE_CURSES is only defined by curses.buildlink3.mk.
+ Improve the detection of native ncurses in ncurses/builtin.mk and
allow headers and libraries to be usable as <ncurses.h> and -lncurses.
Bump revisions of both ncurses and ncursesw packages.
fixes based on Coverity report:
------------------------------
- remove dead code in test/bs.c
- remove dead code in test/demo_defkey.c
- remove an unused assignment in progs/infocmp.c
- fix a limit check in tack/ansi.c tools_charset()
- fix tack/ansi.c tools_status() to perform the VT320/VT420
tests in request_cfss(). The function had exited too soon.
- fix a memory leak in tic.c's make_namelist()
- fix a couple of places in tack/output.c which did not check for
EOF.
- fix a loop-condition in test/bs.c
- add index checks in lib_color.c for color palettes
- add index checks in progs/dump_entry.c for version_filter()
handling of V_BSD case.
- fix a possible null-pointer dereference in copywin()
- fix a possible null-pointer dereference in waddchnstr()
- add a null-pointer check in _nc_expand_try()
- add a null-pointer check in tic.c's make_namelist()
- add a null-pointer check in _nc_expand_try()
- add null-pointer checks in test/cardfile.c
- fix a double-free in ncurses/tinfo/trim_sgr0.c
- fix a double-free in ncurses/base/wresize.c
- add try/catch block to c++/cursesmain.cc
other fixes prompted by inspection for Coverity report:
------------------------------------------------------
- modify ifdef's for c++ binding to use try/catch/throw statements
- add a null-pointer check in tack/ansi.c request_cfss()
- fix a memory leak in ncurses/base/wresize.c
- corrected check for valid memu/meml capabilities in
progs/dump_entry.c when handling V_HPUX case.
we can end up with a circular dependency:
ncurses -> groff -> netpbm -> flex -> bison -> gtexinfo -> ncurses
ncurses may depend on groff because of USE_TOOLS+=tbl.
Break the circular dependency by using --without-manpage-tbl
if there is no tbl(1) available and we're unable to build
groff because of missing curses.
clear that these variables are completely unrelated to
BUILDLINK_TRANSFORM.
Added a legacy check that catches appearances of BUILDLINK_TRANSFORM.*.
XXX: Where should incompatible changes in pkgsrc be documented?
Hashed-databases are supported for storing terminal descriptions.
Support for magic-cookies such as hpterm was improved.
Support for rpath in shared libraries was revised.
Many bugfixes were made for memory leaks, wide-character support, portability,
and more.
and add a new helper target and script, "show-buildlink3", that outputs
a listing of the buildlink3.mk files included as well as the depth at
which they are included.
For example, "make show-buildlink3" in fonts/Xft2 displays:
zlib
fontconfig
iconv
zlib
freetype2
expat
freetype2
Xrender
renderproto
RECOMMENDED is removed. It becomes ABI_DEPENDS.
BUILDLINK_RECOMMENDED.foo becomes BUILDLINK_ABI_DEPENDS.foo.
BUILDLINK_DEPENDS.foo becomes BUILDLINK_API_DEPENDS.foo.
BUILDLINK_DEPENDS does not change.
IGNORE_RECOMMENDED (which defaulted to "no") becomes USE_ABI_DEPENDS
which defaults to "yes".
Added to obsolete.mk checking for IGNORE_RECOMMENDED.
I did not manually go through and fix any aesthetic tab/spacing issues.
I have tested the above patch on DragonFly building and packaging
subversion and pkglint and their many dependencies.
I have also tested USE_ABI_DEPENDS=no on my NetBSD workstation (where I
have used IGNORE_RECOMMENDED for a long time). I have been an active user
of IGNORE_RECOMMENDED since it was available.
As suggested, I removed the documentation sentences suggesting bumping for
"security" issues.
As discussed on tech-pkg.
I will commit to revbump, pkglint, pkg_install, createbuildlink separately.
Note that if you use wip, it will fail! I will commit to pkgsrc-wip
later (within day).
