Code review comment for lp:~alan-griffiths/mir/fix-1314574

> I did consider this - but wasn't sure I wanted to get involved with forking a process that already
> has a mir server running in it (and a socket pair). (Or communicating with one forked earlier after
> obtaining the socket.) I've been thinking that a small scriptable client might be good for several
> tests. But including that work in this MP seemed disproportionate.

The scriptable client would be a solution for some cases, but I prefer the flexebility of running completely custom code for clients if possible (especially if some synchronization is needed).

I think that an "exec_in_clean_process(std::function<>, std::vector<int> excluded_fds)" that does a fork followed by an FD cleaning pass (close all open fds, except the supplied list that contains fd numbers for the server socket and pipes used for synchronization (mtf::CrossProcessSync) etc) would be enough.

Anyway, I am happy as long as we have a plan for this for the short(-ish)-term