Difficulty with Gnumeric's LP Solver

[Marcelo Gomes de Queiroz <mqz ime usp br>]

Hi!

I had trouble solving the following LP problem 
within Gnumeric (1.0.9, Linux kernel 2.4.18):

max 256x1 +  64x2 + 16x3 + 4x4 + x5
s.t.   x1                           <=     1
      8x1 +    x2                   <=    16
     32x1 +   8x2 +   x3            <=   256
    128x1 +  32x2 +  8x3 +  x4      <=  4096
    512x1 + 128x2 + 32x3 + 8x4 + x5 <= 65536
    x1,x2,x3,x4,x5 >= 0

The reported answer is (0,0,0,0,691.2289), the
optimal solution is (0,0,0,0,65536). The standard
form of the same problem (5 extra slack variables,
all equality constraints) gives the result
(0,0,0,0,691.2289,1.999858,10.00085,42.03045,170.5587,0)
which isn't feasible.

I don't think this has anything to do with stability 
issues because I tried the same problems with lp_solve 
and glpsol (embedded in Gnumeric 1.0.x and 1.1.x series)
and both work fine.

Would you please help me? I put both examples in
http://www.ime.usp.br/~mqz/test1.gnumeric
and
http://www.ime.usp.br/~mqz/test2.gnumeric

Thanks in advance!

Marcelo Queiroz