Martin Kellogg
Martin Kellogg
Here are some thoughts about how this should be named (summarizing discussion with @msridhar, @Nargeshdb, and @mernst over a call): * Within the Called Methods Checker context, it's probably better...
If I forgot anything please feel free to add on!
@jyoo980 This is expected (though not really desirable) WPI behavior: effectively, WPI writes all "true" annotations, even those that are equivalent to the default. WPI makes an effort to avoid...
> (There could be a flag to make the checker issue a warning, on the off chance that super() made a call that assigned mySocket.) While this may be common...
> Today my program represents bundles as List. I more or less understand why `List` is disallowed, so I set out to write a small class that can own more...
This is expected behavior: `arg` is a non-owning parameter, so its `@MustCall` type is ignored. The manual section for the Must Call Checker doesn't explain this behavior, specifically, and I...
@jacek-lewandowski Sorry I missed this issue when you initially filed it! I think the issue here might be that the checker is implicitly assuming that `clusterBuilder.build()` may be able to...
There is a longer explanation of this issue in the manual here: https://checkerframework.org/manual/#type-refinement-purity, if you'd like to read more.
I've experienced false negatives related to interaction between Lombok and the Checker Framework before. Unfortunately, I don't have a clean example (all of this is based on my recollection from...
With a few small modifications, the old test I mentioned does still work with the current version of the Called Methods Checker. I opened a draft PR that should allow...