Deprecate numeric conversions that lose precision (e.g., Long to Double)

[som-snytt]

Int to Float, Long to Float and Long to Double are dangerous conversions and should never be done automatically.

Example:

scala> (33554435: Float).toInt
res0: Int = 33554436

Rebase #7405 and use ordinary warning system.

[som-snytt]

Maybe escalate to error under -Xsource:3.

[lrytz]

Thank you, @smarter & @som-snytt

[som-snytt]

@lrytz thanks for improving the comment. #lazy

[lrytz]

Except that it actually doesn't warn on the example I put in the comment 😄 (logged at scala/bug#10773 (comment))

[som-snytt]

[warn] /home/jenkins/workspace/scala-2.13.x-validate-main@2/src/library/scala/jdk/DoubleAccumulator.scala:110:27: Automatic conversion from Long to Double is deprecated (since 2.13.1) because it loses precision. Write `.toDouble` instead.
[warn]         x(x.length - 1) = pv
[warn]                           ^
[warn] /home/jenkins/workspace/scala-2.13.x-validate-main@2/src/library/scala/jdk/DoubleAccumulator.scala:119:27: Automatic conversion from Long to Double is deprecated (since 2.13.1) because it loses precision. Write `.toDouble` instead.
[warn]         x(x.length - 1) = pv
[warn]                           ^

@lrytz it's a long getting stuck in the data array.

Also demonstrates that, like the equals warning in RefChecks, it's in your face.

[Sciss]

Is there a compiler switch to turn the warning off?

[som-snytt]

@Sciss

scala> def f(i: Int): Float = i
                              ^
       warning: Widening conversion from Int to Float is deprecated because it loses precision. Write `.toFloat` instead.
def f(i: Int): Float

scala> :replay -Wconf:cat=deprecation&msg=conversion:s
replay> def f(i: Int): Float = i
def f(i: Int): Float


scala> 

[SethTisue]

@Sciss see #8373 for full details on -Wconf

[som-snytt]

I recommend your favorite beverage and armchair and scalac -Wconf:help.

[lrytz]

@som-snytt @SethTisue it's heart-warming to see you both get up in the morning and be thinking about -Wconf