Derivative Test uses different random numbers on Windows and Linux

[bradbell]

I have a case where IpRandom01 is returning different results on windows than on linux. It is actually aborting on windows because the function cannot be evaluated at the trial point, This results in a test case failing on windows and passing in linux, due to the random point being chosen on window being different. It is hard (for me) to debug on windows, so it took me a while to track this down.

I think it would be better to have a random number generator that is the same for all machines. It appears that std::rand is not such a generator; see
https://stackoverflow.com/questions/7115459/c-rand-and-srand-gets-different-output-on-different-machines

[svigerske]

Yes, that would be nicer. I don't want to change the current choice in a bugfix/minor release, but that could be something for a larger Ipopt update in the future, so keeping this open.

For reference: https://stackoverflow.com/questions/40361041/achieve-same-random-number-sequence-on-different-os-with-same-seed/40361148

[bradbell]

I think backward compatibility is important and suggest a new option that enables one to select random points that are the same for all machines. If this option has its default value, the random selection would be as before.

On the other hand, perhaps it is important to avoid the problem I had without the user having to think about it.