master regression test failing

[polgreen]

From travis:
Running regenerate-entry-function/test.desc [FAILED]

[martin-cs]

Hmmm... not good. I had rather hoped that the submodule set-up would be linked to a particular "last known good" master version. I don't think it would be unreasonable to do this.