py-zulip-term: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-zulip: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

lmdb: Ensure we only build IPROGS.

The test binaries haven't yet been libtoolized.

(jperkin)

qt6-qtbase: Fix build on SunOS.

(jperkin)

doc: Note meta-pkgs/php80-extensions removal.

(jperkin)

meta-pkgs: Remove php80-extensions.

lang/php80 was deleted.

(jperkin)

bulk-medium: php80 was deleted.

(jperkin)

bulk-test-python27: Spell arcticfox correctly.

(jperkin)

snort: Fix build on SunOS.

(jperkin)

snort: Remove snort-prelude option.

This cannot have worked for nearly 2 years since prelude was deleted.

(jperkin)

daq: Use C99 inttypes on SunOS.

(jperkin)

mozilla: Support illumos triple.

(jperkin)

mozilla: Support config-override.

(jperkin)

mozilla: Match rust triple on illumos.

(jperkin)

mk: Allow "illumos" as GNU host os.

For the most part the triple is still *-*-solaris*, but newer software
such as rust defines the target triple as x86_64-unknown-illumos, and
supporting this helps with patching things that try to match them.

(jperkin)

clang: Switch SunOS back to using the GCC runtime.

Upstream LLVM have removed the SunOS code from libcxx et al, despite us
offering to support it and it working fine, and so we're left with no
choice but to go back to using the GCC libraries and headers, which is a
shame.

Tested to compile at least a few different C/C++ files, and is able to
complete the builds of lang/wasi-*.

(jperkin)

qt5-qttools: Disable llvm option on SunOS.

It brings in full clang rather than just llvm.

(jperkin)

cgdb: Switch to ncurses.

This doesn't currently build with builtin NetBSD curses due to a number
of issues.  While here order things correctly.

(jperkin)

lmdb: Finish converting to libtool.

Fixes build at MAKE_JOBS=1.

(jperkin)

ncurses: Put back fake ncurses.pc handling.

Fixes build of e.g. editors/mg on macOS using builtin ncurses.

(jperkin)

nudoku: Switch over to ncurses.

This package was updated in between my posting the ncurses patch for
review and committing, so it was missed.

Remove mk/curses.buildlink3.mk include as it's a nop if ncurses is being
specifically requested, and add USE_CURSES=wide so that wide support is
enabled.

(jperkin)

ncurses: Provide a compat symlink for ncursesw.pc.

Some packages hardcode requests for this file, even though the actual
functionality is in libncurses, so just redirect there.

(jperkin)

mk: Apply pkglint autofix to curses.builtin.mk.

Note that fixes for USE_CURSES were skipped as pkglint does not
currently handle that case correctly.

(jperkin)

mariadb*: Remove ncurses workarounds.

It's unclear what they were required for previously, but with the new
ncurses package they are now actively harmful, and removing them fixes
the build.

(jperkin)

*: Revision bumps for ncurses 6.5 overhaul.

(jperkin)

*: Remove long-obsolete USE_NCURSES references.

(jperkin)

doc: Note removal of devel/ncursesw.

(jperkin)

ncursesw: Remove package.

This has been merged into the main ncurses package as of 6.5.

(jperkin)

doc: Updated devel/ncurses to 6.5

(jperkin)

*: Switch packages to combined ncurses package.

CURSES_TYPE=ncursesw is no longer valid, update matches to use ncurses.

(jperkin)

ncurses: Update to 6.5 and overhaul package.

The first half of this commit message describes the pkgsrc changes, the
second half contains the changes in ncurses 6.5 since 6.4.

This combines ncurses and ncursesw into a single package, and ships
combined versions of libncurses and the ncurses/ include directory.
There are no longer separate versions of libncursesw and ncursesw/.

This aligns with what many other operating systems are doing nowadays,
simplifies the package immensely, provides a consistent interface for
other packages to request curses features, improves performance due to
the removal of questionable code in ncursesw's buildlink3.mk, and fixes
bugs - notably the recent python fallout.

Packages that require wide-curses support can use the standardised
"wide" value for USE_CURSES to ensure they are built against the
wide-curses API, while providing the option for NetBSD's builtin curses
to handle that instead, should support be added in the future.

When opting into the wide-curses API, the NCURSES_WIDECHAR define is
used instead of _XOPEN_SOURCE_EXTENDED as the latter cannot be used on
certain operating systems.

A number of other cleanups are included:

- stop using a custom TERMINFODIR on SunOS
- remove unused PLIST_SUBST
- print-PLIST fixes
- use newer make syntax
- pkglint cleanups

This package now conflicts with and supersedes ncursesw-[0-9]*.  Removal
of support for devel/ncursesw will come next.

Release Notes

  These notes are for ncurses 6.5, released April 27, 2024.

  This  release  is  designed  to  be source-compatible with ncurses 5.0
  through  6.4; providing extensions to the application binary interface
  (ABI).  Although  the  source  can  still be configured to support the
  ncurses  5  ABI, the reason for the release is to reflect improvements
  to the ncurses 6 ABI and the supporting utility programs.

  There  are,  of  course,  numerous  other improvements, listed in this
  announcement.

  The  most  important  bug-fixes/improvements  dealt  with  robustness
  issues.  The  release notes also mention some other bug-fixes, but are
  focused  on  new  features and improvements to existing features since
  ncurses 6.4 release.

  Library improvements

    New features

  These are new features:
    * The low-level terminfo and termcap interfaces are used both by the
      higher-level curses library, as well as by many applications.
      The  functions  which  convert  parameterized  terminal capability
      strings  for output to the terminal (tiparm and tparm) analyze the
      capability string to determine which parameters are strings (i.e.,
      addresses), versus numbers (not addresses).
      The  library's analysis of a capability string may differ from the
      calling  application's design if environment variables are used to
      point  to  an  invalid  terminal  database. This is a longstanding
      problem  with  all  implementations  of  terminfo, dating from the
      early 1980s.
      Two  new  functions  address this problem: by providing a function
      which  allows  the  calling  application  to tell ncurses how many
      string-parameters to expect:
          + tiscan_s  helps  applications  check  formatting capabilities
            that would be passed to tiparm_s.
          + tiparm_s provides applications a way to tell ncurses what the
            expected parameters are for a capability.
    * The  ncurses library supports a compile-time feature (enabled with
      the  configure  --enable-check-size  option)  which  simplifies
      initialization  with  terminals  which  do  not  negotiate  window
      (screen)  size.  This is done in setupterm, by providing for using
      ANSI cursor-position report (in user6/user7 terminfo capabilities)
      to  obtain  the  screen  size  if neither environment variables or
      ioctl is used.
      The  ncurses  test-program  with options "-E -T" demonstrates this
      feature.
    * add functions to query tty-flags in SCREEN

  This release drops compatibility with obsolete versions of tack, e.g.,
  pre-1.08

    Other improvements

  These are improvements to existing features:
    * In  addition  to  the  new,  safer function tiparm_s, ncurses adds
      checks to make the older tiparm, tparm and tgoto functions safer:
          + the  terminfo  functions  tiparm  and  tparm  ensure that the
            capability  string  comes from the terminal description which
            ncurses  loads,  rather  than  from  random  data  which  the
            application happens to have.
          + the  tgoto function disallows capabilities which its analysis
            shows will attempt to use string parameters.
          + ncurses  uses  internal functions which correspond to tiparm,
            and  tgoto which ensure that the capability strings which are
            passed  to  these  functions  come  from  the loaded terminal
            description.
    * improve  check  in  lib_tparm.c, ensuring that a char* fits into a
      TPARM_ARG
    * modify  _nc_syserr_abort  to  use _nc_env_access, rather than only
      checking root uid
    * improve thread lock in lib_trace.c
    * modify  flushinp  to  use  file descriptors in SCREEN, rather than
      from TERMINAL, and check if they are for a terminal, like SVr4
    * modify mcprint to use file descriptor in SCREEN, for consistency
    * modify  internal  function  _nc_read_file_entry  to  show relevant
      filename in warnings
    * improve  checks  in  internal  function convert_string for corrupt
      terminfo entry
    * review/improve handling of out-of-memory conditions
    * limit  delays  to 30 seconds, i.e., padding delays in terminfo, as
      well as napms and delay_output functions
    * fix  reallocation  loop  for  vsnprintf  in  _nc_sprintf_string by
      copying the va_list variable
    * modify  delscreen  to  limit  the windows which it creates to just
      those associated with the screen
    * modify  endwin to return an error if it is called again without an
      intervening screen update
    * modify wenclose to handle pads
    * eliminate use of PATH_MAX in lib_trace.c
    * provide for any CCHARW_MAX greater than 1

  These are corrections to existing features:
    * correct loop termination condition in waddnstr and waddnwstr
    * improve  parsing  in  internal  function _nc_msec_cost, allowing a
      single decimal point
    * amend  parameter check for entire string versus specific length in
      winsnstr  and wins_nwstr to match Solaris; make similar correction
      to wins_nwstr
    * correct  internal  function  wadd_wch_literal  when  adding  a
      non-spacing character to a double-width character
    * correct  definition of Charable macro for non-wide ncurses library
      .

  Program improvements

  Several improvements were made to the utility programs. Some were done
  to  make  the  infocmp  option  "-u" option help refactor the terminal
  database.

  infocmp

          + add  limit  checks  for processing extended capabilities with
            the "-u" option
          + correct  initial  alignment of extended capabilities, so that
            the "-u" option can be used for more than two terminal types
          + modify  "-u"  option  to not report cancels for strings which
            were already cancelled in a use'd chunk.
          + correct  an  assignment  "-u"  for  detecting if a boolean is
            unset  in  a base entry and set in a use'd chunk, i.e., if it
            was cancelled.

  tic

          + correct limit-check when dumping tc/use clause via "-I"
          + check  return  value  of  _nc_save_str, in special case where
            extended  capabilities  are  processed  but  the  terminal
            description was not initialized
          + modify  check for multiply defined aliases to report problems
            within  the  current  runtime  rather than for conflicts with
            pre-existing terminal descriptions.
          + disallow  using $TERMINFO or $HOME/.terminfo when "-o" option
            is used

  tput and tset

          + add "-v" option to tput, to show warnings
          + modify reset command to avoid altering clocal if the terminal
            uses a modem
          + modify  reset feature to avoid 1-second sleep if running in a
            pseudo-terminal

    Examples

  Along  with  the  library and utilities, improvements were made to the
  ncurses-examples:
    * modify test_tparm to account for extended capabilities
    * corrected mouse mask in test/testcurs.c
    * modify test/clip_printw.c to optionally test non-wrapped updates
    * modify test/test_mouse.c to use curses api for raw/noraw
    * modify test/clip_printw.c to optionally test non-wrapped updates

  There is one new demo/test programs:

  test/test_endwin.c
          This program shows the return-status from endwin with different
          combinations of endwin (repeated), initscr, newterm.

  Terminal database

  There are several new terminal descriptions:
    * ansi+apparrows
    * contour
    * linux+kbs for terminals which imitate xterm's behavior with Linux
    * rio, rio-direct
    * mostlike
    * ms-vt100-16color, winconsole
    * vt100+noapp,  vt100+noapp+pc,  xterm+app+pc,  xterm+decedit  from
      xterm #389
    * putty+cursor to reflect amending of modified cursor-keys in 2021
    * wezterm

  There  are  many  changes to existing terminal descriptions. Some were
  updates  to  several  descriptions, using the infocmp "-u" option in a
  script  to  determine  which  building-block  entries could be used to
  replace multiple capability settings (and trim redundant information).

  Other changes include:
    * document XF, kxIN and kxOUT
    * add note on sun regarding wscons/cmdtool/shelltool
    * remove DECCOLM+DECSCLM from foot
    * add xterm+focus to foot+base
    * add ecma+strikeout to putty
    * use CSI 3J in vte-2017
    * use oldxterm+sm+1006 in vte-2014
    * modify xgterm to work around line-drawing bug
    * add  xterm  focus  mode 1004 to xterm+focus as fe/fd capabilities,
      like vim.
    * add xterm+focus to alacritty+common
    * add  XR/xr,  to work with vim, and use RV/rv to denote DA2 and its
      response
    * add  XF  flag  to  xterm+focus so that termcap applications can be
      aware of terminals which may support focus in/out
    * use xterm+focus in xterm-p370 and tmux
    * remove xterm+sm+1006 from tmux
    * NetBSD-related fixes for x68k and wsvt25

  Documentation

  As usual, this release
    * improves documentation by describing new features,
    * attempts  to  improve the description of features which users have
      found confusing
    * fills  in overlooked descriptions of features which were described
      in the NEWS file but treated sketchily in manual pages.

  In  addition  to  providing  background  information  to explain these
  features  and  show  how  they  evolved,  there  are  corrections,
  clarifications, etc.:
    * Corrections:
          + add  assignment  in  CF_MAN_PAGES  to  fill  in  value  for
            TERMINFO_DIRS in ncurses, terminfo and tic manpages.
          + clarify  interaction  of  -R  option  versus -C, -I and -r in
            infocmp manpage.
          + correct manpage description of panel_hidden.
          + improve  manpage  description  for addch versus unctrl format
            used for non-printable characters.
          + improve  manpages  discussing  file  descriptors in low-level
            functions.
          + improve description of search rules for terminal descriptions
            in terminfo manpage.
          + modify  dist.mk  to  avoid  passing  developer's  comments in
            manpages into the generated html documentation.
          + modify  test-package  "ncurses6-doc"  to use manpage-aliases,
            which  in  turn  required a change to the configure script to
            factor in the extra-suffix option when deriving alias names.
    * New/improved history and portability sections:
          + add  information  about  "ttycap",  termcap's  forerunner, to
            tset.1
          + document  limitations  of  tparm,  and  error-returns  in
            curs_terminfo.3x
          + document  limitations  of  tgoto,  and  error-returns  in
            curs_termcap.3x
    * Other improvements:
          + This  release  has many changes to improve the formatting and
            style of the manpages.
          + Manpages  now use consistent section-naming, page headers and
            footers (including the modification date for each page).
          + Table layout has been revised.

  There  are  no new manual pages (all of the manual page updates are to
  existing pages).

  Interesting bug-fixes

  The  changes to tparm, tgoto which improve the design of the low-level
  interfaces are interesting, but are not bug-fixes per se.

  Configuration changes

    Major changes

  These are the major changes (aside from introducing tiparm_s):
    * use wide-character (ncursesw) by default
    * use opaque typedefs by default

  However,  most  of  the  work  on configure scripts was done to reduce
  warnings within the configure script:
    * intrusive warnings from GNU grep regarding fgrep and egrep
    * fatal  errors  in  compile-checks,  arising from recent "Modern C"
      efforts  by  some  developers  which caused longstanding configure
      checks to fail.
      After  repairing  the  configure  script,  none  of  that activity
      affected  ncurses  because stricter warnings are used routinely in
      development.

  Other improvements made to configure checks include
    * use  string-hacks in alloc_entry.c, alloc_type.c and hardscroll.c,
      overlooked due to compiler changes in recent OpenBSD releases
    * revise progs.priv.h to provide for NC_ISATTY reuse
    * configure check for MB_LEN_MAX provides warning as needed
    * trim  a  space  after  some  "-R"  options,  fixing  builds  for
      applications built using clang and ncurses on Solaris
    * work  around  misconfiguration  of  MacPorts  gcc13, which exposes
      invalid  definition  of  MB_LEN_MAX  in  gcc's  fallback  copy  of
      limits.h
    * modified  experimental  Windows  driver  works  with  xterm  mouse
      protocol

    Configuration options

  There are a few new configure options:

  --disable-setuid-environ
          Compile  with  environment  restriction, so certain environment
          variables  are  not  available when running via a setuid/setgid
          application. These are (for example $TERMINFO) those that allow
          the  search  path  for  the  terminfo  or  termcap  entry to be
          customized.

          A  setuid/setgid application inherits its environment variables
          from  the current user, in contrast to sudo which may limit the
          environment variables that ncurses uses.

  --enable-check-size
          Compile-in  feature to detect screensize for terminals which do
          not advertise their screensize, e.g., serial terminals.

  --with-abi-altered=NUM
          Override  the  displayed  (rather  than  compiled-in) ABI. Only
          packagers who have created configurations where the ABI differs
          from ncurses should be interested in this option.

  --with-strip-program=XXX
          When  stripping  executables  during install, use the specified
          program  rather  than  "strip" overriding program chosen by the
          install program for stripping executables.

  These configure options are modified:

  --with-pkg-config-libdir[=DIR]
          The  optional  DIR parameter can now be "auto" to automatically
          use pkg-config's library directory.

          The default is $(libdir).

  --with-xterm-kbs[=XXX]
          The  default  is  "auto"  which  tells  the configure script to
          choose BS or DEL according to platform defaults.

  Portability

  Many  of  the  portability  changes  are implemented via the configure
  script:
    * add/use  configure  check  for  clock_gettime,  to  supersede
      gettimeofday.
    * modify  configure script check for pkg-config library directory to
      take  into  account  an  older  version  0.15.0  which  used
      PKG_CONFIG_PATH but not PKG_CONFIG_LIBDIR
    * allow for MinGW32-/64-bit configurations to use _DEFAULT_SOURCE
    * modify  CF_XOPEN_SOURCE  macro's  amend  default  case  to  avoid
      undefining _XOPEN_SOURCE if _POSIX_C_SOURCE is defined
    * updated configure script macro CF_XOPEN_SOURCE, for uClibc-ng
    * modify version-check for gcc/g++, now works for msys2
    * build-fixes related to configure-options and/or platform:
          + fix for --enable-fvisibility
          + fix for unusual values of --with-rel-version
          + fix for unusual values of --with-abi-version
          + fix for --disable-tcap-names
          + fix for termcap in nc_access.h
    * other configure-script improvements:
          + recent msys2 headers work with _DEFAULT_SOURCE; amend check
          + use  $ac_includes_default in most cases where stdlib.h should
            work
          + use #error consistently vs "make an error"
          + add configure macro for gettimeofday vs inline check

  Here are some of the other portability fixes:
    * modify  configure  scripts/makefiles  to  omit  KEY_RESIZE  if the
      corresponding SIGWINCH feature is disabled
    * increase MB_CUR_MAX to 16, matching glibc's MB_LEN_MAX
    * add BSD erase2 to characters handled by tset/reset
    * use getauxval when available, to improve setuid/setgid checks
    * set dwShareMode in calls to CreateConsoleScreenBuffer
    * use  CreateFile  with "CONIN$", "CONOUT$" rather than GetStdHandle
      to  obtain a handle on the actual console, avoiding redirection in
      the MinGW/Win32 configurations
    * modify  MinGW  driver  to  return KEY_BACKSPACE when an unmodified
      VK_BACK virtual key is entered
    * modify  MinGW  configuration  to provide for running in MSYS/MSYS2
      shells, assuming ConPTY support

(jperkin)

ncurses: Stop installing the demo.cc example.

If more than 2 people have used this in the 25 years it's been added to
this package, I'd be very surprised.

(jperkin)

ncurses: Remove faked-up files.

Some of this has already been moved to mk/curses.buildlink3.mk, and the
rest is obsolete now that this file will only be included if we are
specifically using ncurses and not some other implementation pretending
to be ncurses.

(jperkin)

ncurses: Remove USE_NCURSES support.

This was superseded by USE_CURSES many years ago.  If this file is being
included then we are already using ncurses.

This only removes support for the variable, removing it from individual
packages will come later.

(jperkin)

ncurses: Remove USE_TERMINFO.

This is a long-obsolete variable, support was moved to
mk/terminfo.buildlink3.mk many years ago.

(jperkin)

ncurses: Remove INCOMPAT_CURSES.

This was moved to mk/curses.builtin.mk over 10 years ago.

(jperkin)

ncurses: Remove NetBSD builtin hack.

If this file is being included, then the package has specifically
requested ncurses.  This code was committed in 2005 and since then the
pkgsrc curses handling has been revamped significantly.

(jperkin)

ncurses: Remove pointless comment.

(jperkin)

emacs29: Copy file notification fix from emacs28.

(jperkin)

rust: Bump illumos bootstrap kit.

I don't know why this wasn't done with the package update, the bootstrap
kit has been available for ages.

(jperkin)

python312: Remove gettext-lib from buildlink3.mk.

Unbreaks dependencies since gettext-lib was removed from the main package.

(jperkin)

bsdtar: Include limits.h for INT_MAX.

(jperkin)

mutt: Remove unsupported configure argument.

(jperkin)

mutt: Fix native Solaris Kerberos detection.

(jperkin)

doc: Updated security/pinentry to 1.3.0nb1

(jperkin)

pinentry: Needs wide curses.

If linked against ncurses then pinentry-curses box characters are all
wrong.  Bump PKGREVISION.

(jperkin)

crimsonfields: Disable in bulk builds on Darwin.

The build hangs during mkunitset and isn't CPU-bound, so ulimit doesn't
help, requiring external processes to kill it.

(jperkin)

freeciv-client: Disable SDL test during configure.

Hangs the build on macOS waiting for graphical input.

(jperkin)

xclannad: Disable smpeg test during configure.

Hangs the build on macOS waiting for graphical input.

(jperkin)

python312: Clean up previous change.

Now pkglint clean, no functional change.

(jperkin)

python311: Clean up recent readline addition.

Now pkglint clean, no functional change.

(jperkin)

qt6-qtbase: Adjust logic for PLIST.egldevice.

Avoids errors on systems that do not have the file in question, which will be
the majority of bulk build hosts.

(jperkin)

py-salt-pepper: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-acme-tiny: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-pikepdf: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

emacs29: Disable libgccjit option on SunOS too.

It would probably be better for someone who understands the benefit of this
option to invert the logic as there are going to be a huge number of platforms
that can't support this.

(jperkin)

mk: Remove Darwin aarch64 cross-compile flag hack.

This isn't required any more.  When building inside a macOS chroot the
arch of mksh determines whether Rosetta is in use, and the compiler will
just generate code for that arch by default.

(jperkin)

bootstrap: Use printf builtin in bootstrap mksh.

This mirrors the build of the main mksh binary, and improves bootstrap
performance by around 12% on my M1 Max.

(jperkin)

mk: Correctly engineer previous change.

Rather than breaking many of the 23 platforms that pkgsrc purports to support,
move the EXTRACT_OPTS_TAR logic into the individual package that needed it.
If there are other packages that need this they can easily copy the required
two lines.

It's worth noting here that only bsdtar has a problem with this, gtar extracts
the file fine (with warnings).

(jperkin)

tegaki-*: Remove DEPENDS on tegaki-pygtk.

Package was removed 11 days ago.

(jperkin)

ffmpeg7: Fix COMMENT.

(jperkin)

forked-daapd: Exclude libinotify on SunOS

(jperkin)

messagelib: Exclude libinotify on SunOS

(jperkin)

akonadi: Add socket libraries on SunOS.

(jperkin)

openblas: Do not apply broken -Y handling on SunOS.

(jperkin)

openblas: Remove ONLY_FOR_PLATFORM.

Being untested is no reason to exclude 20 platforms.

(jperkin)

trn: Add NOT_FOR_BULK_PLATFORM for Darwin.

For some reason this software tries to locate libc.  On at least modern
versions of macOS this fails and so it runs an interactive program to
ask the user where it lives.  This causes infinite loops in bulk builds.

(jperkin)

cyrus-sasl: Remove some outdated text from MESSAGE.

Some users found this text confusing and ended up trying to put config
files in the very directory we were telling them not to, e.g.
https://github.com/TritonDataCenter/pkgsrc/issues/378.

This text was added almost 7 years ago, so we can be pretty confident that
users have migrated past this change by now.

(jperkin)

gcc*: Add NOT_FOR_PLATFORM for Darwin/aarch64.

None of these releases support that platform, at least not without a large
patch like the one we currently use for lang/gcc12 (to provide gfortran).

(jperkin)

guile30: Wrap dprintf() in HAVE_DPRINTF.

(jperkin)

doc: Updated pkgtools/bootstrap-mk-files to 20240422

(jperkin)

bootstrap-mk-files: Update to 20240422.

Remove OS variable from remaining sys.mk files.  This was never defined on a
few operating systems (including NetBSD), is not used by pkgsrc, and wasted a
fork of a shell and uname(1) in every bmake call on many systems.

(jperkin)

rust: Revert part of last update to fix lib names.

(jperkin)

doc: Updated editors/emacs29-nox11 to 29.3nb1

(jperkin)

emacs29-nox11: Enable module support.

Discussed on pkgsrc-users a while back.  Negligible (around 24KB) increase
in package size, no change in dependencies.  Bump PKGREVISION.

(jperkin)

openssl: Use consistent socket interfaces on SunOS.

The OpenSSL codebase is a forest of preprocessor usage so it's hard to say
exactly what's going wrong, but there's definitely inconsistent use across
the codebase, resulting in socket failures when trying to use QUIC.

Ensure recvmsg/CMSG_* interfaces are available, fixing test suite as well
as some real world regressions (e.g. devel/py-poetry) noticed by users.

(jperkin)

go122: Support O_DIRECT on illumos.

(jperkin)

go121: Support O_DIRECT on illumos.

(jperkin)

notmuch: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-lieer: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-semver: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

go122: Implement syscall.Mkfifo() on illumos.

(jperkin)

go-bin: pkglint autofix.

(jperkin)

go-bin: Update the illumos bootstrap kit.

(jperkin)

go121: Implement syscall.Mkfifo() on illumos.

(jperkin)

espeak-ng: Add correct TEST_TARGET.

(jperkin)

espeak-ng: Portability fixes.

(jperkin)

cargo-c: Use newer os_info crate.

Contains build fixes for illumos.

(jperkin)

nbsed: Add __dead compat.  Fixes build on SmartOS.

Test suite looks clean compared to previous nbsed and native sed.

(jperkin)

nbsed: Add __UNCONST compat.

This now builds on macOS, still failing on other OS.  Running the test suite
appears to show no differences, at least compared to macOS 14.3.1's sed.

(jperkin)

nbsed: Remove REG_GNU code.

This is a feature only available on newer NetBSD.  Neither -G nor -g are
supported by any other sed's that I can find, so just completely remove the
options, as we only need to be compatible.  The libnbcompat regcomp is very
old, and so trying to merge in REG_GNU support would also be very difficult.

(jperkin)

libuuid: Add /usr/include/uuid to the fake .pc.

Fixes software that only looks for <uuid.h>.

(jperkin)

mbrola: Support x86_64, not MAKE_JOBS safe, fix perms.

(jperkin)

qgpgme: Avoid "single" conflict.

(jperkin)

xscreensaver: Avoid "sun" conflict.

(jperkin)

xymon: Further fixes to use system c-ares.

(jperkin)

qalculate: Avoid sun define.

(jperkin)

opencv: Avoid ambiguous function call.

(jperkin)

capstone: Handle illumos getopt.

(jperkin)

gst-plugins1-base: Include sys/stat.h for S_*.

(jperkin)

taglib: Avoid offset_t conflict on illumos.

(jperkin)

doc: Updated pkgtools/pkg_install to 20240307

(jperkin)

pkg_install: Bump version to 20240307.

(jperkin)

pkg_install: Respect quiet flag for rebuild commands.

While here remove a bunch of trailing whitespace.

(jperkin)

pkg_install: Add sanity checks for pkgdb correctness.

A number of users have hit "Conflicting PLIST" errors in the past that are
almost certainly caused by duplicate PKGBASE entries in the pkgdb, most likely
as a result of unpacking a bootstrap kit over the top of an existing install.

This adds a check to ensure there are no duplicate PKGBASE entries, displaying
an error message with the list of duplicates if they are found, and is called
for the "check", "rebuild", and "rebuild-tree" commands.

(jperkin)

net-snmp: Support long fds_bits.

(jperkin)

R: Fix building extensions with C++.

(jperkin)

php-mrbs: Fix DEPENDS.

(jperkin)

bulk-test-php: Catch up with wordpress PKGNAME change.

(jperkin)

py-numpy: Fix build on SunOS.

(jperkin)

kitemviews: Avoid ambiguous function call.

(jperkin)

bibtexu: Fix PKGMANDIR.

(jperkin)

mk/fetch: Remove -O from curl's FETCH_AFTER_ARGS.

Perhaps this was required in some ancient version of curl, but in any modern
version having both "-o filename" and "-O" (which means to use the remote
filename) only results in "Warning: Got more output options than URLs" messages
for every single download.

(jperkin)

doc: Updated security/mozilla-rootcerts-openssl to 2.14

(jperkin)

doc: Updated security/mozilla-rootcerts to 1.0.20240214

(jperkin)

mozilla-rootcerts*: Update to 20240214 data.

(jperkin)

grafana: Restore illumos support.

(jperkin)

ipmitool: Use standard autoreconf.

The supplied bootstrap script is broken.  Fixes build on Darwin.

(jperkin)

py-matrix-nio: Not for python38.

(jperkin)

rspamd: Build fix (_L conflict with ctype).

(jperkin)

po4a: Various build and test fixes.

(jperkin)

kde4: kfloppy was deleted, remove DEPENDS.

Fixes bulk builds.

(jperkin)

*: PostgreSQL 11 is no more.

Fixes scan failures.

(jperkin)

py-keyrings.alt: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-phonenumbers: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-waitress: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

festvox*: Relax DEPENDS matches.

There's no need to be absolutely specific about version requirements, it makes
certain operations slower, and has been broken since the mbrola updates last
week.

(jperkin)

rust: Update illumos pkgsrc bootstrap libraries.

(jperkin)

p5-Net-CUPS: Re-order includes.

Fixes indirect library dependencies as well as ensuring they are sorted.

(jperkin)

freeradius: Remove DEPMETHOD=build for dependencies.

The modules already have a full dependency on freeradius, which in turn has
a full dependency on all of these dependencies, so this does nothing but
cause indirect library dependency issues.

(jperkin)

xsc: Fix indirect library depends.

(jperkin)

qbrew: Fix indirect library dependencies.

(jperkin)

eterm: Remove bogus BUILDLINK_DEPMETHOD.

While here remove pointless bsd.prefs.mk include, and sort includes.

(jperkin)

py-scikit-image: Ensure python is properly buildlinked.

(jperkin)

py-macs3: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

mozjs78: Add direct include of iconv.

Previously it was coming in via the clang build dependency, and thus falling
foul of indirect library checks.

(jperkin)

udunits: Add expat to bl3.

Fixes indirect library checks in various dependencies.

(jperkin)

*: Remove BUILDLINK_DEPMETHOD.libXt=build

Fixes indirect library depends.

(jperkin)

*: Remove BUILDLINK_DEPMETHOD.qt5-qttools=build.

This appears to just be cargo-culted around various packages.  Fixes indirect
library dependencies.

(jperkin)

py-hmmlearn: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

libcxx: GCC 13 build fixes.

(jperkin)

libcxxabi: Backport GCC 13 build fix.

(jperkin)

p5-Config-MVP: Remove syntax error and pkglint.

(jperkin)

link-grammar: SunOS build fixes.

(jperkin)

wxGTK32: SunOS and non-wayland fixes.

(jperkin)

alsa-lib: Build fixes.

(jperkin)

gmime3: libgpg-error is not a build-only dependency.

Fixes indirect library checks.

(jperkin)

protobuf-c: Restore section removed in last update.

Without this the build fails with "google/protobuf/descriptor.proto: File not
found.".  It's unclear how this could work for some and not others.

(jperkin)

ocaml-dune: SunOS build fix.

(jperkin)

c-icap: Export pcre option correctly.

(jperkin)

py-matplotlib: Add missing part of previous fix.

(jperkin)

xindy: Work around clisp ignoring buildlink.

(jperkin)

ups-nut: Fix PLIST.SunOS.

(jperkin)

libunwind: Use correct flags to link.

The configure script only tests if CXX supports various flags, but then uses
CC to compile and link various objects.  While compiling only issues warnings
about incorrect flags, attempting to link with the wrong front end causes
fatal errors.

Specifically fixes trying to link with gcc using "-nostdlib++".

(jperkin)

mautrix-telegram: Fix impossible py-mautrix constraint.

(jperkin)

ghc96: Build with LC_ALL=en_US.UTF-8.

There appear to be a number of UTF-8 issues in ghc, some of which are patched
in this package and the resulting binaries, but there are others that may be
caused by the bootstrap kit.  Package now builds correctly on SunOS.

(jperkin)

ghc96: Remove more SunOS hacks.

(jperkin)

ghc96: Switch to updated SunOS bootstrap kit.

Generated using the same version as others, and modified to use self-contained
libiconv and ncurses libraries.

(jperkin)

postgresql16: Requires mit-krb5 1.11 for gss_store_cred_into.

(jperkin)

py-matplotlib: Enforce PYTHON_FOR_BUILD_ONLY=no.

Required as py-pybind11's bl3 comes first, and it defaults to ?=yes.

(jperkin)

mk: Support Arch Linux lib64 directories.

Reported by zagreus on IRC.

(jperkin)

harfbuzz: Put glib2 bl3 include back where it belongs.

The ordering hack is no longer required now that gtk-doc's buildlink3.mk
no longer includes glib2.

(jperkin)

gtk-doc: Remove glib2 from buildlink3.mk.

It appears to no longer be needed (added in 2015) and causes indirect
dependency problems due to this file defaulting to DEPMETHOD of "build".

(jperkin)

python: Exclude dependencies for build.

When python bl3 files are included with PYTHON_FOR_BUILD_ONLY=yes, their
DEPMETHOD is set to "build", in which case we do not want to include
indirect dependencies as they will then be tagged as "build" also.  Fixes
potential runtime issues exposed by indirect dependency checks.

(jperkin)

gobject-introspection: Exclude dependencies for build.

This package is often included with its DEPMETHOD set to "build", in which
case we do not want to include its indirect dependencies as they will then
be tagged as "build" also.  Fixes potential runtime issues exposed by
indirect dependency checks.

(jperkin)

py-pybind11: Default to python for build only.

This buildlink3.mk has a default DEPMETHOD of "build", so this helps avoid
unwanted indirect buildlink3 dependencies from python.

(jperkin)

py-aioresponses: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

xsel: Remove libXt=build depmethod.

This package relies on libXt pulling in other X11 libraries, and it definitely
requires them at runtime.  Exposed by indirect dependency checks.

(jperkin)

vncviewer: Remove unused libXt.  Fix SunOS build.

libXt is no longer a build-only dependency and this fixed indirect depends
checks.

(jperkin)

gmime: Remove bogus libgpg-error=build depmethod.

This is certainly not a build-only dependency, and is already included by
at least two different full dependencies.  Fixes issue exposed by indirect
dependency checks.

(jperkin)

botan1: Use python tool.mk.

Avoids python buildlink3 problems with indirect dependencies.

(jperkin)

compiz: Remove bogus BUILDLINK_DEPMETHOD.pango=build.

pango is already included at least twice by other full dependencies so it's
never going to be build-only, and it causes failures with indirect dependency
checks.

(jperkin)

gmpc: Remove default BUILDLINK_DEPMETHOD.gmpc=build.

It doesn't make any sense, plugins are useless without the main package, and
it causes indirect depends issues.

(jperkin)

doc: Updated wm/windowmaker to 0.96.0nb4

(jperkin)

windowmaker: Remove bogus BUILDLINK_DEPMETHOD.libXt=build.

It is definitely not a build-only dependency.  Exposed by indirect depends
checks.  Bump PKGREVISION.

(jperkin)

doc: Updated graphics/freetype-utils to 1.5nb2

(jperkin)

freetype-utils: Remove bogus BUILDLINK_DEPMETHOD.libXt=build.

It is definitely not a build-only dependency.  Exposed by indirect depends
checks.  Bump PKGREVISION.

(jperkin)

doc: Note py-gobject* revbumps.

(jperkin)

py-gobject*: Change buildlink3 to match requirements.

py-gobject is the package that needs a dependency on glib2, py-gobject-shared
only needs it via gobject-introspection to build the files that it installs.

Fixes issue with indirect depends, tested with py-gtk2.  Bump PKGREVISIONs.

(jperkin)

gcc12: Stop subverting buildlink with LDFLAGS.

It's not clear why this was ever added in the first place (back in gcc46),
but it results in the build gcc finding libraries in PREFIX that aren't
buildlinked.

Discovered on macOS with the indirect DEPENDS checks where iconv was being
pulled in directly from PREFIX.  While here fix a non-ASCII space.

(jperkin)

check-shlibs: Include DSO name in failure output.

Clarify the use of "DSO", previously it referred to both the original object
that was being checked, as well as any libraries it depended upon, which was
just confusing.  While here reduce some of the diffs between the two versions.
At some point they should really just be merged together as there is a huge
amount of duplication.

(jperkin)

*: Simplify vala buildlink3.

The default DEPMETHOD for vala's bl3 is "build", so there's no point
duplicating that in each package.  Given that it is only ever a build
dependency, completely remove the indirect dependencies as they should not
be made available.  Also remove obsolete BUILDLINK_API_DEPENDS settings.

Tested with a few of the affected packages.

(jperkin)

vala: Limit buildlink3 includes to DEPMETHOD=full.

This file is generally only included when the valac and vapigen tools are
required, and it even defaults to DEPMETHOD=build.  The library dependencies
are not required for running the tools, and this avoids a number of implicit
dependency issues.

(jperkin)

p11-kit: Requires msgfmt.

(jperkin)

boost-libs: Pull in xz for lzma support in iostreams.

This could have been found by accident previously, so enable explicitly.

(jperkin)

nodejs*: Consolidate and fix python includes.

Ensures that PYTHON_FOR_BUILD_ONLY is set prior to including pyversion.mk,
and ensures python and its dependencies are not buildlinked.

(jperkin)

py-aiostream: Unbreak impossible DEPENDS.

(jperkin)

p11-kit: SunOS needs -D__EXTENSIONS__.

(jperkin)

python*: Remove -luuid hack on SunOS.

This ends up leaking into the shipped python-config, which wouldn't normally
be a problem, but broken build systems such as waf end up linking against
libraries that are not buildlinked, resulting in missing libuuid references.

If this is still required for builtin libuuid support then that will need to
be done in a different way that doesn't end up in the exported libraries.

(jperkin)

gettext-tools: Add termcap to the buildlink includes.

This is required for dependencies that use libtextstyle.

(jperkin)

harfbuzz: Ensure glib2 is included before gtk-doc.

This works around an issue with detecting whether all required libraries are
runtime dependencies.  gtk-doc is a build-only dependency, which causes bl3
to tag glib2 and all of its dependencies as build too.  Later inclusion of
glib2 marks it as a full dependency, but include guards mean all of its
dependencies are not.  Including glib2 first is a hack, but does resolve the
problem for now.

(jperkin)

fontconfig: Remove PYTHON_FOR_BUILD_ONLY=yes.

This should really have been "tool" all along, and now that tool.mk defaults to
that we can remove it completely.

(jperkin)

llvm: Switch PYTHON_FOR_BUILD_ONLY to "tool".

Avoids problems where python is buildlinked, and buildlink dependencies of
python ending up being available during the build but not registered as full
dependencies, resulting in them potentially being unavailable at runtime.

(jperkin)

python: Default tool.mk to PYTHON_FOR_BUILD_ONLY=tool.

Helps to avoid buildlinking unnecessarily which can lead to dependency
issues with packages that are pulled in by python's buildlink3.mk.

(jperkin)

cairo: SunOS build fixes.

(jperkin)

nodejs: Switch PYTHON_FOR_BUILD_ONLY to "tool".

Avoids problems where python is buildlinked, and buildlink dependencies of
python ending up being available during the build but not registered as full
dependencies, resulting in them potentially being unavailable at runtime.

(jperkin)

py-ldapdomaindump: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-tinydb: Fix PYTHON_VERSIONS_INCOMPATIBLE.

(jperkin)

py-gobject32: Fix py-cairo include.

(jperkin)

go: Add support for GO_VERSION_REQD.

This isn't full multi support like other languages yet, but for now simply a
way for a package to indicate that it requires a specific version of go.

Useful for things like wireguard-go which currently do not build with 1.21.

(jperkin)