[PATCH] D77866: [analyzer][CallAndMessage] Add checker options for each bug category

Gábor Horváth via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Tue Jun 2 10:58:25 PDT 2020


xazax.hun added a comment.

In D77866#2069144 <https://reviews.llvm.org/D77866#2069144>, @NoQ wrote:

> Like, i mean, the tree of packages that we currently have is a wrong abstraction.


I totally agree with this.

> The most ad-hoc approach that's better than the current situation would be twitter-like hashtags.

While this sounds really good for basic use cases I think it quickly becomes unmanageable for power users. For example, if the user wants to enable undefined behavior related checks but disable null checks how should this affect a check that is tagged with both categories? I am not sure whether using the enable/disable order is user-friendly enough. Moreover, we still need to be able to suppress each check individually (I meant checks as in one checker could implement multiple checks or diagnostics and the checker might not be the right granularity for suppression). I really like the idea of this approach but I wonder how to make this intuitive.

> A better - stricter but still sufficiently flexible - approach would be to have some actual different classifications: "language:C++ platform:unix stability:alpha severity:undefinedbehavior" etc.

I do not see a big difference between this method and the tags but I might be missing something.

I really support the idea of having a more intuitive way to handle checks and tags sound promising. But there are a lot of details we need to think about. Nevertheless, I'd love to see a more detailed RFC about the topic if someone has the time/up to it. I think this approach is probably something that the clang-tidy checks could also profit from.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D77866/new/

https://reviews.llvm.org/D77866





More information about the cfe-commits mailing list