mirror of
https://github.com/Ponce/slackbuilds
synced 2024-11-25 10:03:03 +01:00
libraries/zziplib: Simplify doinst.sh for mandb.
Signed-off-by: B. Watson <urchlay@slackware.uk> Signed-off-by: Willy Sudiarto Raharjo <willysr@slackbuilds.org>
This commit is contained in:
B. Watson
Willy Sudiarto Raharjo
parent
cb71abc9f4
commit
984c47af33
2 changed files with 37 additions and 45 deletions
34
libraries/zziplib/doinst.sh
Normal file
34
libraries/zziplib/doinst.sh
Normal file
|
|
@ -0,0 +1,34 @@
|
|||
# 20240916 bkw: this may end up in the template, so explanation:
|
||||
#
|
||||
# The "man -k", "apropos", and "whatis" commands in Slackware's
|
||||
# man-db rely on a database of man pages, that gets built nightly by
|
||||
# /etc/cron.daily/man-db, which runs the mandb command. This means any
|
||||
# man pages installed by SBo packages should get added to the database
|
||||
# within 24 hours of the time they're installed.
|
||||
#
|
||||
# Well and good, except it doesn't always work: if the timestamps on
|
||||
# the /usr/man/man* directories in the package are older than the last
|
||||
# time the cron job ran, then the next time it runs, mandb will see
|
||||
# that the timestamp is older than the database, and will not search
|
||||
# for new man pages. In fact, when this happens, the man pages *never*
|
||||
# get added to the database.
|
||||
#
|
||||
# If you only ever install packages right after building them, you
|
||||
# won't have this problem. However, if you install an older package
|
||||
# you built yesterday (last week, etc), its /usr/man/man* timestamps
|
||||
# will be older than the man database...
|
||||
#
|
||||
# It's easy enough to avoid the problem. Duncan Roe came up with the
|
||||
# idea for this on the mailing list: Touch the man directories in the
|
||||
# doinst.sh script for any build that installs man pages. The new
|
||||
# man pages will get added to the database the next time the cronjob
|
||||
# runs.
|
||||
#
|
||||
# It does *not* mean that "man -k" will be able to find newly
|
||||
# installed man pages *immediately* after installing a new
|
||||
# package... but then, as pointed out by KB_SBo (aka King Beowulf)
|
||||
# on the mailing list, "locate" can't find the new files immediately
|
||||
# either (it updates nightly, same as mandb), and nobody complains
|
||||
# about that.
|
||||
|
||||
[ -d usr/man ] && find usr/man -type d -exec touch {} +
|
||||
|
|
@ -23,6 +23,7 @@
|
|||
# OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
|
||||
# ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
|
||||
# 20240916 bkw: BUILD=3, simplify doinst.sh for mandb.
|
||||
# 20240909 bkw: BUILD=2, generate doinst.sh for mandb.
|
||||
# 20240823 bkw: update for v0.13.78.
|
||||
# 20240808 bkw: update for v0.13.77.
|
||||
|
|
@ -33,18 +34,11 @@
|
|||
# - switch to cmake.
|
||||
# - make slack-desc more concise.
|
||||
|
||||
# Note: 0.13.71 had the file "usr/lib64/pkgconfig/zzip-zlib-config.pc",
|
||||
# and 0.13.72 doesn't have it. Dependees are alephone and milkytracker [*].
|
||||
# These build correctly without this .pc file. If anyone has issues
|
||||
# with new builds because of this, please contact me.
|
||||
# [*] desmume had zziplib listed in REQUIRES, but it doesn't actually
|
||||
# use it (maybe it did at one time?).
|
||||
|
||||
cd $(dirname $0) ; CWD=$(pwd)
|
||||
|
||||
PRGNAM=zziplib
|
||||
VERSION=${VERSION:-0.13.78}
|
||||
BUILD=${BUILD:-2}
|
||||
BUILD=${BUILD:-3}
|
||||
TAG=${TAG:-_SBo}
|
||||
PKGTYPE=${PKGTYPE:-tgz}
|
||||
|
||||
|
|
@ -121,43 +115,7 @@ cat $CWD/$PRGNAM.SlackBuild > $PKGDOC/$PRGNAM.SlackBuild
|
|||
|
||||
mkdir -p $PKG/install
|
||||
cat $CWD/slack-desc > $PKG/install/slack-desc
|
||||
cat $CWD/doinst.sh > $PKG/install/doinst.sh
|
||||
|
||||
cd $PKG
|
||||
|
||||
# 20240909 bkw: This bit may get added to the SBo templates, so I'll
|
||||
# comment it in excruciating detail.
|
||||
#
|
||||
# The idea is that, immediately after installing the package, the
|
||||
# "man -k", "whatis", "apropos" commands should be able to find the
|
||||
# newly-installed man pages. This means that, at package install time,
|
||||
# they have to be added to the man database with the mandb command.
|
||||
#
|
||||
# We *could* run "mandb -c" in the doinst.sh, which would rebuild the
|
||||
# whole cache, but it takes a long time to do in Slackware 15.0 (several
|
||||
# minutes). Running "mandb" without the -c is much quicker, but
|
||||
# doesn't always pick up the new man pages (see next paragraph). So
|
||||
# instead, we have to add them one-by-one with "mandb -f" commands.
|
||||
#
|
||||
# In theory, we could run mandb with arguments in doinst.sh, or just
|
||||
# rely /etc/cron.daily/man-db to pick up the new man pages every
|
||||
# day... but in practice, it fails to pick up new ones, if the
|
||||
# timestamp on /usr/man/man1 (man3, etc) is older than the mandb
|
||||
# database file. This happens if we install or reinstall a package
|
||||
# that was built yesterday or earlier, if it contains any man pages.
|
||||
#
|
||||
# Note that I'm *not* proposing removing the man pages from the
|
||||
# database on package removal. It could be done with a simple "mandb"
|
||||
# command in doinst.sh (takes ~10 seconds to run with an SSD, longer
|
||||
# with a platter drive), but the daily cron job already handles this
|
||||
# correctly, regardless of the timestamps on the man page dirs. So
|
||||
# when you remove a package, you'll have to live with its man pages
|
||||
# being in the database for up to 24 more hours. This is no worse
|
||||
# than its files still being listed in the locate database, and nobody
|
||||
# complains about that.
|
||||
|
||||
find usr/man -type f -a -name '*.gz' | \
|
||||
sed -e 's,^,chroot . /usr/bin/mandb -f "/,' \
|
||||
-e 's,$," \&>/dev/null,' \
|
||||
>> install/doinst.sh
|
||||
|
||||
/sbin/makepkg -l y -c n $OUTPUT/$PRGNAM-$VERSION-$ARCH-$BUILD$TAG.$PKGTYPE
|
||||
|
|
|
|||
Loading…
Reference in a new issue