fix Issue 17807 - Spurious dead code warnings on enum and static variables.

[tgehr]

Local declarations (except variable declarations that are neither enum nor static) have no associated code, and therefore should not give dead code warnings.

[dlang-bot]

Thanks for your pull request, @tgehr! We are looking forward to reviewing it, and you should be hearing from a maintainer soon.

Some tips to help speed things up:

• smaller, focused PRs are easier to review than big ones

• try not to mix up refactoring or style changes with bug fixes or feature enhancements

• provide helpful commit messages explaining the rationale behind each change

Bear in mind that large or tricky changes may require multiple rounds of review and revision.

Please see CONTRIBUTING.md for more information.

Bugzilla references

Auto-close	Bugzilla	Description
✓	17807	Spurious dead code warnings on enum and static variables.

[ibuclaw]

Indentation

[ibuclaw]

And virtual shouldn't be needed.

[jpf91]

Can you elaborate why the compiler should not complain about static variables? They do not generate code but take up space in the data segment, so a warning seems reasonable nevertheless?

[tgehr]

Control flow passing by the static variable declaration does not do anything. Your argument justifies warnings for unused static variables, but not static variables that control flow does not reach.

[WalterBright]

@tgehr I agree.

[jpf91]

OK, but is there actually a case where a static variable declaration is not reached by control flow and the variable is still accessible from some place reached by code flow?

void main()
{
    return;

    {
        static int x;
    }
}

My line of thought goes like this: If the declaration is unreachable, all code in the scope of the declaration following the declaration is unreachable as well. The variable is only accessible in the same scope (or a nested scope) defined after the variable declaration. So, such a variable seems to be always unused/used only by dead code.

OTOH int x = void; also does not generate any code, so I don't really understand the passing by the static variable declaration does not do anything point. Or are you just arguing that these cases are still invalid, but should produce a variable unused message instead of the statement not reachable message?

EDIT: I see hasCode is used not only for the warning. So hasCode should indeed return false for static variable declarations. Maybe void initializers should also be handled then.

[WalterBright]

Trailing whitespace in expression.h is what is preventing this from passing the autotester. Please fix. I use the tool detab to do that.

[ibuclaw]

Interestingly, if you remove static here, there's a ICE in the backend. Has a bug report been raised for this?

[tgehr]

Nice catch. I couldn't find an existing issue. I have opened a new issue here: https://issues.dlang.org/show_bug.cgi?id=17834
(However, this bug will disappear once the deprecation messages are turned into compilation errors.)