mixxxdj
User more gsl::not_null
I couldn't resist to try to get rid of the party stopping runtime checks of gsl::not_null.
This is the case when the compiler sees null checks in our own code. So I went ahead creating a wrapper header that checks this by turning the runtime checks into link-time checks. It is a drop in replacement for #include <gsl/pointers> and re-defines the Expects(a) and Ensures(a) macros as link-time assertions.
https://github.com/daschuer/mixxx/blob/not_null_link_assert/src/util/not_null_link_assert.h
With this in place, I found the changes presented in this PR, required to have no run-time null check in the constructor. Unfortunately without this wrapping header there is still a null check involved with each copy by a pedantic "Postcondition" that verifies that gsl::not_null template itself works.
Here you see the failing workflow without this PR: https://github.com/daschuer/mixxx/actions/runs/2807983581
And this is the workflow using not_null_link_assert.h and the changes from this PR: https://github.com/daschuer/mixxx/actions/runs/2807984837
I would like to include not_null_link_assert.h in a second PR to make sure all null checks are in place before creating a not_null pointer in future changes as well.
The changes of this PR should be non-controversial.
What do you think?
Merged since this is small improvements that have been open for a while.
