NagyDonat wrote: > I'm in general open for changing the diagnostic message, but we should also > consider the friction such renames can cause. Usually tools like CodeChecker > or other frontends hash the messages, thus if it changes, the suppressions > may not work ask expected before and after the rename.
There is ample precedent for changing messages that are objectively bad, but you're right that they can break some suppressions. (I don't have an intuition about the impact/severity of this, because I'm the clang analyzer backend guy within the CodeChecker team :sweat_smile: .) ----- On the other hand we need to preserve that in the original phrasing - "undefined" is a specific term that's defined by the standard (relying on that value is _undefined behavior™_ aka may launch the nukes if the compiler wants to do so); - "garbage" represents a variety of other cases that do not trigger _undefined behavior™_ but are still problematic values that should not be used within a normal code (e.g. values that are "implementation-defined", moved-from objects etc.). https://github.com/llvm/llvm-project/pull/126596 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
