Initial CNSGA pop size is wrong in working example

[ColwynGulliford]

I ran a real test case of the latest Xopt using the dev setup. The example is a dcgun tracked with GPT, optimizing the emittance out of the gun vs. the bunch charge. Generally, it looks like the example ran successfully in parallel at NERSC:

Investigating the population stats, I see that the first population has the wrong size (121 instead of 120):

I've attached the full working example. dcgun_distgen.zip

[ColwynGulliford]

I restarted the optimization from the last pop file in the attached working example, and the initial pop size is incorrect on restart:

[roussel-ryan]

Is it possible to reproduce this error in a minimal example?

[ColwynGulliford]

mwe_tnk_with_subprocess.zip

The initial pop size looks off. The problem still optimizes...