coverage: 99.797% (-0.03%) from 99.822%
when pulling ce64fc2 on james-powis:symlink-dry
into 3cb3a20 on adrienverge:master.

This pull request is not reviewable as is: function-local code style not respected, quotes, inclusion of yamllint/.cli.py.swp (congrats for using Vim by the way 😉), remains of your tests like maxDiff = None, independant changes like set() that should have their own commit / review, etc.

But I understand that it's intended, and you want to start a discussion about symbolic links.
To be honest I have a hard time understanding your commit message (e.g. the "strategies"), and what "this solution" consists in. You should describe it.

And my current recommendation is to skip them entirely.

I don't think all yamllint users would agree, this would be a breaking change.

The goal of this PR was not to actually get this accepted, it was to demonstrate one of the few problems symlinks present. Especially around excluding... Do you apply exclude the symlink or the target, or both, none? How do you handle the reporting of the lint issue, do you report the link, or the target?. How do you prevent 100 million symlinks to the same file not result in the running of yamllint 100 million times.

Ultimately, if the file should be linted, it should be linted at the source not the one of potentially numerous links to it. By excluding the symlink entirely, the users of yamllint have the option to target the target via include if so desired file. But that is just my opinion.