Take type parameters for this(...) constructor calls from prefix

Previously, those were inferred from arguments, but this is wrong
because we implicitly assume that the type parameters of the constructor
and the type parameters of the class are the same.

I could not find a test that fails for this. But if you look at the
-Xprint:front output of pos/i941.scala, you notice that the
inferred argument to the this(...) call was `Nothing` where
it should have been `A`.

1 files changed, 1 insertions, 1 deletions

diff --git a/src/dotty/tools/dotc/typer/Typer.scala b/src/dotty/tools/dotc/typer/Typer.scala
index fac24e440..3d36905f2 100644
--- a/src/dotty/tools/dotc/typer/Typer.scala
+++ b/src/dotty/tools/dotc/typer/Typer.scala
@@ -1677,7 +1677,7 @@ class Typer extends Namer with TypeAssigner with Applications with Implicits wit
           if (pt.isInstanceOf[PolyProto]) tree
           else {
             var typeArgs = tree match {
-              case Select(New(tpt), nme.CONSTRUCTOR) => tpt.tpe.dealias.argTypesLo
+              case Select(qual, nme.CONSTRUCTOR) => qual.tpe.widenDealias.argTypesLo
               case _ => Nil
             }
             if (typeArgs.isEmpty) typeArgs = constrained(poly, tree)._2