Override Failure.toString implementation

[jilen]

Avoid the very confusing failure message like

1 targets failed
ws.resolvedIvyDeps mill.api.Result$Failure
    mill.api.Result$Failure.map(Result.scala:75)
    mill.api.Result$Failure.map(Result.scala:74)
    mill.scalalib.Lib$.resolveDependencies(Lib.scala:77)
    mill.scalalib.CoursierModule$Resolver.resolveDeps(CoursierModule.scala:146)
    mill.scalalib.JavaModule.$anonfun$resolvedIvyDeps$3(JavaModule.scala:535)

[lolgab]

I don't understand what happened between Mill 0.11.10 and 0.11.11 which broke the error messages.

[lihaoyi]

Probably was broken by #3330

[lolgab]

getOrThrow is a lossy conversion in case of Failure. Isn't it better to return Result[Agg[PathRef]] in resolveDeps?

[lihaoyi]

Possibly; throwing exceptions is quick and dirty, but the current loss of error message is just a bug AFAICT and isn't fundamental to the question of exception vs result. I thought I propagated the error message in that PR, but clearly i missed something.

We could add an overload to return a Result, but I think that's orthogonal to fixing the issue at hand

[lolgab]

The bug should be fixed in #3391.
Nevertheless, we can improve the toString implementation so that in the rare case that someone calls .toString, the result has meaning. It can also be "Result.Failure($msg, $value)"

[lefou]

I think the provided patch won't improve the error message presented in the initial comment. In contrast to the merged PR #3391. We should keep the default toString of the case class.

[jilen]

I think the provided patch won't improve the error message presented in the initial comment. In contrast to the merged PR #3391. We should keep the default toString of the case class.

The toString already overrided by Failing trait. Override toString actually ensure the msg not lost at every corner case.