Initial import

1 files changed, 471 insertions, 0 deletions

diff --git a/PKGBUILD b/PKGBUILD
new file mode 100644
index 000000000000..c6cee8c5ba78
--- /dev/null
+++ b/PKGBUILD
@@ -0,0 +1,471 @@
+# Maintainer:  Chris Severance aur.severach aATt spamgourmet dott com
+# Contributor: Christian Neukirchen <chneukirchen@gmail.com>
+# Contributor: Dave Reisner <d@falconindy.com>
+# Contributor: Mark Taylor <skymt0@gmail.com>
+# Contributor: Stefan Husmann <stefan-husmann@t-online.de>
+# Contributor: xduugu
+# Category: system
+
+# Recommended build command: makepkg -scCfi
+
+# TODO: An option to install more of the ast toolchain. Which parts?
+# TODO: /etc/skel as found in the RPM
+# TODO: Do we need to include BSD licenses? Probably not.
+# TODO: Verify this builds on i686
+# TOFIX: AT&T et al need to add a copyright to the end of the ksh man page similar to the one on nmake.
+# TOFIX: AT&T et al need to fix <YEAR>... in the BSD copyright.
+# TOFIX: AT&T et al need to make fixes for the PKGBUILD patches.
+# TOFIX: AT&T et al need to fix the build errors that happen after ksh.
+# TOFIX: test wouldn't complete after 16 hours. This might be related to the
+#        build errors after ksh.
+# TOFIX: If ksh is to be a serious contender in the interactive shell game, ksh needs mc integration
+
+# To build you must view and accept the EPL license agreement then insert the
+# username and password from the EPL license page below. Using AT&T sources
+# is necessary because the Debian sources won't build any more. The new sources 
+# are hard enough to build.
+
+# http://www.kornshell.com/
+#   _Software_
+#   _The Official AT&T Release of KornShell 93_
+#   ... _EPL license page_
+# http://www2.research.att.com/~astopen/download/licenses/epl-1.0.html
+
+###############################################################################
+# Please visit the EPL License page for the username and password
+_opt_EPLUSER=''
+_opt_EPLPASS=''
+_opt_EPLURL='http://www.research.att.com/sw/download'
+_opt_InstallBETA=1  # Default: 1, 0 to build release version
+_opt_InstallNMAKE=0 # Default: 0, 1 to Install nmake
+###############################################################################
+
+# This is a multi core build. It takes less than 3 minutes on a big processor 
+# like my Intel Haswell E3-1245v3 with SAS drives! Watch it go in htop!
+# My Celeron G530 with SATA drives at home wasn't even close!
+
+# Use Debian-provided sources since original AT&T tarball requires
+# license agreements.
+#_debpkgver='93u+20120801'
+#source=("http://ftp.de.debian.org/debian/pool/main/k/ksh/ksh_${_debpkgver}.orig.tar.gz" 'LICENSE')
+
+# The Debian package is INIT.2012-08-01.tgz+ast-ksh.2012-08-01.tgz
+# We build the latest INIT + ast-base
+
+# If Arch distributed this as a binary then building from the Debian sources 
+# would be fine. Build it once then wait until the next version. This is in 
+# the Arch AUR so it must build any time someone wants it. The progression 
+# of Linux soon renders static sources in the complex AT&T build environment 
+# unable to build.
+
+# To build we had to mix the beta INIT and the release or beta ast-base which
+# neither the Debian source nor the AT&T downloader is capable of.
+
+# The patches herein are simple clerical errors resulting from a complex build 
+# environment failing to support Linux as well as we'd like. Others haven't
+# found the errors becase they build ast-ksh instead of ast-base.
+
+# An advanced PKGBUILD that uses (and abuses) all the features of the AT&T
+# build system is the best way to keep up with the ever changing build 
+# requirements. Building ast-base may also attract the attention of the AT&T 
+# devs so that Linux support improves.
+
+# Gentoo also builds on demand so if you have build problems, check for new
+# tricks in their eBuilds.
+
+set -u
+pkgname='ksh'
+pkgdesc="AT&T's Korn shell ksh93 nmake from ast-base"
+arch=('i686' 'x86_64')
+url='http://www.kornshell.com/'
+license=('EPL' 'CPL') # When a new release replaces the 2012 version, CPL can be dropped. http://www2.research.att.com/~astopen/download/licenses/licenses.html
+depends=('glibc')
+makedepends=('binutils') # for ar to unpack debian package
+conflicts=('pdksh') # I'm not sure why this is a conflict! There are many ksh clones. Why aren't they all in conflict?
+install='ksh.install'
+
+# URL encoding, the sleazy way for the few characters we need!
+_opt_EPLUSERE="${_opt_EPLUSER// /%20}"
+_opt_EPLUSERE="${_opt_EPLUSERE//\//%2F}"
+_debfile='ksh_93u+20120801-1_amd64.deb' # We don't want the binaries so the arch doesn't matter.
+
+if [ "${_opt_InstallBETA}" -ne 0 ]; then
+pkgver='2014.06.25beta'
+pkgrel='1'
+source=("http://${_opt_EPLUSERE}:${_opt_EPLPASS}@www2.research.att.com/~astopen/download/beta/INIT.2014-12-24.tgz"
+        "http://${_opt_EPLUSERE}:${_opt_EPLPASS}@www2.research.att.com/~astopen/download/beta/ast-base.2014-06-25.tgz"
+        "http://http.us.debian.org/debian/pool/main/k/ksh/${_debfile}" # man page and misc files
+)
+# :; is to keep updpkgsums from removing all our sums
+:;sha256sums=('8852b9d37d5034e3780aeb5f963726381eeb4e08bb5bee1fbfa7e3f529c10e1b'
+            '58588b07b076f05dbbd5f4f095d5753309a8356ba1e5475262ce77d6bff42dae'
+            '37495cc625a2174b22a43542acac1d69402ee4992ee084a84690546c5b932b39')
+else
+pkgver='2012.08.01'
+pkgrel='5'
+#source=("http://${_opt_EPLUSERE}:${_opt_EPLPASS}@www2.research.att.com/~astopen/download/tgz/INIT.2013-05-24.tgz"
+source=("http://${_opt_EPLUSERE}:${_opt_EPLPASS}@www2.research.att.com/~astopen/download/beta/INIT.2014-12-24.tgz"
+        "http://${_opt_EPLUSERE}:${_opt_EPLPASS}@www2.research.att.com/~astopen/download/tgz/ast-base.2012-08-01.tgz"
+        "http://http.us.debian.org/debian/pool/main/k/ksh/${_debfile}" # man page and misc files
+)
+:;sha256sums=('8852b9d37d5034e3780aeb5f963726381eeb4e08bb5bee1fbfa7e3f529c10e1b'
+            '3bd668dbd922790f1d24c78e3522ef4816501e9cf80abf3c480f554c74f4dbac'
+            '37495cc625a2174b22a43542acac1d69402ee4992ee084a84690546c5b932b39')
+fi
+# makepkg unpacking deb files into the src folder isn't particularly useful.
+noextract=("${_debfile}")
+
+# Build Notes.
+# This AT&T package builds itself up from the few crude tools supplied in INIT
+# and the sources in ast-base. The general progression is
+# INIT bin/package -> iffe + mamake + early tools and libs -> nmake
+# -> late tools & libs -> ksh -> test
+# Each step builds the tools needed to progress to the next step. 
+
+# Their general build technique is to make modified copies of select sources 
+# in src to arch. gcc -I- is used to sub in the modified copies. Each 
+# target like ksh can use the same src files by making different changes 
+# to them then using gcc -I- to sub in the changes. With -I- deprecated
+# in 2005 and yet in 2015 it still hasn't been removed gives you an idea of
+# how much work it is to fix the build environment to use -iquote, to say
+# nothing of the platforms where -I- is still available. Rather than move 
+# to -iquote it seems that AT&T et al are just hacking the build so select
+# sources compile with or without -I-.
+
+# sed hacking sources in arch/ generally doesn't work because they don't exist
+# before the make and they are generated at the time they are used. I had to 
+# use chattr +i to lock arch/ sources to test changes for effectiveness then 
+# figure out how to make those changes over in the src files that are 
+# available before the build starts and not modified during the build.
+
+# Had I not found ways to do that I'd need to let the make crash, fix the
+# file in arch, spirit away the src file so it can't be generated again, 
+# then restart the make.
+
+# Hopefully if I document some things future maintainers won't need to spend
+# so much time learning the AT&T build system to fix builds.
+
+# The release ast-base will build with the beta INIT. This PKGBUILD can do that
+# but you won't be able to trick bin/package into making this combination. 
+# From this success I can tell:
+
+# INIT adapts to the ever advancing environments, including Linux.
+# ast-base moves forwards as they improve the code.
+
+# Release file sizes: 2012.08.01
+# 1543096 2015-06-06 03:30 usr/bin/shcomp
+# 1547192 2015-06-06 03:30 usr/bin/ksh
+
+# Beta file sizes: 2014.06.25beta
+# 2012304 2015-06-06 03:35 usr/bin/shcomp
+# 2008208 2015-06-06 03:35 usr/bin/ksh
+
+# I didn't see enough source diffs to justify a 4th larger binary. Maybe it's
+# debug code. Announcements on the ast lists indicate that the improvements
+# are pretty substantial in the direction of keeping ksh in the game as an
+# interactive shell. This seems useful mainly in environments that use ksh
+# as their primary shell as the future is grim for any shell wanting to
+# unseat zsh as the best interactive shell.
+
+# -CJS
+
+_pkginit() {
+  if [ "${SOURCEONLY:-0}" -ne 0 ]; then # see makepkg -S if this var changes
+    if [ ! -z "${_opt_EPLPASS}" -o ! -z "${_opt_EPLUSER}" ]; then
+      echo '_opt_EPLUSER and _opt_EPLPASS must be blank' 1>&2
+      echo 'to produce a source package.' 1>&2
+      exit 1
+    fi
+  else
+    if [ -z "${_opt_EPLPASS}" -o -z "${_opt_EPLUSER}" ]; then
+      echo '_opt_EPLUSER and _opt_EPLPASS must be filled in.' 1>&2
+      echo 'Please retrieve these from the EPL License page.' 1>&2
+      echo 'Edit the PKGBUILD for instructions and changes.' 1>&2
+      echo 'This PKGBUILD must be edited to install so 1 step' 1>&2
+      echo "packages like packer can't be used." 1>&2
+      if [ "${GENINTEG:-0}" -eq 0 ]; then # see makepkg -g if this var changes
+        exit 1
+      fi
+    fi
+  fi
+}
+declare -f srcinfo_write_attr > /dev/null || _pkginit
+
+prepare() {
+  set -u
+
+  # makepkg just dumps the deb file directly into the src folder. This wouldn't
+  # work if we needed 2 DEB files. We unpack what we want into a sane
+  # folder name.
+  # Unpack Debian file. No binaries needed, mostly package QC and the ready
+  # made shcomp man pages.
+  mkdir -p "${srcdir}/debian"
+  cd "${srcdir}/debian"
+  # http://www.g-loaded.eu/2008/01/28/how-to-extract-rpm-or-deb-packages/
+  ar p "../${_debfile}" 'data.tar.gz' | tar zxf -
+  set +u
+}
+
+# I haven't found any help on package parameters other than here:
+# bin/package help 2>&1 | less
+
+build() {
+  set -u
+  cd "${srcdir}"
+
+  # Interesting, but too old to be of much help.
+  # http://www.linuxfromscratch.org/blfs/view/6.3/postlfs/ksh.html
+
+  # These were the least unhelpful instructions. This full featured PKGBUILD
+  # can now serve as the complete build and install instructions for ksh.
+  # http://www2.research.att.com/~astopen/download/gen/SOURCE.html
+
+  # Make package authorize/read use our persistent downloads rather than 
+  # downloading again.
+  #mkdir -p 'lib/package/tgz'
+  #cp -pn *.tgz 'lib/package/tgz/'
+  # This unpacks the packages we just placed
+  # sh bin/package read
+
+  # These are here to show alternate ways to get the packages. INIT is 
+  # automatically downloaded as a dependency. We don't do this because:
+  # * The PKGBUILD downloads persist when we wipe the src folder.
+  # * We can guarantee that the package name always matches the content.
+  # * We can mix the beta INIT with the release ast-base.
+  if [ "${_opt_InstallBETA}" -ne 0 ]; then
+    : # sh bin/package authorize "${_opt_EPLUSER}" password "${_opt_EPLPASS}" quiet setup beta "${_opt_EPLURL}" ast-base
+  else
+    # This likely worked in 2013 but on Arch as of June 2015 it is no longer 
+    # feasible to get any release to compile with INIT.2013-05-24.
+    : # sh bin/package authorize "${_opt_EPLUSER}" password "${_opt_EPLPASS}" quiet setup source "${_opt_EPLURL}" ast-base
+  fi
+
+  # Arch typically does not have a /usr/tmp. My system happens to because of 
+  # another bum package. Gentoo also patched this.
+  # The grep nonsense is so we don't needlessly change the timestamp of 
+  # already fixed files. Helpful for running makepkg -ef over and over.
+  sed -i -e 's:/usr/tmp:/tmp:g' 2>/dev/null `grep -l '/usr/tmp' 'src/cmd/nmake/Makerules.mk' 'src/cmd/tw/updatedb.sh' 'src/lib/libast/features/stdio' 'src/lib/libast/features/mmap' 'src/lib/libast/path/pathtemp.c'` || :
+
+  # fix L_tmpname, sizeof("/tmp")+15 = 20 (not 19) but we'll use 25 like the 
+  # rest of the code.
+  # This doesn't work. Despire being in src, something in make overwrites this file.
+  # (This was early in the build attempts before I learned how iffe works.
+  #  This claim may be wrong. Other iffe/FEATURE files are source files and 
+  #  cannot be modified by make so this one shouldn't be modified either.)
+  #if ! grep '^\t<<"#define L_tmpnam">> L_tmpnam$' 'src/lib/libast/features/stdio'; then
+  #  sed -i -e 's:^\(\t<<"#define L_tmpnam\)\(">> L_tmpnam\)$:\1 (25)\2:g' 'src/lib/libast/features/stdio'
+  #fi
+  # Until that's figured out, I'll just enable the defaults in the .c files.
+  sed -i -e 's:^#ifndef L_tmpnam$:#undef L_tmpnam\n &:g' \
+    'src/lib/libast/comp/tmpnam.c' 'src/lib/libast/path/pathtmp.c'
+
+  # These patches are worthless. They were a desperate attempt to make 
+  # INIT.2013-05-24 build ast-base.2012-08-01. All I had to do was use the
+  # beta INIT and the release ast-base builds fine. No sense making bad changes
+  # to the ast-base source code when the problems are in INIT and the AT&T devs
+  # have already fixed them. This is left here as a reminder that for code
+  # that formerly compiled, most build problems are from Linux evolution 
+  # breaking INIT and not that unchanged formerly working code has somehow
+  # broken itself.
+  if ! : && [ "${_opt_InstallBETA}" -eq 0 ]; then
+    # These patches are very bad. Unfortunately the nmake files are so terse
+    # that brute force is the only way I can find to get this stuff to work.
+
+    # Library -ldbm seems to have been replaced with -ldb + -lm, fixed in BETA
+    sed -i -e 's: -ldbm: -ldb -lm:g' `grep -l ' -ldbm' 'src/cmd/coshell/Makefile' 'src/cmd/cs/Makefile'` || :
+
+    # The -l switches on the :LIBRARY: line are being ignored.
+    sed -i -e 's|^SHCOMP := shcomp.*$|CCFLAGS = -lm\n &|g' 'src/cmd/ksh93/Makefile'
+
+    # The -l switches on the :LIBRARY: line are being ignored.
+    #sed -i -e 's:^CCFLAGS = \(.*\)$:CCFLAGS= \1 -ldl -ldld:g' 'src/lib/libdll/Makefile'
+  fi
+
+  # Due to the gcc -I- -iquote problem this appears to include the non
+  # prototyped expr.h in src rather than the prototyped version in arch.
+  # On Arch Linux it seems to work. The key feature of -I- is that it removes 
+  # "." as a search folder from #include "expr.h". The build makes a modified 
+  # eval.h in the arch/ folder and uses '-Iarch -I-' to force the arch folder 
+  # in front of "." so that one is taken even though there's a eval.h right 
+  # next to sources.c.
+  # This key functionality is not at all duplicated by -iquote so eliminating 
+  # -I- hasn't been done. This technique is used so that each target in the 
+  # product group ast was extracted from can make custom mods to standard tools 
+  # without changing the sources. 
+  # The only technique that works reasonably well is to use #include <eval.h> 
+  # but that requires changing the sources, something you can't really do on a 
+  # multi core build or when the source files are on a network share and many 
+  # tool chains are being built at once. Another way is to make a proto eval.h 
+  # with an alternate name that is never available but that forces every target 
+  # to make custom mods. Smaller targets may not want the extra burden.
+
+  # Once you see how -I- is being used you get an idea of how much damage gcc
+  # has done to the large tool chains by deprecating -I-.
+  # http://stackoverflow.com/questions/11605515/how-do-i-prevent-a-quoted-include-from-searching-the-directory-of-the-current-so
+  # https://gcc.gnu.org/ml/gcc-help/2005-04/msg00341.html
+
+  # Looks like someone changed a #define to a typedef, added another #define,
+  # but forgot to fix the defined() for the include upline.
+  sed -i -e 's:defined(YYSTYPE):defined(YYSTYPE_IS_DECLARED):g' \
+    'src/lib/libexpr/expr.h'
+
+  # The simplicity of these patches do not adequately reflect the difficulty 
+  # I had in finding them.
+
+  # This is from the Gentoo ebuild.
+  # sh bin/package flat only make ast-ksh SHELL=sh SHOPT_SYSRC=1
+  # See src/cmd/ksh93/Makefile for brief notes on options like SHOPT_SYSRC
+
+  # Remove quiet to see the cc command lines.
+  echo '>>>Starting multi core quiet build, watch htop for activity!'
+  sh bin/package quiet only make 'ast-ksh' SHOPT_COSHELL=0 || :
+  # Make errors can be found in src/arch/linux.i386-64/lib/package/gen or by
+  # bin/package results
+
+  # All the build errors happen after ksh so they don't seem too important.
+
+  # Sometimes package make will return false and halt the PKGBUILD.
+  # We'll implement our own detection. ksh guarantees that nmake also built.
+  set +u
+  if [ ! -x "arch/$(sh bin/package host)/bin/ksh" ]; then
+    echo '>>>ksh not found, build failed'
+    false
+  else
+    echo '>>>ksh build complete, ready to package'
+  fi
+}
+
+check() {
+  cd "${srcdir}"
+  # This takes longer than the build so don't run it unless something big 
+  # changes. It's not multi core so doesn't go much faster with a big CPU.
+  # sh bin/package test
+
+  # 2012.08.01 not tested
+  # 2014.06.25 No errors up to where it freezes at
+  # "test exit begins at 2015-06-05+06:51:17"
+  # I had to kill the process.
+}
+
+# http://stackoverflow.com/questions/462100/bash-script-to-create-symbolic-links-to-shared-libraries
+# $1: 0=cascade links    a.so -> a.so.1 -> a.so.1.1
+#     1=consistent links a.so -> a.so.1.1, a.so.1 -> a.so.1.1
+# $2: path of lib files
+# Links will overwrite copied files but it's generally recommended to not copy
+# unnecessary files.
+_liblinks() {
+# liblinks - generate symbolic links 
+# given libx.so.0.0.0 this would generate links for libx.so.0.0, libx.so.0, libx.so
+  local _FILE
+  for _FILE in "$2"/*.so*; do
+    #echo "${_FILE}"
+    local _shortlib="$(basename "${_FILE}")"
+    local _basename="${_shortlib}"
+    local _dirname="$(dirname "${_FILE}")"
+    local _extn
+    while _extn="$(echo "${_shortlib}" | sed -n '/\.[0-9][0-9]*$/s/.*\(\.[0-9][0-9]*\)$/\1/p')" &&
+        [ -n "${_extn}" ]; do
+      _shortlib="$(basename "${_shortlib}" "${_extn}")"
+      ln -sf "${_basename}" "${_dirname}/${_shortlib}"
+      if [ "$1" -eq 0 ]; then
+        _basename="${_shortlib}"
+      fi
+    done
+  done
+}
+
+package() {
+  set -u
+  cd "${srcdir}"
+  install -dm755 "${pkgdir}/usr/bin"
+  install -dm755 "${pkgdir}/usr/lib"
+  if ! :; then
+    install -dm755 "${pkgdir}/usr/include"
+    # ksh doesn't install ksh
+    # ast-base installs way too much including a bunch of conflicting tools
+    # ast-ksh doesn't install nmake
+    # ast-ksh+ast-make same as ast-base
+    sh bin/package flat install "${pkgdir}/usr/" 'ast-ksh' 'ast-make'
+  else
+    ## Install manually to avoid nmake dep
+    # The Debian package was ast-ksh which doesn't include nmake. The entire
+    # package is built from Mamfiles and mamake. I build ast-base which 
+    # includes nmake and other utilities.
+    # We do it manually to avoid the binaries we already have and we don't want
+    # a full AT&T toolchain.
+
+    # Copy the Debian man pages for shcomp. We even get a French version. 
+    # Whoo-hoo! Like Ubuntu, I'm leaving the Debian branding in.
+    cd "${srcdir}/debian/usr/share/man"
+    local _doc
+    for _doc in `find . -type f -name 'shcomp*.gz'`; do
+      install -Dm644 "${_doc}" "${pkgdir}/usr/share/man/${_doc}"
+    done
+
+    # Install the shells with links to the alternates.
+    cd "${srcdir}"
+    cd "arch/$(sh bin/package host)"
+    install -Dm755 'bin/ksh' "${pkgdir}/usr/bin/ksh"
+    # The Gentoo ebuild shows us that softlinks work for the alternate shells.
+    # Running rksh does show 'restricted' so I know that one works.
+    local _exe
+    for _exe in 'ksh93' 'rksh' 'pfksh'; do
+      ln -sf 'ksh' "${pkgdir}/usr/bin/${_exe}"
+    done
+    install -Dm755 'bin/shcomp' "${pkgdir}/usr/bin/shcomp"
+    install -Dm644 'man/man1/sh.1' "${pkgdir}/usr/share/man/man1/ksh.1"
+
+    # Install Debian's kshrc file. I didn't check it for correctness to Arch Linux.
+    install -dm755 "${pkgdir}/usr/share/doc/${pkgname}"
+    install -Dm644 "${srcdir}/debian/usr/share/doc/ksh/example.kshrc" "${pkgdir}/usr/share/doc/${pkgname}/example.kshrc"
+    # Install the same misc files that Debian installs.
+    local _doc
+    for _doc in 'COMPATIBILITY' 'OBSOLETE' 'PROMO.mm' 'RELEASE' 'RELEASE88' 'RELEASE93'; do
+      install -Dm644 "${srcdir}/src/cmd/ksh93/${_doc}" "${pkgdir}/usr/share/doc/${pkgname}/${_doc}"
+      gzip "${pkgdir}/usr/share/doc/${pkgname}/${_doc}"
+    done
+
+    # Debian wanted these too. It's not clear to me how they are used, why 
+    # we change fun to functions, or why ast-base has made them 755 and Debian
+    # has them 644. I'm just following along for no good reason.
+    install -dm755 "${pkgdir}/usr/share/${pkgname}/functions"
+    local _fun
+    for _fun in 'dirs' 'popd' 'pushd'; do
+      install -Dm644 "fun/${_fun}" "${pkgdir}/usr/share/${pkgname}/functions/${_fun}"
+    done
+
+    # Install the nmake option, just in case someone thinks
+    # http://www2.research.att.com/~astopen/download/faq.html
+    # "Is nmake(1) really the greatest thing since sliced bread?"
+    if [ "${_opt_InstallNMAKE}" -ne 0 ]; then
+      echo '>>>Installing optional nmake'
+      install -Dm755 'bin/nmake' "${pkgdir}/usr/bin/nmake"
+      install -Dm644 'man/man1/nmake.1' "${pkgdir}/usr/share/man/man1/nmake.1"
+      install -Dm755 'lib/'libast.so.* "${pkgdir}/usr/lib/"
+      _liblinks 1 "${pkgdir}/usr/lib"
+    fi
+  fi
+
+  # Install modern LICENSE file. There are also BSD and ZLIB copyrights and 
+  # licenses to consider. Noone else is adding them so they are likely 
+  # unnecessary.
+  cd "${srcdir}"
+  sh bin/package copyright 'ast-ksh' > "${srcdir}/LICENSE"
+  sh bin/package license 'ast-ksh' >> "${srcdir}/LICENSE"
+  install -Dm644 "${srcdir}/LICENSE" "${pkgdir}/usr/share/licenses/${pkgname}/LICENSE"
+  set +u
+}
+set +u
+# any sums down here need to be moved in place above, with a preceeding :;
+srcinfo_write_attr () 
+{ 
+    local attrname=$1 attrvalues=("${@:2}");
+    attrvalues=("${attrvalues[@]//+([[:space:]])/ }");
+    attrvalues=("${attrvalues[@]#[[:space:]]}");
+    attrvalues=("${attrvalues[@]%[[:space:]]}");
+    printf "	$attrname = %s
+" "${attrvalues[@]}"
+}
+sha256sums=('8852b9d37d5034e3780aeb5f963726381eeb4e08bb5bee1fbfa7e3f529c10e1b'
+            '58588b07b076f05dbbd5f4f095d5753309a8356ba1e5475262ce77d6bff42dae'
+            '37495cc625a2174b22a43542acac1d69402ee4992ee084a84690546c5b932b39')