[Clp-tickets] [COIN-OR Linear Programming Solver] #58: cbc.exe -- Linear relaxation infeasible

[COIN-OR Linear Programming Solver]

#58: cbc.exe -- Linear relaxation infeasible
----------------------------------------------+----------------------
 Reporter:  vdevil                            |      Owner:  somebody
     Type:  defect                            |     Status:  new
 Priority:  major                             |  Milestone:
Component:  component1                        |    Version:
 Keywords:  cbc Linear relaxation infeasible  |
----------------------------------------------+----------------------
 Hello,

 I'm using cbc.exe (built from CoinAll-1.6.0) to solve a linear problem.
 The presolve is set to off.

 If I use DualSimplex Cbc doesn't find the solution and says :

 {{{
 Coin:dualS
 Perturbing problem by 0.001 % of 15328.673 - largest nonzero change
 0.014293713
 (% 0.13843725) - largest zero change 0.014850191
 0  Obj -1.0011806e+012 Primal inf 1.228312e+015 (120)
 4  Obj 3.7294764e+019 Primal inf 1.0874013e+024 (114)
 7  Obj 1.7568362e+023 Primal inf 5.0281442e+024 (113)
 9  Obj 2.1954681e+028 Primal inf 5.6899824e+029 (121)
 10  Obj 2.1954682e+028 Primal inf 5.6899957e+029 (124)
 11  Obj 2.1954682e+028 Primal inf 5.6900909e+029 (122)
 13  Obj 2.1956923e+028 Primal inf 5.6901586e+029 (123)
 15  Obj 2.1956936e+028 Primal inf 5.6913518e+029 (126)
 16  Obj 2.1956936e+028 Primal inf 5.6913518e+029 (130)
 18  Obj 2.1956936e+028 Primal inf 5.6913518e+029 (131)
 20  Obj 2.1956936e+028 Primal inf 5.6913518e+029 (133)
 21  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (133)
 23  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (140)
 24  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (142)
 25  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (143)
 26  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (144)
 28  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (145)
 30  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (144)
 32  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (144)
 75  Obj 6.5876509e+028 Primal inf 1.7075533e+030 (109)
 76  Obj 6.5888849e+028 Primal inf 1.7075533e+030 (112)
 77  Obj 6.5888849e+028 Primal inf 1.7075533e+030 (111)
 79  Obj 6.5890839e+028 Primal inf 1.7075533e+030 (111)
 80  Obj 6.5890839e+028 Primal inf 1.7075533e+030 (113)
 81  Obj 6.5890839e+028 Primal inf 1.7075533e+030 (114)
 82  Obj 6.5891369e+028 Primal inf 1.7075533e+030 (114)
 83  Obj 6.5891439e+028 Primal inf 1.7075533e+030 (122) Dual inf
 0.00034051428 (1
 )
 84  Obj 6.5891439e+028 Primal inf 1.7075533e+030 (124)
 85  Obj 6.5894741e+028 Primal inf 1.7075533e+030 (140)
 97  Obj 6.5894741e+028 Primal inf 1.7075533e+030 (114)
 98  Obj 2.6350363e+029 Primal inf 1.7075735e+030 (115)
 98  Obj 2.6350363e+029 Primal inf 1.7075735e+030 (115)
 Primal infeasible - objective value 2.0628922e+029
 PrimalInfeasible objective 2.062892195e+029 - 98 iterations time 0.042

 Result - Linear relaxation infeasible

 Enumerated nodes:           0
 Total iterations:           0
 Time (CPU seconds):         5271.35
 Time (Wallclock Seconds):   5271.35
 }}}


 But if I directly make another call to DualSimplex Cbc finds the solution
 ?! :

 {{{

 Coin:dualS
 Perturbing problem by 0.001 % of 15328.673 - largest nonzero change
 0.013690577
 (% 5.2125114) - largest zero change 0.014466123
 0  Obj 1.2372034e+011 Primal inf 2.5945057e+009 (37) Dual inf 2110055.5
 (7)
 6  Obj 4.8619578e+012 Primal inf 6.528013e+008 (39)
 10  Obj 6.5771001e+012 Primal inf 3.6421095e+008 (30)
 15  Obj 7.1419465e+012 Primal inf 2.7443405e+008 (30)
 17  Obj 7.1419509e+012 Primal inf 2.5510148e+008 (28)
 32  Obj 7.1461345e+012 Primal inf 2.5386987e+008 (10)
 34  Obj 7.1461345e+012 Primal inf 3.3829262e+008 (16)
 35  Obj 7.1461345e+012 Primal inf 3.3819038e+008 (17)
 36  Obj 7.1461345e+012 Primal inf 3.4818665e+008 (16)
 36  Obj 7.1461345e+012 Primal inf 3.4818665e+008 (16)
 36  Obj 7.1461345e+012 Primal inf 3.4818665e+008 (16)
 54  Obj -1.2740085e+013 Primal inf 7.3715614e+012 (41)
 55  Obj -1.2740085e+013 Primal inf 7.3713401e+012 (43)
 57  Obj -1.0186659e+013 Primal inf 5.4809212e+012 (42)
 60  Obj 5.0106785e+011 Primal inf 31121055 (15)
 64  Obj 5.2745316e+011 Primal inf 88664.554 (7)
 69  Obj 5.2816431e+011 Primal inf 4398607.9 (15)
 76  Obj 5.2848012e+011 Primal inf 109.27733 (2)
 82  Obj 5.2860406e+011
 82  Obj -1.1280161e+011 Primal inf 1.1778729e+015 (25) Dual inf
 7.0426814e+014
 7)
 94  Obj 3.8456318e+011 Primal inf 9.7273275e+008 (10) Dual inf 61020822
 (1)
 107  Obj 2.2089191e+011 Dual inf 0.022728333 (2)
 108  Obj 2.2089191e+011
 Optimal - objective value 2.2089191e+011
 Optimal objective 2.208919053e+011 - 108 iterations time 0.042

 }}}

 Clp and Cplex both find the solution. I know it would be more convenient
 to use Clp but for some reason I have to go with Cbc.

 Is there a way to get rid of the infeasible linear relaxation problem ?
 (if it's the origin of the problem)
 What should I do ?

 Thanks in advance.

-- 
Ticket URL: <https://projects.coin-or.org/ticket/58>
COIN-OR Linear Programming Solver <http://projects.coin-or.org/Clp>
A linear programming solver.