ignore match errors to expected types of tuple constructor elements (#24611)

fixes #24609

A tuple may have an incompatible expected type if there is a converter
match to it. So the compiler should not error when trying to match the
individual elements in the constructor to the elements of the expected
tuple type, this will be checked when the tuple is entirely constructed
anyway.

(cherry picked from commit 8d0e853e0a)

compiler/semexprs.nim

@@ -2928,7 +2928,10 @@ proc semTupleFieldsConstr(c: PContext, n: PNode, flags: TExprFlags; expectedType
       # hasEmpty/nil check is to not break existing code like
       # `const foo = [(1, {}), (2, {false})]`,
       # `const foo = if true: (0, nil) else: (1, new(int))`
-      n[i][1] = fitNode(c, expectedElemType, n[i][1], n[i][1].info)
+      let conversion = indexTypesMatch(c, expectedElemType, n[i][1].typ, n[i][1])
+      # ignore matching error, full tuple will be matched later which may call converter, see #24609
+      if conversion != nil:
+        n[i][1] = conversion
 
     if n[i][1].typ.kind == tyTypeDesc:
       localError(c.config, n[i][1].info, "typedesc not allowed as tuple field.")

tests/tuples/ttupleconverter.nim

@@ -0,0 +1,18 @@
+# issue #24609
+
+import std/options
+
+type
+  Config* = object
+    bits*: tuple[r, g, b, a: Option[int32]]
+
+# works on 2.0.8
+#
+# results in error on 2.2.0
+#   type mismatch: got 'int literal(8)' for '8' but expected 'Option[system.int32]'
+#
+converter toInt32Tuple*(t: tuple[r,g,b,a: int]): tuple[r,g,b,a: Option[int32]] =
+  (some(t.r.int32), some(t.g.int32), some(t.b.int32), some(t.a.int32))
+
+var cfg: Config
+cfg.bits = (r: 8, g: 8, b: 8, a: 16)