Primal starting value on 7.0.3 about scip.jl

[matbesancon]

the 4 tests failing are:
Primal start values | 8 4 12

[rschwarz]

Testing "MIP Start" is always tricky. I think the idea of the test is to

• submit a sub-optimal solution,
• disable heuristics to prevent SCIP from finding something better,
• limiting the search to stop the process prematurely,
• and finally asserting that our submitted solution is (still) the best-known solution.

I could guess that an "improvement" in SCIP might have sneaked in a better solution here?

[matbesancon]

yes that might be

[matbesancon]

So for more details:
SCIP 7.0.2:

julia> MOI.optimize!(optimizer)
feasible solution found by completesol heuristic after 0.0 seconds, objective value 1.000000e+00
presolving:
presolving (1 rounds: 1 fast, 1 medium, 1 exhaustive):
 2 deleted vars, 1 deleted constraints, 0 added constraints, 2 tightened bounds, 0 added holes, 0 changed sides, 0 changed coefficients
 0 implications, 0 cliques
presolved problem has 0 variables (0 bin, 0 int, 0 impl, 0 cont) and 0 constraints
transformed objective value is always integral (scale: 1)
Presolving Time: 0.01


SCIP Status        : solving was interrupted [solution limit reached]
Solving Time (sec) : 0.01
Solving Nodes      : 0
Primal Bound       : +1.00000000000000e+00 (2 solutions)
Dual Bound         : +1.00000000000000e+20
Gap                : infinite

with the new bugfix:

julia> MOI.optimize!(optimizer)
feasible solution found by completesol heuristic after 0.0 seconds, objective value 1.000000e+00
presolving:
presolving (1 rounds: 1 fast, 1 medium, 1 exhaustive):
 2 deleted vars, 1 deleted constraints, 0 added constraints, 2 tightened bounds, 0 added holes, 0 changed sides, 0 changed coefficients
 0 implications, 0 cliques
transformed 1/3 original solutions to the transformed problem space
Presolving Time: 0.02

SCIP Status        : problem is solved [optimal solution found]
Solving Time (sec) : 0.02
Solving Nodes      : 0
Primal Bound       : +2.00000000000000e+00 (3 solutions)
Dual Bound         : +2.00000000000000e+00
Gap                : 0.00 %

[matbesancon]

Solution was to deactivate presolving, which changed some behavior on other constraint handlers, the
@testset "NoGoodCounter (2 binary vars)"
was counting the number of solutions, which has been reduced from 4 to 3 by the presolver between the two versions

[matbesancon]

Closed by #195