If ntp is used, boinc should start afterwards - otherwise work units tends to fail, if the system time steps toward via ntpdate.
You probably shouldn't use ntpdate at all in this case, or at least not in the same runlevel.
Please further explain the problem boinc has - setting a faux depend in the init.d script isn't the solution. (Also, I'd already have marked this bug report invalid if you admitted to running ntpdate while the system services are completely up and running.)
I run ntpdate only during boot of the system in runlevel default.
After moving a system from one hard disc to another disc a lot of WU failed. Looking for the root cause I stumbled over error entries both in boinc log files and in the reported result files (the later however I didn't see anymore on the wep page of "world community grid" and "Einstein@home").
I tested a week ago or so to set the clock to "UTC", but run into too much side effects - I went back to "local".
Might be that the boinc errors are the results of my doing, but nevertheless ntpdate should be called as early as possible, isn't it ?
You can set up both ntp-client and ntpd in the default runlevel, and the ntpd init.d script should already take care that ntp-client is run first. Maybe adding "after ntp-client" to the boinc init.d script might work.
(In reply to comment #3)
> Maybe adding "after ntp-client" to the boinc init.d script might work.
I just putted ntpd into the use line of the boinc init.d script - that works. thx
+*boinc-7.2.0 (30 Jun 2013)
+ 30 Jun 2013; Justin Lecher <email@example.com> -boinc-7.0.29.ebuild,
+ +boinc-7.2.0.ebuild, -files/7.0.2-fix_subdirs.patch,
+ +files/boinc-7.2.0-fix_subdirs.patch, files/boinc.init,
+ files/generate_tarball.sh, metadata.xml:
+ Version bump and drop old, fixes: security issue, #467908; link problems,
+ #464772; correlation with abrupt time changes through ntp, #413041; menu
+ items with newer x11-libs/wxGTK, #379979 & #376237