Message ID | 20210913195454.3679513-1-ppalka@redhat.com |
---|---|
State | New |
Headers | show |
Series | c++: empty union member activation during constexpr [PR102163] | expand |
On 9/13/21 3:54 PM, Patrick Palka wrote: > Here, the union's constructor is defined to activate its empty data > member _M_rest, but during constexpr evaluation of this constructor the > subobject constructor call to O::O(&_M_rest, 42) produces no side > effects that actually activates the member, so the union still appears > uninitialized after the fact. This patch fixes this by faking up a > dummy MODIFY_EXPR in this situation, whose evaluation ensures the member > gets activated. > > Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK > for trunk? OK. > PR c++/102163 > > gcc/cp/ChangeLog: > > * constexpr.c (cxx_eval_call_expression): After evaluating a > constructor call for an empty union member, produce a side > effect that makes sure the member is activated. > > gcc/testsuite/ChangeLog: > > * g++.dg/cpp0x/constexpr-empty17.C: New test. > --- > gcc/cp/constexpr.c | 34 +++++++++++++++---- > .../g++.dg/cpp0x/constexpr-empty17.C | 21 ++++++++++++ > 2 files changed, 49 insertions(+), 6 deletions(-) > create mode 100644 gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C > > diff --git a/gcc/cp/constexpr.c b/gcc/cp/constexpr.c > index 7772fe62d95..40b0b80b438 100644 > --- a/gcc/cp/constexpr.c > +++ b/gcc/cp/constexpr.c > @@ -2787,12 +2787,34 @@ cxx_eval_call_expression (const constexpr_ctx *ctx, tree t, > &jump_target); > > if (DECL_CONSTRUCTOR_P (fun)) > - /* This can be null for a subobject constructor call, in > - which case what we care about is the initialization > - side-effects rather than the value. We could get at the > - value by evaluating *this, but we don't bother; there's > - no need to put such a call in the hash table. */ > - result = lval ? ctx->object : ctx->ctor; > + { > + /* This can be null for a subobject constructor call, in > + which case what we care about is the initialization > + side-effects rather than the value. We could get at the > + value by evaluating *this, but we don't bother; there's > + no need to put such a call in the hash table. */ > + result = lval ? ctx->object : ctx->ctor; > + > + if (!result && new_obj > + && TREE_CODE (new_obj) == COMPONENT_REF > + && TREE_CODE (TREE_TYPE > + (TREE_OPERAND (new_obj, 0))) == UNION_TYPE > + && is_really_empty_class (TREE_TYPE (new_obj), > + /*ignore_vptr*/false)) > + { > + /* This constructor call for an empty union member might not > + have produced a side effect that actually activated the > + member. So produce such a side effect now to ensure the > + union appears initialized. */ > + tree activate = build2 (MODIFY_EXPR, TREE_TYPE (new_obj), > + new_obj, > + build_constructor (TREE_TYPE (new_obj), > + NULL)); > + cxx_eval_constant_expression (ctx, activate, lval, > + non_constant_p, overflow_p); > + ggc_free (activate); > + } > + } > else if (VOID_TYPE_P (TREE_TYPE (res))) > result = void_node; > else > diff --git a/gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C b/gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C > new file mode 100644 > index 00000000000..9d753a3bb69 > --- /dev/null > +++ b/gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C > @@ -0,0 +1,21 @@ > +// PR c++/102163 > +// { dg-do compile { target c++11 } } > + > +struct O { > + constexpr O(int) { } > +}; > + > +union _Variadic_union { > + constexpr _Variadic_union(int __arg) : _M_rest(__arg) { } > + > + int _M_first; > + O _M_rest; > +}; > + > + > +struct _Variant_storage { > + constexpr _Variant_storage() : _M_u(42) {} > + _Variadic_union _M_u; > +}; > + > +constexpr _Variant_storage w; >
diff --git a/gcc/cp/constexpr.c b/gcc/cp/constexpr.c index 7772fe62d95..40b0b80b438 100644 --- a/gcc/cp/constexpr.c +++ b/gcc/cp/constexpr.c @@ -2787,12 +2787,34 @@ cxx_eval_call_expression (const constexpr_ctx *ctx, tree t, &jump_target); if (DECL_CONSTRUCTOR_P (fun)) - /* This can be null for a subobject constructor call, in - which case what we care about is the initialization - side-effects rather than the value. We could get at the - value by evaluating *this, but we don't bother; there's - no need to put such a call in the hash table. */ - result = lval ? ctx->object : ctx->ctor; + { + /* This can be null for a subobject constructor call, in + which case what we care about is the initialization + side-effects rather than the value. We could get at the + value by evaluating *this, but we don't bother; there's + no need to put such a call in the hash table. */ + result = lval ? ctx->object : ctx->ctor; + + if (!result && new_obj + && TREE_CODE (new_obj) == COMPONENT_REF + && TREE_CODE (TREE_TYPE + (TREE_OPERAND (new_obj, 0))) == UNION_TYPE + && is_really_empty_class (TREE_TYPE (new_obj), + /*ignore_vptr*/false)) + { + /* This constructor call for an empty union member might not + have produced a side effect that actually activated the + member. So produce such a side effect now to ensure the + union appears initialized. */ + tree activate = build2 (MODIFY_EXPR, TREE_TYPE (new_obj), + new_obj, + build_constructor (TREE_TYPE (new_obj), + NULL)); + cxx_eval_constant_expression (ctx, activate, lval, + non_constant_p, overflow_p); + ggc_free (activate); + } + } else if (VOID_TYPE_P (TREE_TYPE (res))) result = void_node; else diff --git a/gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C b/gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C new file mode 100644 index 00000000000..9d753a3bb69 --- /dev/null +++ b/gcc/testsuite/g++.dg/cpp0x/constexpr-empty17.C @@ -0,0 +1,21 @@ +// PR c++/102163 +// { dg-do compile { target c++11 } } + +struct O { + constexpr O(int) { } +}; + +union _Variadic_union { + constexpr _Variadic_union(int __arg) : _M_rest(__arg) { } + + int _M_first; + O _M_rest; +}; + + +struct _Variant_storage { + constexpr _Variant_storage() : _M_u(42) {} + _Variadic_union _M_u; +}; + +constexpr _Variant_storage w;