[Joao-Dionisio]

Supersedes #1228. @abb-omidi , do you mind taking a look?

The example's conslock was empty. The handler is included with needscons=False, so SCIP relies on it to lock the variables it enforces. Without the locks, presolve aggregated the relaxed y variables away (a weak dual reduction, so allowstrongdualreds=0 didn't help), and the cuts added later cut off the optimum (75 instead of 57).

Now conslock locks x and y, the allowstrongdualreds workaround is gone, and a test runs the example and checks that both objectives are 57.

[abb-omidi]

@Joao-Dionisio, do I have to check something?

[Joao-Dionisio]

@abb-omidi you don't have to, of course, just if you have the time :)

You can see here the changes I made, essentially adding variable locks.

[Joao-Dionisio]

Later on we might want to run the examples before releases as well, this is something we've been neglecting.