Hi,
this one appears to be a rather simple case of missing diagnostic: in
convert_to_void we aren't calling maybe_warn_nodiscard when we strip an
INDIRECT_REF wrapping a CALL_EXPR thus we don't issue the diagnostic
that we normally provide for plain CALL_EXPRs (eg, for a func returning
a plain int). Tested x86_64-linux.
Thanks, Paolo.
//////////////////
/cp
2017-05-31 Paolo Carlini <paolo.carl...@oracle.com>
PR c++/80896
* cvt.c (convert_to_void): Possibly call maybe_warn_nodiscard
for case INDIRECT_REF too in the main switch.
/testsuite
2017-05-31 Paolo Carlini <paolo.carl...@oracle.com>
PR c++/80896
* g++.dg/cpp1z/nodiscard5.C: New.
Index: testsuite/g++.dg/cpp1z/nodiscard5.C
===================================================================
--- testsuite/g++.dg/cpp1z/nodiscard5.C (revision 0)
+++ testsuite/g++.dg/cpp1z/nodiscard5.C (working copy)
@@ -0,0 +1,7 @@
+// PR c++/80896
+// { dg-do compile { target c++11 } }
+
+int x = 42;
+[[nodiscard]] int& func() { return x; }
+
+int main() { func(); } // { dg-warning "ignoring return value" }
Index: cp/cvt.c
===================================================================
--- cp/cvt.c (revision 248728)
+++ cp/cvt.c (working copy)
@@ -1296,6 +1296,8 @@ convert_to_void (tree expr, impl_conv_void implici
&& !is_reference)
warning_at (loc, OPT_Wunused_value, "value computed is not
used");
expr = TREE_OPERAND (expr, 0);
+ if (TREE_CODE (expr) == CALL_EXPR)
+ maybe_warn_nodiscard (expr, implicit);
}
break;