Add note about Scala 3 compiler flags

[ivan-klass]

The import-guarded feature is safe when proper compiler flags are there

[raquo]

I wish -Wvalue-discard could save us, but sadly, it won't.

Take this for example:

          val v = Var(0)
          div(
            text <-- v,
            onClick.mapTo(v.now() + 1) --> { if (v.now() < 5) v.writer }
          )

With -Wvalue-discard, it gives a warning, yes. But it warns us that we're discarding v.writer, not that we're discarding { if (v.now() < 5) v.writer }. This warning is entirely because of this bug: scala/scala3#20450. It wasn't there in earlier versions of Scala 3, and it will be fixed eventually. See also the poster's subsequent comment that explains the difference in intended behaviour between Scala 2 and 3.

[ivan-klass]

it gives a warning, yes. But it warns us that we're discarding v.writer, not that we're discarding { if (v.now() < 5) v.writer }

I'm sorry, I didn't get it.
First, there's still a warning. If there's an error in one part of the expression, it means the whole expression has an error - i.e. if there's a code considered "dirty", the entire program is "dirty".

Second, the problem in your example is indeed that we discard the writer - so it's not subscribed to clicks. The branch itself is executed - see scastie example.

[raquo]

First, there's still a warning.

My understanding is that this warning is considered a "bug" in Scala, the one I linked. Personally, I don't think it should be considered a bug, but as today, it is. I've just added a comment to that issue with my counter-argument to the proposed change, but until and unless they change their mind, this warning is sadly slated to go away in a future release of Scala 3.

the problem in your example is indeed that we discard the writer - so it's not subscribed to clicks. The branch itself is executed.

Well, yes. If the user tries to return the writer conditionally, it's because they expect to trigger it conditionally. If the writer is ignored without a warning instead, that's a problem.

BTW I looked into the history a bit, and it turns out that the wvalue-discard compiler flag was only implemented in Scala 3.3.0 – I probably wrote those docs before I switched to that version.