dolfin team mailing list archive

[Johan Hake <johan.hake@xxxxxxxxx>]

On Monday August 23 2010 12:09:32 Harish Narayanan wrote:
> On 8/24/10 12:18 AM, Johan Hake wrote:
> > On Monday August 23 2010 11:40:52 Harish Narayanan wrote:
> >> On 8/23/10 11:38 PM, Anders Logg wrote:
> >>> On Mon, Aug 23, 2010 at 11:04:00AM -0700, Johan Hake wrote:
> >>>> On Monday August 23 2010 10:52:12 Anders Logg wrote:
> >>>>> On Mon, Aug 23, 2010 at 10:33:56AM -0700, Johan Hake wrote:
> >>>>>> On Monday August 23 2010 10:15:32 Anders Logg wrote:
> >>>>>>> On Mon, Aug 23, 2010 at 10:00:48AM -0700, Johan Hake wrote:
> >>>>>>>> On Monday August 23 2010 09:50:12 Anders Logg wrote:
> >>>>>>>>> On Mon, Aug 23, 2010 at 09:48:02AM -0700, Johan Hake wrote:
> >>>>>>>>>> On Monday August 23 2010 09:32:40 Anders Logg wrote:
> >>>>>>>>>>> On Mon, Aug 23, 2010 at 09:25:29AM -0700, Johan Hake wrote:
> >>>>>>>>>>>> Hello!
> >>>>>>>>>>>> 
> >>>>>>>>>>>> CMake works smooth here. Much nicer config messages too. You
> >>>>>>>>>>>> are doing a great job getting this up and running!
> >>>>>>>>>>>> 
> >>>>>>>>>>>> The only test that wont pass for me now is SLEPc. It
> >>>>>>>>>>>> complains about PETSC_ARCH not beeing set or set to empty.
> >>>>>>>>>>>> 
> >>>>>>>>>>>> I have installed PETSc and SLEPc using dorsal.
> >>>>>>>>>>>> 
> >>>>>>>>>>>> Johan
> >>>>>>>>>>> 
> >>>>>>>>>>> Which platform are you running?
> >>>>>>>>>> 
> >>>>>>>>>> Lucid
> >>>>>>>>>> 
> >>>>>>>>>>> Is PETSC_ARCH set?
> >>>>>>>>>> 
> >>>>>>>>>> Nope.
> >>>>>>>>>> 
> >>>>>>>>>>> It's very easy to debug. Just add stuff like
> >>>>>>>>>>> 
> >>>>>>>>>>>   message("PETSC_ARCH = ${PETSC_ARCH}")
> >>>>>>>>>>> 
> >>>>>>>>>>> in the file cmake/FindSLEPc.cmake.
> >>>>>>>>>> 
> >>>>>>>>>> I see that the FindPETSc.cmake includes some logic for when
> >>>>>>>>>> PETSC_ARCH is not set. This is lacking in FindSLEPc.cmake, which
> >>>>>>>>>> requires PETSC_ARCH to be set.
> >>>>>>>>> 
> >>>>>>>>> ok, now I see what the problem is. I'll see if I can fix it.
> >>>>>>>>> 
> >>>>>>>>> The way Dorsal installs PETSc has changed. It now does the
> >>>>>>>>> PETSc-style installation = no installation which required
> >>>>>>>>> PETSC_DIR and PETSC_ARCH to be set.
> >>>>>>>> 
> >>>>>>>> Ok, still on old dorsal. I liked that everything is installed in
> >>>>>>>> one place skipping the need to PETSC_ARCH.
> >>>>>>> 
> >>>>>>> Yes, but it turns out that didn't work very well with SLEPc 3.1 so
> >>>>>>> we had to change.
> >>>>>> 
> >>>>>> Ok?
> >>>>>> 
> >>>>>> I am on SLEPc 3.1 but I had to switch to manual install. All tests
> >>>>>> compiles
> >>>>>> 
> >>>>>> and runs for SLEPc using:
> >>>>>>   make examples
> >>>>> 
> >>>>> What do you mean by manual install?
> >>>> 
> >>>> I go to the src directory of SLEPc and configure and make it myself.
> >>>> SLEPC_DIR is set to prefix, but I temporarily set it to PWD when
> >>>> compiling. If I have a slepc installation in prefix I need to remove
> >>>> that first...
> >>>> 
> >>>>   SLEPC_DIR=$(pwd) ./configure --prefix=$HOME/local
> >>>>   make install
> >>> 
> >>> ok. You should be able to use Dorsal for this now.
> >>> 
> >>> It would be good if as many as possible (even us developers) used
> >>> Dorsal so that it gets enough testing.
> > 
> > I am using it for most part. Trilinos and SLEPc beeing the exceptions.
> 
> Oh, so your religion allows you to /use/ it...

:D

A bash free life is still far fetched...

Johan