Great - but where do we put this?

Might this be a good argument for a super-repository?

--
Anders


Wed Jan 07 2015 at 7:13:40 PM skrev Martin Sandve Alnæs <[email protected]
>:

>  Here's an experimental script draft that does a full install of fenics
> components on the side after using hashdist for the dependencies.
>
>  Martin
>
> On 7 January 2015 at 13:45, Martin Sandve Alnæs <[email protected]>
> wrote:
>
>> Great, that's the way I wanted it myself. It's much cleaner not to have a
>> hashdist build if fenics in the path hidden behind my own build.
>>
>> Martin
>>
> 7. jan. 2015 13:24 skrev "Anders Logg" <[email protected]>:
>>
>  Something I just noticed is that it does not work to first build all of
>>> FEniCS with fenics-install.sh (HashDist) and then use fenics-dev-install.sh
>>> to build e.g. DOLFIN. The build itself works fine, but when one later runs
>>> an application, the DOLFIN built by HashDist will be picked up (because of
>>> the use of rpath).
>>>
>>
>>>  So developers need to build only the dependencies of FEniCS using
>>> fenics-install.sh and then manually install FEniCS components using
>>> fenics-dev-install.sh.
>>>
>>>  Johannes has just pushed a fix to fenics-install.sh that helps with
>>> this. There is now an option to only build the dependencies of FEniCS. The
>>> script will ask the following question when run:
>>>
>>>  The following build types are supported:
>>>
>>>    [0] development version of FEniCS [default]
>>>   [1] latest stable version of FEniCS, currently 1.5
>>>   [2] only dependencies, including e.g. MPI, PETSc, Swig, Boost
>>>
>>>  Please select build type [0, 1, 2]:
>>>
>>>  Even if the installation script is not yet foolproof, at the time of
>>> release of 1.5, I would like to promote the use of fenics-install.sh as the
>>> default way to install FEniCS on the web page (next to the
>>> binaries/packages). Dorsal is no longer actively supported and we have been
>>> lacking a working and recommended install process of development versions
>>> for some time now.
>>>
>>>  Any objections to this?
>>>
>>>  --
>>> Anders
>>>
>>>
>>>  _______________________________________________
>>> fenics mailing list
>>> [email protected]
>>> http://fenicsproject.org/mailman/listinfo/fenics
>>>
>>>
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to