Imported From: https://issues.scala-lang.org/browse/SI-7466?orig=1
Reporter: Steve Lawrence (slawrence)
Affected Versions: 2.10.1
See #7369

@paulp said (edited on May 9, 2013 2:16:53 PM UTC):
The warning is correct, because the types of Yes1/Yes2/No1/No2 are "Boolean", they are not true constants - so you haven't actually covered any of the boolean space. It goes away with Ints because the pattern matcher doesn't aspire to exhaustively checking four billion values, but it can check two.

Declare the constants so they will be seen as such, and the warning goes away:

final val Yes1 = true
final val Yes2 = true
final val No1 = false
final val No2 = false

@paulp said:
Oh, sorry "unreachable" is clearly not true, I was thinking it said "not exhaustive".

@paulp said:
However the remedy remains.

@paulp said:
Ah, I was thinking it says "not exhaustive" because that's all it says in master. So the unreachable warning is already fixed.

<console>:14: warning: match may not be exhaustive.
It would fail on the following inputs: (??, false), (??, true), (false, ??), (false, false), (false, true), (true, ??), (true, false), (true, true)
           (b1, b2) match {
           ^
defined object Test

@retronym said (edited on May 9, 2013 2:28:23 PM UTC):
Similar to #7369. After its fix, (scala/scala@62713964), the reachability warning is gone, but the spurious (or at least, over-reaching) exhaustiveness warning remains.

@paulp said:
Adriaan: when stable ids and not constants are involved, the warning should probably say "it may fail", not "it would fail".

@paulp said:
On the subject of the warning, I think a warning should remain, but it should be a lot more useful, such as suggesting "final" for any stable ids which may be constants, and perhaps not enumerating the cases but only saying it can't verify exhaustiveness without constants so either give it constants or write a default case (or mark it unchecked.)

Spent a good while studying this one and I can't think of a good way to fix it. It's not false unreachable warnings anymore, it's just saying that it can't prove exhaustiveness, and gives you the 8 ways it thinks 2 booleans can fail, which is a bit dumb.