[llvm-dev] RFC: Modernizing our use of auto

[David Greene via llvm-dev]

Stephen Kelly via llvm-dev <llvm-dev at lists.llvm.org> writes:

> From the perspective of a relatively new contributor, it looks like
> patches are accepted or rejected based on the use of `auto` depending
> on who is doing it, and who is reviewing.

For better or for worse, that's just kind of the way things are.  Code
reviewers are human and different humans are going to have different
opinions.  I certainly have had suggestions made in the past that I
disagreed with but in the end we just have to try to understand each
other.

That said, I think it is a fool's errand to try to codify everything.
LLVM overall has done a pretty good job of providing some structure
through its coding standards while retaining flexibility for discretion.
I personally am of the opinion that "auto" is a great help and we could
use more of it but I also understand that not everyone agrees.  It seems
impossible to me to have a prescriptive rule about when "auto" is
acceptable.  There will always be exceptions.  If one can argue the
technical merits of one's patches, it can go a long way toward educating
all of us.  :)

I tend to be more in the camp of, "no rules and let reviewers decide,"
while others prefer to have more knowledge of what will be acceptable
before they submit patches.  I've always bristled at rules like, "use
C++14 except this and that feature."  If we're using C++ X, let's use
C++ X.  Again, I think LLVM strikes a pretty good balance overall.

                            -David