binvar often needs to be declared in the beginning of the script

[zoharl3]

Gurobi MILP reports infeasibility. I move

x = binvar(n, 1);

to the top of the script, and the infeasibility is resolved. It had happened quite often until I started the habit of declaring binvars early.

I don't have an example at the moment, but if it's interesting, I can try and come back in the future with something.

[johanlofberg]

That should not have any influence (except in some rare complicated mixed nonconvex quadratic + SOCP cones where there is a known bug causing an issue which I haven't fixed) so you would have to create a reproducible example