[openib-general] DevCon "decision" on userspace

[Sasha Khapyorsky]

On 14:07 Tue 21 Nov     , Woodruff, Robert J wrote:
> 
> >Everything (except ibutils) was converted from gen2/*/src/userspace/.
> >libibverbs and libmthca are not in the list because it is already
> >converted by Roland and published on kernel.org.
> 
> Should we mirror Roland's libibverbs and libmthca git trees somewhere on
> the openfabrics
> server so that the openfabrics server contains a copy of all of the
> code?

Don't know yet. It is up to our (and Roland) decision. In general I
think would be better to avoid the situation when somebody (or crontab)
will need to mirror Roland's trees from kernel.org - it is trivial to
do, but in case of problems we will need to check that "mirroring" is
not a source of this problems.

Other option could be "fake" git tree which will have only
objects/info/altenatives file with pointer to kernel.org location, but
I didn't evaluate this technically yet.

> >5. I will place symbolic links to maintainer's trees at /pub/scm , so
> >  maintainer's trees will be visible via gitweb (and also accessible by
> >  git as git://staging.openfabrics.org/<project-name>)
> 
> Or can you have a symboic link to kernel.org for Roland's libibverbs and
> libmthca ?

We cannot have symbolic link to other machines.

> >Any comments, suggestions, objections?
> 
> Should we have either a mirror git tree or symbolic link in /pub/scm to
> the 
> kernel code too ? Roland's git tree, and the git tree's for the other
> kernel components
> that are not yet in Roland's tree, SDP, etc. 

As above this is probably possible via "fake" trees or so, but not sure.
Do we really want to host kernel sources? It is hosted fine on kernel.org
today.

Sasha