trinity-devel@lists.pearsoncomputing.net

Message: previous - next
Month: April 2014

Never take vacation & Bug 1998

From: "David C. Rankin" <drankinatty@...>
Date: Thu, 10 Apr 2014 01:27:34 -0500
All,

  I'm back - I hope. What started as a planned week "family trip" (not exactly
vacation -- 3 kids, car 16 hour road trip)( was actually great), but on return
left us a week behind in all other aspects of life, with a case scheduled for
4/21 multiple depos remaining, pre-trial, etc.. -- then my client had a stroke.
(enough said, but I'm just finally getting unburried in time for taxes :-)

  Slavek, have you had a chance to look at bug 1998. I am stuck after
implementing the k4 multiseat patches in TDE, which compile beautifully and
appear to solve part of the problem with tde_dbus_hardwarecontrol, but I am lost
following the effects of the patch through the remainder of the code. (the
patches include a new tdebase/cmake/modules/FindSystemd.cmake, that makes
systemd available to cmake when found on the target system) I need help getting
this patch implemented. TDE is NOT enterprise ready without it. TDE must allow
user access to sound, user mount (cd/dvd/usb, etc..) via normal udev/polkit, and
must prevent the race conditions with tdehwlib/tdepowersave, etc. with the purse
systemd implementation. Right now it doesn't.

  To be blunt, TDE is unusable on pure systemd distros like Arch due to these
problems and I can't release TDE to Arch as usable until we get some way of
solving this problem. Right now I can build and run TDE, but it takes ugly hacks
to provide user access to sound or cd/dvd/usb. (essentially add every user to
the audio, sys and disk groups just to get access which is a glaring no-no)
Additionally, you have to disable tdepowersave to prevent filling the logs with
hundreds-of-thousands of errors (per hour).

  So far all examples that have been provided where TDE 'works' with systemd
have NOT been systemd implementations, but instead init-script implementations
of TDE that hand off control to systemd-sysvcompat, not true services where
'systemctl start tdm.service' is used to launch tde/tdm directly by systemd
itself. The init-script case is not systemd, but the compatibility mode for the
old init-script implementation. Many distros are working through the transition
and will be pure-systemd in the very near future (even Debian:
https://wiki.debian.org/systemd). So we need to get this fixed. Let's get a plan
together for testing/implementing the needed fix. I'm happy to do the work, but
it is not something I can do myself in any reasonable amount of time. I need
help finding/following all the needles through the TDE haystack. Let me know
when you have a bit of time to devote to the issue. All help is welcome.

-- 
David C. Rankin, J.D.,P.E.