[nauticacom]

I've genuinely never understood this one, especially as a flat-out error instead of a warning. When could not using a variable ever be a bug?

[pcwalton]

That warning has caught bugs for me when I was copy and pasting code, or writing code on "autopilot". Usually it's pretty obvious mistakes though.

Unused variable warnings are a good idea, but making them hard errors is a language design mistake. Warnings are good because they allow programmers to quickly make changes and test things, while providing a reminder to clean things up in the end (which is why the "just use tooling that removes unused variables" response misses the point--when making quick temporary changes for debugging, you want the warnings as reminders to go back and undo the change). Additionally, warnings allow for adding helpful static analyses to the compiler over time without breaking existing code like introducing new errors does. As I recall, there were some cases in which Rust 1.0 accidentally didn't flag unused variables, which was fixable post 1.0 without breaking existing code precisely because it was a warning, not an error.

[AndyKelley]

Pretty bold to call it a language design mistake so confidently, when you have Rust users committing warning-emitting code to their source control.

Does Rust emit warnings for cached compilation units?

[ben0x539]

    foo := someDefaultValue
    if someCondition {
        foo := someMoreSpecificValue
    }

Whoops, accidentally created a fresh, unused variable in the nested scope instead of changing the value of the original variable.

I do that frighteningly often.

[int_19h]

Seems to be more an indication that Go's variable declaration syntax is just not good.

[lijogdfljk]

It happens. Hell, it even happens in Rust, i had one happen a few weeks ago that the warning lint caught. I had a rename and then added an inner scoped var of the old, previous name - but neglected to use that new var in that inner scope. Compiled fine, but was very much a bug.

Luckily though my Rust setup doesn't fail to compile with unused stuff, it just warns - and then on CI i have it reject all warnings.

I agree it's very frustrating not having a -dev flag or something less restrictive.

[macintux]

When you mistakenly used the wrong variable name instead, I suppose.

[1_player]

It is only useful in the very niche case that the incorrect variable name you're using is defined (otherwise you'd get an unknown identifier error) and the correct name you should have been using isn't used anywhere.

Would catch this:

  var a, b
  c = a + a // whoops, should have been a + b, compiler complains about unused b

Doesn't catch this:

  var a, b
  foo(a, b)
  c = a + a // whoops, should have been a + b, but still compiles

All in all, unused variables being errors is an awful feature that isn't very helpful in practice, at the cost of making experimentation a pain in the arse.

There is nothing that kills my state of flow more than having to comment a piece of code that is unreferenced, because the compiler complains, while I'm trying to hack and explore some idea.

[alpaca128]

In which case a warning would also do the trick, unless the compiler produces so many warnings that people stop caring.