From: Thomas Petazzoni Date: Sun, 6 Mar 2016 23:15:18 +0100 Subject: [Buildroot] [PATCH v6 3/3] ypbind-mt: new package In-Reply-To: <1453832900-25757-4-git-send-email-yba@tkos.co.il> References: <1453832900-25757-1-git-send-email-yba@tkos.co.il> <1453832900-25757-4-git-send-email-yba@tkos.co.il> Message-ID: <20160306231518.63807855@free-electrons.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Dear Jonathan Ben-Avraham, Sorry for the slow answer :/ On Tue, 26 Jan 2016 20:28:20 +0200, Jonathan Ben-Avraham wrote: > diff --git a/package/ypbind-mt/S70ypbind b/package/ypbind-mt/S70ypbind > new file mode 100755 > index 0000000..e73bec6 > --- /dev/null > +++ b/package/ypbind-mt/S70ypbind > @@ -0,0 +1,93 @@ > +#!/bin/sh > +# Based on /etc/init.d/ypbind from SuSe Linux by Thorsten Kukuk > + > +DAEMON=/usr/sbin/ypbind > +pidfile=/var/run/ypbind.pid Capital letters for the variable. > +[ -x ${DAEMON} ] || exit 0 Not needed, the daemon is guaranteed to be installed. > +[ -f /etc/default/ypbind ] && source /etc/default/ypbind > + > +case "$1" in > + start) > + echo -n "Starting ypbind: " echo -n => printf > + # Check for preset YP domainname > + ypdomainname &> /dev/null > + if [ $? -ne 0 ]; then > + if [ ! -f /etc/defaultdomain ]; then > + # No configured YP domainname > + echo "FAIL" > + exit 1 > + fi > + # Assume that /etc/defaultdomain contains > + # a valid domainname and assume ypdomainname > + # succeeds > + XDOMAINNAME=$(cat /etc/defaultdomain) > + ypdomainname "${XDOMAINNAME}" > + fi > + > + if [ ! -f /etc/yp.conf -a "${YPBIND_BROADCAST}" != "yes" ]; then > + # We need a local configuration if not in broadcast mode > + echo "FAIL" > + exit 1 > + fi > + > + OPTIONS="" > + test "${YPBIND_LOCAL_ONLY}" = "yes" && \ > + OPTIONS="-local-only ${OPTIONS}" > + test "${YPBIND_BROADCAST}" = "yes" && \ > + OPTIONS="-broadcast ${OPTIONS}" > + test "${YPBIND_BROKEN_SERVER}" = "yes" && \ > + OPTIONS="-broken-server ${OPTIONS}" Why are we doing this complicated dance instead of just having YPBIND_OPTIONS defined in /etc/default/ypbind ? > + start-stop-daemon --start --quiet --pidfile ${pidfile} \ > + --exec ${DAEMON} -- ${OPTIONS} > + > + if [ $? -ne 0 ]; then > + echo "FAIL" > + exit 1 > + fi You should print OK if start-stop-daemon succeeded. Look at other init scripts in Buildroot. > + > + # Make sure we have a server so that later scripts can know > + notfound=1 > + for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15; do for i in $(seq 1 16) > + if ypwhich &>/dev/null; then > + notfound=0 > + break > + fi > + echo -n " ." > + sleep 1; > + done > + > + if [ ${notfound} -eq 1 ]; then > + # No NIS server found > + echo "FAIL" > + exit 1 > + fi > + > + echo "OK" Could you move all of this into a start() function ? It would nicely reduce the indentation level. > + ;; > + > + stop) > + echo "Shutting down ypbind " printf > + start-stop-daemon --stop --quiet --pidfile $pidfile And print OK / FAIL. Look at other init scripts in Buildroot to see which options we typically pass when calling start-stop-daemon to stop a daemon. > + restart) > + ${0} stop > + sleep 1 > + ${0} start Then you can use the stop() and start() functions. > @@ -0,0 +1,24 @@ > +################################################################################ > +# > +# ypbind-mt > +# > +################################################################################ > + > +# The github releases of ypbind-mt are often a release or more ahead of the > +# "Download" links on http://www.linux-nis.org/nis/ypbind-mt/ > +YPBIND_MT_VERSION = ypbind-mt-2_2 > +YPBIND_MT_SITE = $(call github,thkukuk,ypbind-mt,$(YPBIND_MT_VERSION)) > +YPBIND_MT_LICENSE = GPLv2 > +YPBIND_MT_LICENSE_FILES = COPYING > +YPBIND_MT_AUTORECONF = YES > +YPBIND_MT_DEPENDENCIES = host-pkgconf yp-tools > +YPBIND_MT_CONF_ENV = \ > + PKG_CONFIG_SYSROOT_DIR="$(TARGET_DIR)" \ > + PKG_CONFIG_PATH="$(TARGET_DIR)/usr/lib/pkgconfig" This is wrong. There is no reason to have .pc files in $(TARGET_DIR), and they are in fact removed at the end of the build. Can you explain why you override those variables ? Could you rework the patch to take into account those comments and send an updated version? Thanks a lot! Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux, Kernel and Android engineering http://free-electrons.com