This PR complains about -Wint-in-bool-context warning where it shouldn't. Here, in "f<1 * 1>()" the MULT_EXPR isn't really in a boolean context, but we warn anyway during overload resolution, when adding various template candidates. I believe the way to prevent this is to honor complain & tf_warning which is what this patch does. Does that seem reasonable or is there a better way?
Bootstrapped/regtested on x86_64-linux, ok for trunk? 2017-02-10 Marek Polacek <pola...@redhat.com> PR c++/79184 * cvt.c (ocp_convert): Add a sentinel against -Wint-in-bool-context if warnings shouldn't be given. * g++.dg/warn/Wint-in-bool-context-1.C: New. diff --git gcc/cp/cvt.c gcc/cp/cvt.c index ae9991a..5f4b5e3 100644 --- gcc/cp/cvt.c +++ gcc/cp/cvt.c @@ -798,7 +798,15 @@ ocp_convert (tree type, tree expr, int convtype, int flags, to the underlying type first. */ if (SCOPED_ENUM_P (intype) && (convtype & CONV_STATIC)) e = build_nop (ENUM_UNDERLYING_TYPE (intype), e); - return cp_truthvalue_conversion (e); + if (complain & tf_warning) + return cp_truthvalue_conversion (e); + else + { + /* Prevent bogus -Wint-in-bool-context warnings coming + from c_common_truthvalue_conversion down the line. */ + warning_sentinel w (warn_int_in_bool_context); + return cp_truthvalue_conversion (e); + } } converted = convert_to_integer_maybe_fold (type, e, dofold); diff --git gcc/testsuite/g++.dg/warn/Wint-in-bool-context-1.C gcc/testsuite/g++.dg/warn/Wint-in-bool-context-1.C index e69de29..117eca4 100644 --- gcc/testsuite/g++.dg/warn/Wint-in-bool-context-1.C +++ gcc/testsuite/g++.dg/warn/Wint-in-bool-context-1.C @@ -0,0 +1,16 @@ +// PR c++/79184 +// { dg-do compile } +// { dg-options "-Wint-in-bool-context" } + +enum { E = 2 }; +template <bool> void f(int) { } +template <int> void f() {} + +int +main () +{ + f<1 * 1>(); // { dg-bogus "in boolean context" } + f<1 << 1>(); // { dg-bogus "in boolean context" } + f<1 ? 3 : 2>(); // { dg-bogus "in boolean context" } + f<E>(); // { dg-bogus "in boolean context" } +} Marek