[Lldb-commits] [lldb] [lldb] Add unreachable after fully covered switches, avoid GCC warnings. NFC. (PR #159327)

[Martin Storsjö via lldb-commits]

mstorsjo wrote:

> Taking one as an example:
> 
> ```
>   enum class FunctionNameRepresentation {
>     eName,
>     eNameWithArgs,
>     eNameWithNoArgs
>   };
> ```
> 
> And we do have a return for each of the cases.
> 
> So it feels like a limitation in GCC's analysis, is that right or did they make a decision to warn in these cases? I can see some value in making people mark the "unused" exit path like you might a fallthrough case.

See https://github.com/llvm/llvm-project/blob/main/llvm/docs/CodingStandards.rst#don-t-use-default-labels-in-fully-covered-switches-over-enumerations - the root cause here is that GCC assumes that an enum variable still technically can have any value outside of the enum, while Clang is ok with it.

> If you have some background on it please add it to the PR description. I'll probably want to cite it again at some point :)


> Also if we added an entry to one of these enums, would the "switch doesn't cover value" warning still happen? I think it would right?

Yes, we separately use `-Wswitch` which does produce this kind of warning with both GCC and Clang if we're missing enum elements in the switch:
```
../../lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp:2181:11: warning: enumeration value 'eName' not handled in switch [-Wswitch]
 2181 |   switch (representation) {
      |           ^~~~~~~~~~~~~~
```

https://github.com/llvm/llvm-project/pull/159327