[Lldb-commits] [clang] [lldb] [Clang] Introduce OverflowBehaviorType for fine-grained overflow control (PR #148914)

[Miguel Ojeda via lldb-commits]

ojeda wrote:

> Looking at Rust's arithmetic, there is just simply _no_ mismatched bit widths allowed. Things need to be explicitly cast, so there is no ambiguity about bit width. Going all the way to this requirement feels like the furthest swing away from usability in C, especially with the goal of slowly adding `__ob_trap` types/annotations to an existing codebase.

To be honest, not having to worry all the time about that in Rust is actually quite nice, and while skimming this discussion I felt it would be the simplest approach here too. But, yeah, some kernel developers may not be thrilled about it being an error -- it is very not-C-like. It would also require building the kernel often with a newer Clang, to catch all missing casts. It also feels a bit orthogonal to the overflow behavior (but since they do not behave the same anyway, it is a fair opportunity to introduce integers with less surprises...).

https://github.com/llvm/llvm-project/pull/148914