[Neurodebian-users] FSL and neurodebian

Michael Hanke michael.hanke at gmail.com
Fri Jan 16 16:31:50 UTC 2015


Hi,

On Fri, Jan 16, 2015 at 2:49 PM, Jonathan Brooks <Jon.Brooks at bristol.ac.uk>
wrote:

> I wonder if I could make a plea to have 2 separate packages for fsl-core.
> One which still plays nice with SGE, and leaves the queue configuration in
> fsl_sub untouched, and the other which has the changes necessary for Condor
> (currently the default)? I think it is down to the user to make this
> choice, and not to have it forced on them.
>
> As it stands there are plenty of people who still use SGE and would prefer
> to have the “vanilla” FSL
>

The short answer is: this won't happen. Here is why:

- the current modified implementation does not favor any particular system
(condor is not the default, there is no default), it should works just fine
with SGE -- if not that is a bug, please report it and it will get fixed.
It is true, though, that the condor support is much better tested -- solely
for the fact that we use condor and nobody using SGE contributes feedback
on any testing efforts -- any such feedback is greatly appreciated

- shipping an unmodified fsl_sub is not possible, as it significantly
overfits the SGE setup at Oxford -- and most SGE instances will be a lot
different.

  a) e.g. it tries to email accounts with @fmrib.ox.ac.uk addresses -- most
admins won't like that

  b) the three-queue setup is arbitrary and mostly unnecessary, because SGE
should be perfectly capable of picking the right queue from the supplied
runtime estimate

Moreover, anyone interested in FSL packages that must not have that
particular generalization of cluster support can rebuild the Debian package
with this patch disable (simply comment out the respective line in
debian/patches/series). But keep the patch that fixes some more bashisms
that are still present in the upstream sources ;-)

Cheers,

Michael

-- 
Michael Hanke
http://mih.voxindeserto.de
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/neurodebian-users/attachments/20150116/af29d64b/attachment.html>


More information about the Neurodebian-users mailing list