Should we make large (but mostly cosmetic) changes to our source code to help new contributors write clean C++?
We now have a .clang-format file to help new contributors write code with correct whitespace formatting. Many other parts of our C++ style can be enforced by clang-tidy, a stronger tool included in our toolchain that understands much more than whitespace. There is a cost, however -- clang-tidy is not nearly as good at targeting its warnings at newly-written code. To use it well(*), we will need to fix a lot of warnings about our existing code. Should we fix the existing code to make clang-tidy accept it? This would be a big change from what we did with clang-format, in which we ask that new code be clang-formatted, but we allow old code to remain unformatted unless it is touched again. Some randomly-selected examples of the kind of things clang-tidy can enforce (**): * TypeNamesAreCamelCase * "if (b == true)" should be written "if (b)" * Function prototypes should have the same parameter names as function definitions * Single-argument constructors should be "explicit" We can configure clang-tidy, just like clang-format, to be more or less strict for any given check. (*) without building up a lot of ad-hoc tooling to suppress warnings about existing code (**) partial list here: http://clang.llvm.org/extra/clang-tidy/checks/list.html. This list does not include some diagnostics that are part of clang proper and that clang-tidy can warn about, though we can also acquire a list of those and tune them for our needs.
