Currently, the version 1.2.97 is checking whether hddtemp is listening on a socket during configure time: AC_MSG_CHECKING([for hddtemp being queryable via netcat]) NETCAT_RUN=([`"$NETCAT_PATH" "$NETCAT_HOST" "$HDDTEMP_PORT" 2> /dev/null`]) However, hddtemp may not be running at that moment even if it's installed. This is actually typical for Gentoo if hddtemp is pulled in as a dependency for xfce4-sensors-plugin by the package manager and installed without any user involvement. In this scenario, the package manager will install hddtemp without starting the daemon and proceed to compile the plugin. It is also possible to configure the package manager so that it doesn't allow any socket connections inside the sandbox as it's compiling the package even if hddtemp is actually running. It'd be nice if there's a way to bypass this, something like '--without-hddtemp-socket-check'.
You're right; the intention has already been to allow forcing some settings where required. As ./configure --enable-hddtemp=yes --enable-netcat --with-pathnetcat=/bin/netcat --with-pathhddtemp=/usr/bin/hddtemp doesn't suffice due to the mentioned check, I will try adding a disable-pathchecks option or similar (--enable-pathchecks=no). I wonder why the issue hasn't been mentioned years earlier?
(In reply to Fabian Nowak from comment #1) > I wonder why the issue hasn't been mentioned years earlier? I suspect many users enabled the setuid bit after they see a warning about the plugin getting -EPERM with hddtemp. Others probably just started the hddtemp daemon and reinstalled the plugin as a workaround, instead of reporting it upstream. My original intention was to make hddtemp configurable at runtime via the settings, never managed to allocate the time for that however :)
Committed: 1fe2ce3