Open iPapatsoris opened 2 years ago
Looking closer, this seems to be beacuse the TSP example is written with the assumption that 0 represents an invalid part, while br17
has 0-cost segments that are valid. This should be fixed, but requires checking all the examples to make sure that they are all as they should be.
Describe the bug
Running tsp.cpp in the examples directory, the third hardcoded problem instance (br17 from TSPLIB) returns a solution of supposed optimal cost of 87, while the optimal solution reported by TSPLIB is 39. By copy and pasting the hardcoded input into a custom model that uses the cost-less version of the circuit constraint, I was able to get the 39, suggesting that it isn't a typo between the hardcoded data inside tsp.cpp and the one posted by TSPLIB.
To Reproduce
Compile
/examples/tsp.cpp
and run with argument2
.Gecode and Platform Configuration
Gecode version commit
bbefcea214fec798a0f5acc442581984555acd21
Ubuntu 18.04.1 LTS g++ 7.5.0