[llvm-dev] Implementing clang-tidy check to prefer scoped enumerations

[János Benjamin Antal via llvm-dev]

Hi,

As my first contribution to LLVM I would like to implement a clang-tidy
check that enforces the usage of scoped enumerations based on the C++ Core
Guidelines
<https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#Renum-class>.
I already submitted a review <https://reviews.llvm.org/D85697#2209272> and
now I am working on the changes based on the feedback.

I think I addressed most of the feedback, but I have some difficulties to
figure out when an unscoped enum cannot be changed to scoped enumerations.
A good example for this is already mentioned in the review: when any of the
decls or the usages (except the ones that are properly qualified by the
enum name) are in a macro.

I think another example for the unfixable case might be if any of the
decls or the usages are in a source from where the diagnostics are emitted,
i.e.: filtered out with header-filter or line-filter arguments. Can I
somehow extract that information within a derived class of ClangTidyCheck?
I think the ClangTidiyCheck::Context member variable holds that
information, but it doesn't expose that member nor the necessary
information to the derived classes.

To sum up, my questions are:

   1. Can I somehow check from a derived ClangTidyCheck if the current
   location is going to be filtered out?
   2. If the answer for the previous question is no, then how should I
   handle this situation?
   3. Do you know any other case where the decls/usages shouldn't be fixed?


Thanks for the help,
Benjamin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200818/5bed2b15/attachment.html>