D-Opt porting:

Some problems under the port: 

1) gcc claims that some variables are used in different
  places with different precision, specifically, argument
  3 and 5 to subroutine INV are declared in the routine
  as REAL but called as double precision.  This didn't 
  seem to cause problems (?). I corrected this.

2) Still, after that change, I get sometimes a General
  Protection Fault, when using blocks. That seems to
  depend on size&number of block: When using 7 blocks
  of size 3 each, everything works right, but 2 blocks of
  size 12 cause GPF.

  More details: 

  l <- c("-","+")
  fac <- expand.grid(l,l,l,l,l,l,l,l,l,l,l)
  names(fac) <- LETTERS[-6][1:11]      # omit F as factor name

  dsn <- Dopt(~ A+B+C+D+E+G+H+I+J+K+L, 12, fac, initial=1, fixed=T) 
==>  OK
  dsn <- Dopt(~ A+B+C+D+E+G+H+I+J+K+L, 24, fac, initial=1, fixed=T) 
==>  OK
  dsn <- Dopt(~ A+B+C+D+E+G+H+I+J+K+L, c(rep(1,7), rep(2,7)), fac,
initial=1, fixed=T)  ==>  GPF (R terminates)
  dsn <- Dopt(~ A+B+C+D+E+G+H+I+J+K+L, c(rep(1,8), rep(2,8)), fac,
initial=1, fixed=T)  ==>  GPF 
  dsn <- Dopt(~ A+B+C+D+E+G+H+I+J+K+L, rep(1:8, rep(3,8)), fac,
initial=1, fixed=T)     ==>  OK
  dsn <- Dopt(~ A+B+C+D+E+G+H+I+J+K+L, rep(1:3,rep(6,3)), fac,
initial=1, fixed=T)      ==>  OK


3) I cannot get the formula handling for blocks to work, 
   but simply disabling it and using vectors for
   designing blocks is OK. Do I really loose
   something in this?




Kjetil Halvorsen.

