Recommendations for enabling more compiler checks

Nadie ha tomado este issue todavía.

Compilers enable by default, a set of warnings they will emit on dodgy or less-than-ideal code. These defaults:

- a. change over time
- b. vary between compiler versions
- c. do not include some of the most valuable warnings that might help prevent bugs

**_This issue is designed to focus on the problem of `c` from the perspective of the latest clang++ 4.x versions._**

Enabling more warnings than compilers do by default can be an important way of catching bugs early. For example integer truncation bugs can often be detected by enabling `-Wconversion`, which is not on by default.

Enabling more warnings is very difficult to do after a project is big (e.g. https://github.com/Project-OSRM/osrm-backend/pull/4495 and https://github.com/mapnik/mapnik/issues/2907 and https://github.com/mapnik/mapnik/issues/3204). It is best not to wait and rather to start a project with aggressive warnings from the beginning.

So, the question then becomes: what is a good set of aggressive warnings to enable at the start of a project (or to try to integrate into existing projects)?

In particular:

1) 🍇 Which flags we should always enable for all code no matter what?

2) 🍊 What additional flags may be very useful in some scenarios/some code bases?

3) 🍏 For small projects where it is feasible, can we actually start with `clang++`s `-Weverything`? This, when feasible, might be ideal. How to do it?

4) 🍎 What compiler specific flags should we recommend (that only currently work for clang++ or g++)?

Empieza revisando las cuatro preguntas del issue sobre las advertencias agresivas en clang++ 4.x, incluidas -Wconversion y -Weverything. Investiga qué flags son ampliamente útiles, dependen del escenario o son específicos del compilador; después documenta una recomendación clara y explica cómo podrían aplicarla los proyectos pequeños; el trabajo estará terminado cuando las indicaciones sobre las advertencias respondan a las cuatro preguntas.

Escrito por el modelo de indexación a partir del texto del issue.