Re: New slurm-llnl package for squeeze
On Sat, 2010-11-06 at 10:09 -0500, Dirk Eddelbuettel wrote:
> On 6 November 2010 at 14:14, Adam D. Barratt wrote:
> | On Sat, 2010-11-06 at 09:01 -0500, Dirk Eddelbuettel wrote:
> | > On 4 November 2010 at 17:56, Adam D. Barratt wrote:
> | > | It looks like the current Debian package doesn't set LD_LIBRARY_PATH at
> | > | all in the slurm-llnl-slurmdbd init script. Was the addition of that to
> | > | the init script in the updated package intentional? (Upstream's init
> | > | script does set LD_LIBRARY_PATH, so it makes sense there).
> | >
> | > So I am totally confused. Does that mean there is no bug?
> | Gennaro's proposed fix modifies two init scripts, only one of which
> | previously set LD_LIBRARY_PATH. The fix for that one script is fine,
> | the question was whether the change to the second script was intentional
> | and appropriate (as it will lead to an extra directory being added to
> | LD_LIBRARY_PATH which was not previously included).
> Thank you -- that's how I was reading the discussion to.
> But now I need a consensus recommendation as to what I should upload. As the
> package was prepared by Gennaro, or with another alteration?
So far as I can see, Gennaro's modification to the second init script -
as is - doesn't make sense. It adds $LIBDIR to LD_LIBRARY_PATH, but
never defines LIBDIR.
I'd be happy with either reverting the modification to the
slurm-llnl-slurmdbd init script, or adding the definition of LIBDIR to
it (bringing it closer to upstream), but doing one without the other
seems wrong; it also means that if the environment happens to contain a
$LIBDIR at the time the init script is called then that will end up
being prepended to the LD_LIBRARY_PATH used in the script.
Gennaro - which approach you take is up to you.