[Coin-discuss] Sbb bug wrt to branch selection in bug tracking database...

[Edwards, Nathan J.]

...but it seems like the mailing lists are more active...

Bug #3871, was updated on 02/20/04 14:02
Here is a current snapshot of the bug.

Project: coin
Category: None
Status: Open
Resolution: None
Bug Group: None
Priority: 5
Summary: Segmentation fault in sbb...

Details:
...in SbbBranchDecision::bestBranch at SbbBranchBase.cpp:159.

bestObject never gets set to a non-zero value, so

bestObject->way(bestWay);

causes a segmentation fault.

When I protect this statement with a "if (bestObject) {...}" wrapper, I
get a primal infeasility indication. Is this the correct outcome?

I can send the 1Mb (gzipped) MPS file to anyone who has time to look
into this...

Follow-Ups:

Date: 02/20/04 16:59
By: None

Comment:
When the above if statement is added to protect against the seg-fault, I
get the following output that **seems** to indicate that there are, in
fact, feasible solutions to the integer program...

Coin0001I At line 1 NAME          8_8800_m
Coin0001I At line 2 ROWS
Coin0001I At line 2906 COLUMNS
Coin0001I At line 158153 RHS
Coin0001I At line 159485 ENDATA
Coin0002I Problem 8_8800_m has 2902 rows, 3270 columns and 306632
elements
strongBranching was changed from 5 to 10
Result 7.94204e-06 iterations 778 took 5.22 seconds - total 10.42
Sbb0013I At root node, 0 cuts changed objective from 7.94207e-06 to
7.94207e-06 in 1 passes
Sbb0014I Cut generator 0 (Gomory) - 0 row cuts (0 active), 0 column cuts
- new frequency is 100
Sbb0014I Cut generator 1 (Probing) - 0 row cuts (0 active), 0 column
cuts - new frequency is 10
Sbb0014I Cut generator 2 (Knapsack) - 0 row cuts (0 active), 0 column
cuts - new frequency is 100
Sbb0014I Cut generator 3 (OddHole) - 0 row cuts (0 active), 0 column
cuts - new frequency is 100
Sbb0010I After 0 nodes, 1 on tree, 1e+50 best solution, best possible
7.94207e-06
Sbb0012I Integer solution of 2.25644 found by rounding after 9677
iterations and 34 nodes
Sbb0012I Integer solution of 2.16737 found by rounding after 11585
iterations and 42 nodes
Sbb0012I Integer solution of 2.13768 found by rounding after 13081
iterations and 50 nodes
Sbb0012I Integer solution of 2.04861 found by rounding after 17160
iterations and 65 nodes
Sbb0012I Integer solution of 1.98923 found by rounding after 18007
iterations and 70 nodes
Sbb0012I Integer solution of 1.95954 found by rounding after 18120
iterations and 71 nodes
Sbb0010I After 100 nodes, 72 on tree, 1.95954 best solution, best
possible 7.94207e-06
Sbb0012I Integer solution of 1.88903 found by rounding after 23872
iterations and 104 nodes
Sbb0012I Integer solution of 1.85934 found by rounding after 24051
iterations and 105 nodes
Sbb0004I Integer solution of 1.82965 found after 38140 iterations and
126 nodes
Sbb0010I After 200 nodes, 112 on tree, 1.82965 best solution, best
possible 7.94207e-06
checkSolution infeas! Retrying wihout scaling.
SbbModel.cpp:3497: failed assertion `solver_->isProvenOptimal()'

Cheers!

nathan

Nathan Edwards, Ph.D.
Snr Scientist, Bioinformatics
Informatics Research
Applied Biosystems
45 West Gude Drive
Rockville, MD 20850
Phone: (240) 453-3612
Fax:   (240) 453-3324
Email: EdwardNJ at AppliedBiosystems.Com
Email: EdwardNJ at Celera.Com