libraries/zziplib: Simplify doinst.sh for mandb.

Signed-off-by: B. Watson <urchlay@slackware.uk>

Signed-off-by: Willy Sudiarto Raharjo <willysr@slackbuilds.org>

2 files changed, 37 insertions, 45 deletions

diff --git a/libraries/zziplib/doinst.sh b/libraries/zziplib/doinst.sh
new file mode 100644
index 0000000000000..6bca1a9d276a4
--- /dev/null
+++ b/libraries/zziplib/doinst.sh
@@ -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 {} +
diff --git a/libraries/zziplib/zziplib.SlackBuild b/libraries/zziplib/zziplib.SlackBuild
index bcdd10b0783a4..35a61b8e90331 100644
--- a/libraries/zziplib/zziplib.SlackBuild
+++ b/libraries/zziplib/zziplib.SlackBuild
@@ -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