Make "value of" lines work with StreamSubject.

...by migrating off the deprecated no-arg `check()` overload.

Usages of that method [cause problems for "value of" lines](https://github.com/google/truth/blob/37fd8bea90c0ab4528c4c922c88fa176eb45f65b/core/src/main/java/com/google/common/truth/FailureMetadata.java#L230-L232).

Also, fill in a `typeDescriptionOverride` value for `MultisetSubject`. This is largely a no-op, but it helps under GWT/J2CL (where [we can't infer the type name](https://github.com/google/truth/blob/37fd8bea90c0ab4528c4c922c88fa176eb45f65b/core/src/main/java/com/google/common/truth/super/com/google/common/truth/Platform.java#L72)).

I left `IterableSubject` itself without a `typeDescriptionOverride`, since it has subclasses outside the package, some of which may prefer the more specific names (which are normally automatically derived from the `Subject` class's name—e.g., `AttributeListSubject` would result in `value of: attributeList.size()`). In hindsight, I guess that I could have given _direct_ usages of `IterableSubject` a `typeDescriptionOverride` (by making `StandardSubjectBuilder` pass one), but that seems like more effort than is worthwhile now that I've already tested this CL as it stands.

RELNOTES=n/a
PiperOrigin-RevId: 598964798

core/src/main/java/com/google/common/truth/IterableSubject.java

@@ -97,7 +97,15 @@ public class IterableSubject extends Subject {
    * {@link Subject#check(String, Object...) check(...)}{@code .that(actual)}.
    */
   protected IterableSubject(FailureMetadata metadata, @Nullable Iterable<?> iterable) {
-    super(metadata, iterable);
+    this(metadata, iterable, null);
+  }
+
+  /** Constructor for use by package-private callers. */
+  IterableSubject(
+      FailureMetadata metadata,
+      @Nullable Iterable<?> iterable,
+      @Nullable String typeDescriptionOverride) {
+    super(metadata, iterable, typeDescriptionOverride);
     this.actual = iterable;
   }
 

core/src/main/java/com/google/common/truth/MultisetSubject.java

@@ -31,7 +31,7 @@ public final class MultisetSubject extends IterableSubject {
   private final @Nullable Multiset<?> actual;
 
   MultisetSubject(FailureMetadata metadata, @Nullable Multiset<?> multiset) {
-    super(metadata, multiset);
+    super(metadata, multiset, /* typeDescriptionOverride= */ "multiset");
     this.actual = multiset;
   }
 

core/src/main/java/com/google/common/truth/StreamSubject.java

@@ -336,10 +336,16 @@ public void isNotEqualTo(@Nullable Object unexpected) {
 
   // TODO(user): Do we want to support comparingElementsUsing() on StreamSubject?
 
-  // TODO: b/134064106 - Migrate off no-arg check (to a direct IterableSubject constructor call?)
-  @SuppressWarnings("deprecation")
   private IterableSubject checkThatContentsList() {
-    return check().that(listSupplier.get());
+    /*
+     * Calling Subject constructors directly is usually not advisable: It does not update the
+     * metadata, so the resultant failure message might say (for example) "value of: foo" when it
+     * should say "value of: foo.size()." However, in this specific case, that's exactly what we
+     * want: We're testing the contents of the stream, so we want a "value of" line for the stream,
+     * even though we happen to implement the contents check by delegating to IterableSubject.
+     */
+    return new IterableSubject(
+        metadata, listSupplier.get(), /* typeDescriptionOverride= */ "stream");
   }
 
   private static Supplier<@Nullable List<?>> listCollector(@Nullable Stream<?> actual) {

extensions/java8/src/test/java/com/google/common/truth/StreamSubjectTest.java

@@ -158,8 +158,9 @@ public void testHasSize() throws Exception {
 
   @Test
   public void testHasSize_fails() throws Exception {
-    AssertionError unused =
+    AssertionError failure =
         expectFailure(whenTesting -> whenTesting.that(Stream.of("hello")).hasSize(2));
+    assertThat(failure).factValue("value of").isEqualTo("stream.size()");
   }
 
   @Test