steakhal wrote: > Hello @steakhal, I have just looked through the changes again. What is the > advantage of using checkPreCall instead of checkLocation? I would very much > appreciate some background information. Thanks for your help!
Oh yes, I should have explained. So, `checkLocation` can be triggered for all loads and store operations. Each time in the AST you have a `LValueToRValue` cast for reads, and others for write operations. This means that you can't really define a test where you have clear control of what region you wanted to dump. Using `PreCall`, you can have your own API, that does what you think, exactly when it encounters that call. This is similar to what we do in the `ExprInspection` checker, where we have analyzer debug intrinsics defined. Such as `clang_analyzer_dump(T)`. Search for it in the tests, and you will see what I'm talking about. There is one more benefit of doing this way: There is already a communication channel for the stringified descriptive name: diagnostics. One can easily capture and compare the diagnostics, like in the rest of the tests. I was happy to help as I don't expect new contributors to be really involved with our specific testing. To me what moved the needle was: well described expectation, well described and reproducible actual outcome. You even provided a fix, so I figured I'll help out with the rest. Ask me if there is anything I left out. https://github.com/llvm/llvm-project/pull/85104 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
