fPortfolio ERRORs (caused by quadprog?)

[Berwin A Turlach]

G'day all,

On Wed, 07 Apr 2010 11:29:21 +0200
Uwe Ligges <ligges at statistik.tu-dortmund.de> wrote:

> Dear Rmetrics team,
> 
> fPortfolio is failing under Windows these days:
> It either crashes or hangs these days under all flavors of R under 
> check. Please see the CRAN check summaries (R-release results are not 
> yet available there), e.g.
> http://www.r-project.org/nosvn/R.check/r-prerel-windows-ix86/fPortfolio-00check.html
> 
> It seems to be related to the recent quadprog update. Hence CCing
> Berwin Turlach.
> 
> Any ideas?

The FORTRAN code in quadprog was modified to return the Lagrangian
multipliers too.  I see in fPortfolio's solveRquadprog.R that in the
function .rquadprog the FORTRAN code of quadprog is called directly
instead of using quadprog's R interface.  

Changing 

    optim = .Fortran("qpgen2",
        as.double(Dmat),
        dvec = as.double(dvec),
        as.integer(n),
        as.integer(n),
        sol = as.double(rep(0, n)),
        crval = as.double(0),
        as.double(Amat),
        as.double(bvec),
        as.integer(n),
        as.integer(q),
        as.integer(meq),
        iact = as.integer(rep(0, q)),
        nact = as.integer(0),
        iter = as.integer(rep(0, 2)),
        work = as.double(work),
        ierr = as.integer(0),
        PACKAGE = "quadprog")

to

    optim = .Fortran("qpgen2",
        as.double(Dmat),
        dvec = as.double(dvec),
        as.integer(n),
        as.integer(n),
        sol = as.double(rep(0, n)),
	lagr = as.double(rep(0, q)),   ### here is the change
        crval = as.double(0),
        as.double(Amat),
        as.double(bvec),
        as.integer(n),
        as.integer(q),
        as.integer(meq),
        iact = as.integer(rep(0, q)),
        nact = as.integer(0),
        iter = as.integer(rep(0, 2)),
        work = as.double(work),
        ierr = as.integer(0),
        PACKAGE = "quadprog")

should do the trick.

HTH.

Cheers,

	Berwin