Arch_rnd bugs .. patch attached

[John W. Eaton]

On 31-Jan-2008, Dave Goel wrote:

| (I only have octave2.1.73, but I notice that this bug remains unfixed
| in octave 2.9).
| 
| Problem: 
| 
| Trying to generate an AR(2) or higher process always fails, for
| example:
| 
| octave:74>  seq=arch_rnd ([1 1 1], [.5; .5;.5]', 10)
| error: operator *: nonconformant arguments (op1 is 1x2, op2 is 0x1)
| error: evaluating binary operator `*' near line 84, column 27
| error: evaluating binary operator `+' near line 84, column 17
| error: evaluating assignment expression near line 84, column 10
| error: evaluating for command near line 82, column 3
| error: called from `arch_rnd' in file
| `/usr/share/octave/2.1.73/m/signal/arch_rnd.m'
| error: evaluating assignment expression near line 74, column 5
| 
| 
| The problem is that at several points in the source file, a range is
| specified as higher:lower, where it should be lower:higher.
| 
| The problem never shows up for AR(1) or AR(0) because lower=higher in
| that case.
| 
| The attached patch fixes the problem, I believe, for octave2.1.73, and
| can be applied to the current arch_rnd as well.  The patch is included
| here, and also attached.

I applied this patch and checked it in.

Thanks,

jwe