From: Jason Merrill <jason@redhat.com>
To: Patrick Palka <ppalka@redhat.com>
Cc: gcc-patches@gcc.gnu.org
Subject: Re: [PATCH 1/2] c++: improve "NTTP argument considered unused" fix [PR53164, PR105848]
Date: Fri, 31 Mar 2023 23:50:01 -0400 [thread overview]
Message-ID: <74ccb540-b964-fc2f-d52d-c5832b5df563@redhat.com> (raw)
In-Reply-To: <7c047507-896a-5fbf-f1ce-14a1377b0b9f@idea>
On 3/31/23 15:06, Patrick Palka wrote:
> On Fri, 31 Mar 2023, Jason Merrill wrote:
>
>> On 3/31/23 11:55, Patrick Palka wrote:
>>> On Fri, 31 Mar 2023, Patrick Palka wrote:
>>>
>>>> On Thu, 30 Mar 2023, Jason Merrill wrote:
>>>>
>>>>> On 3/30/23 14:53, Patrick Palka wrote:
>>>>>> On Wed, 29 Mar 2023, Jason Merrill wrote:
>>>>>>
>>>>>>> On 3/27/23 09:30, Patrick Palka wrote:
>>>>>>>> On Thu, 23 Mar 2023, Patrick Palka wrote:
>>>>>>>>
>>>>>>>>> r13-995-g733a792a2b2e16 worked around the problem of
>>>>>>>>> FUNCTION_DECL
>>>>>>>>> template arguments not always getting marked as odr-used by
>>>>>>>>> redundantly
>>>>>>>>> calling mark_used on the substituted ADDR_EXPR callee of a
>>>>>>>>> CALL_EXPR.
>>>>>>>>> This is just a narrow workaround however, since using a
>>>>>>>>> FUNCTION_DECL
>>>>>>>>> as
>>>>>>>>> a template argument alone should constitutes an odr-use; we
>>>>>>>>> shouldn't
>>>>>>>>> need to subsequently e.g. call the function or take its address.
>>>>>>>
>>>>>>> Agreed. But why didn't we already wrap it in an ADDR_EXPR? Even
>>>>>>> for
>>>>>>> reference tparms convert_nontype_argument should do that.
>>>>>>
>>>>>> Indeed we do, the commit message was just rather sloppy/inaccurate...
>>>>>> I'll try to correct it.
>>>>>>
>>>>>>>
>>>>>>>>> This patch fixes this in a more general way at template
>>>>>>>>> specialization
>>>>>>>>> time by walking the template arguments of the specialization and
>>>>>>>>> calling
>>>>>>>>> mark_used on all entities used within. As before, the call to
>>>>>>>>> mark_used
>>>>>>>>> as it worst a no-op, but it compensates for the situation where
>>>>>>>>> we end
>>>>>>>>> up
>>>>>>>>> forming a specialization from a template context in which
>>>>>>>>> mark_used is
>>>>>>>>> inhibited. Another approach would be to call mark_used whenever
>>>>>>>>> we
>>>>>>>>> substitute a TEMPLATE_PARM_INDEX, but that would result in many
>>>>>>>>> more
>>>>>>>>> redundant calls to mark_used compared to this approach.
>>>>>>>>>
>>>>>>>>> Bootstrapped and regtested on x86_64-pc-linux-gnu, does this
>>>>>>>>> look OK
>>>>>>>>> for
>>>>>>>>> trunk?
>>>>>>>>>
>>>>>>>>> PR c++/53164
>>>>>>>>> PR c++/105848
>>>>>>>>>
>>>>>>>>> gcc/cp/ChangeLog:
>>>>>>>>>
>>>>>>>>> * pt.cc (instantiate_class_template): Call
>>>>>>>>> mark_template_arguments_used.
>>>>>>>>> (tsubst_copy_and_build) <case CALL_EXPR>: Revert r13-995
>>>>>>>>> change.
>>>>>>>>> (mark_template_arguments_used): Define.
>>>>>>>>> (instantiate_template): Call mark_template_arguments_used.
>>>>>>>>>
>>>>>>>>> gcc/testsuite/ChangeLog:
>>>>>>>>>
>>>>>>>>> * g++.dg/template/fn-ptr3a.C: New test.
>>>>>>>>> * g++.dg/template/fn-ptr4.C: New test.
>>>>>>>>> ---
>>>>>>>>> gcc/cp/pt.cc | 51
>>>>>>>>> ++++++++++++++++--------
>>>>>>>>> gcc/testsuite/g++.dg/template/fn-ptr3a.C | 25 ++++++++++++
>>>>>>>>> gcc/testsuite/g++.dg/template/fn-ptr4.C | 14 +++++++
>>>>>>>>> 3 files changed, 74 insertions(+), 16 deletions(-)
>>>>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr3a.C
>>>>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr4.C
>>>>>>>>>
>>>>>>>>> diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
>>>>>>>>> index 7e4a8de0c8b..9b3cc33331c 100644
>>>>>>>>> --- a/gcc/cp/pt.cc
>>>>>>>>> +++ b/gcc/cp/pt.cc
>>>>>>>>> @@ -220,6 +220,7 @@ static tree make_argument_pack (tree);
>>>>>>>>> static tree enclosing_instantiation_of (tree tctx);
>>>>>>>>> static void instantiate_body (tree pattern, tree args, tree
>>>>>>>>> d, bool
>>>>>>>>> nested);
>>>>>>>>> static tree maybe_dependent_member_ref (tree, tree,
>>>>>>>>> tsubst_flags_t,
>>>>>>>>> tree);
>>>>>>>>> +static void mark_template_arguments_used (tree);
>>>>>>>>> /* Make the current scope suitable for access checking
>>>>>>>>> when we
>>>>>>>>> are
>>>>>>>>> processing T. T can be FUNCTION_DECL for instantiated
>>>>>>>>> function
>>>>>>>>> @@ -12142,6 +12143,9 @@ instantiate_class_template (tree type)
>>>>>>>>> cp_unevaluated_operand = 0;
>>>>>>>>> c_inhibit_evaluation_warnings = 0;
>>>>>>>>> }
>>>>>>>>> +
>>>>>>>>> + mark_template_arguments_used (INNERMOST_TEMPLATE_ARGS
>>>>>>>>> (args));
>>>>>>>>> +
>>>>>>>>> /* Use #pragma pack from the template context. */
>>>>>>>>> saved_maximum_field_alignment = maximum_field_alignment;
>>>>>>>>> maximum_field_alignment = TYPE_PRECISION (pattern);
>>>>>>>>> @@ -21173,22 +21177,10 @@ tsubst_copy_and_build (tree t,
>>>>>>>>> }
>>>>>>>>> /* Remember that there was a reference to this entity.
>>>>>>>>> */
>>>>>>>>> - if (function != NULL_TREE)
>>>>>>>>> - {
>>>>>>>>> - tree inner = function;
>>>>>>>>> - if (TREE_CODE (inner) == ADDR_EXPR
>>>>>>>>> - && TREE_CODE (TREE_OPERAND (inner, 0)) ==
>>>>>>>>> FUNCTION_DECL)
>>>>>>>>> - /* We should already have called mark_used when taking
>>>>>>>>> the
>>>>>>>>> - address of this function, but do so again anyway to
>>>>>>>>> make
>>>>>>>>> - sure it's odr-used: at worst this is a no-op, but if
>>>>>>>>> we
>>>>>>>>> - obtained this FUNCTION_DECL as part of ahead-of-time
>>>>>>>>> overload
>>>>>>>>> - resolution then that call to mark_used wouldn't have
>>>>>>>>> marked
>>>>>>>>> it
>>>>>>>>> - odr-used yet (53164). */
>>>>>>>>> - inner = TREE_OPERAND (inner, 0);
>>>>>>>>> - if (DECL_P (inner)
>>>>>>>>> - && !mark_used (inner, complain) && !(complain &
>>>>>>>>> tf_error))
>>>>>>>>> - RETURN (error_mark_node);
>>>>>>>>> - }
>>>>>>>>> + if (function != NULL_TREE
>>>>>>>>> + && DECL_P (function)
>>>>>>>>> + && !mark_used (function, complain) && !(complain &
>>>>>>>>> tf_error))
>>>>>>>>> + RETURN (error_mark_node);
>>>>>>>>> if (!maybe_fold_fn_template_args (function, complain))
>>>>>>>>> return error_mark_node;
>>>>>>>>> @@ -21883,6 +21875,31 @@ check_instantiated_args (tree tmpl,
>>>>>>>>> tree
>>>>>>>>> args,
>>>>>>>>> tsubst_flags_t complain)
>>>>>>>>> return result;
>>>>>>>>> }
>>>>>>>>> +/* Call mark_used on each entity within the template
>>>>>>>>> arguments
>>>>>>>>> ARGS of
>>>>>>>>> some
>>>>>>>>> + template specialization, to ensure that each such entity is
>>>>>>>>> considered
>>>>>>>>> + odr-used regardless of whether the specialization was first
>>>>>>>>> formed
>>>>>>>>> in
>>>>>>>>> a
>>>>>>>>> + template context.
>>>>>>>>> +
>>>>>>>>> + This function assumes push_to_top_level has been called
>>>>>>>>> beforehand,
>>>>>>>>> and
>>>>>>>>> + that processing_template_decl has been set iff the template
>>>>>>>>> arguments
>>>>>>>>> + are dependent. */
>>>>>>>>> +
>>>>>>>>> +static void
>>>>>>>>> +mark_template_arguments_used (tree args)
>>>>>>>>> +{
>>>>>>>>> + gcc_checking_assert (TMPL_ARGS_DEPTH (args) == 1);
>>>>>>>>> +
>>>>>>>>> + if (processing_template_decl)
>>>>>>>>> + return;
>>>>>>>>> +
>>>>>>>>> + auto mark_used_r = [](tree *tp, int *, void *) {
>>>>>>>>> + if (DECL_P (*tp))
>>>>>>>>> + mark_used (*tp, tf_none);
>>>>>>>>> + return NULL_TREE;
>>>>>>>>> + };
>>>>>>>>> + cp_walk_tree_without_duplicates (&args, mark_used_r,
>>>>>>>>> nullptr);
>>>>>>>>
>>>>>>>> Here's v2 which optimizes this function by not walking if
>>>>>>>> !PRIMARY_TEMPLATE_P (since in that case the innermost arguments
>>>>>>>> have
>>>>>>>> already been walked when instantiating the context), and walking
>>>>>>>> only
>>>>>>>> non-type arguments:
>>>>>>>>
>>>>>>>> -- >8
>>>>>>>>
>>>>>>>> Subject: [PATCH] c++: improve "NTTP argument considered unused"
>>>>>>>> fix
>>>>>>>> [PR53164,
>>>>>>>> PR105848]
>>>>>>>>
>>>>>>>> PR c++/53164
>>>>>>>> PR c++/105848
>>>>>>>>
>>>>>>>> gcc/cp/ChangeLog:
>>>>>>>>
>>>>>>>> * pt.cc (instantiate_class_template): Call
>>>>>>>> mark_template_arguments_used.
>>>>>>>> (tsubst_copy_and_build) <case CALL_EXPR>: Revert r13-995
>>>>>>>> change.
>>>>>>>> (mark_template_arguments_used): Define.
>>>>>>>> (instantiate_template): Call mark_template_arguments_used.
>>>>>>>>
>>>>>>>> gcc/testsuite/ChangeLog:
>>>>>>>>
>>>>>>>> * g++.dg/template/fn-ptr3a.C: New test.
>>>>>>>> * g++.dg/template/fn-ptr4.C: New test.
>>>>>>>> ---
>>>>>>>> gcc/cp/pt.cc | 57
>>>>>>>> +++++++++++++++++-------
>>>>>>>> gcc/testsuite/g++.dg/template/fn-ptr3a.C | 25 +++++++++++
>>>>>>>> gcc/testsuite/g++.dg/template/fn-ptr4.C | 14 ++++++
>>>>>>>> 3 files changed, 80 insertions(+), 16 deletions(-)
>>>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr3a.C
>>>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr4.C
>>>>>>>>
>>>>>>>> diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
>>>>>>>> index 3bb98ebeac1..a87366bb616 100644
>>>>>>>> --- a/gcc/cp/pt.cc
>>>>>>>> +++ b/gcc/cp/pt.cc
>>>>>>>> @@ -220,6 +220,7 @@ static tree make_argument_pack (tree);
>>>>>>>> static tree enclosing_instantiation_of (tree tctx);
>>>>>>>> static void instantiate_body (tree pattern, tree args, tree d,
>>>>>>>> bool
>>>>>>>> nested);
>>>>>>>> static tree maybe_dependent_member_ref (tree, tree,
>>>>>>>> tsubst_flags_t,
>>>>>>>> tree);
>>>>>>>> +static void mark_template_arguments_used (tree, tree);
>>>>>>>> /* Make the current scope suitable for access checking when
>>>>>>>> we are
>>>>>>>> processing T. T can be FUNCTION_DECL for instantiated
>>>>>>>> function
>>>>>>>> @@ -12152,6 +12153,9 @@ instantiate_class_template (tree type)
>>>>>>>> cp_unevaluated_operand = 0;
>>>>>>>> c_inhibit_evaluation_warnings = 0;
>>>>>>>> }
>>>>>>>> +
>>>>>>>> + mark_template_arguments_used (templ, CLASSTYPE_TI_ARGS (type));
>>>>>>>> +
>>>>>>>> /* Use #pragma pack from the template context. */
>>>>>>>> saved_maximum_field_alignment = maximum_field_alignment;
>>>>>>>> maximum_field_alignment = TYPE_PRECISION (pattern);
>>>>>>>> @@ -21192,22 +21196,10 @@ tsubst_copy_and_build (tree t,
>>>>>>>> }
>>>>>>>> /* Remember that there was a reference to this entity.
>>>>>>>> */
>>>>>>>> - if (function != NULL_TREE)
>>>>>>>> - {
>>>>>>>> - tree inner = function;
>>>>>>>> - if (TREE_CODE (inner) == ADDR_EXPR
>>>>>>>> - && TREE_CODE (TREE_OPERAND (inner, 0)) ==
>>>>>>>> FUNCTION_DECL)
>>>>>>>> - /* We should already have called mark_used when taking
>>>>>>>> the
>>>>>>>> - address of this function, but do so again anyway to
>>>>>>>> make
>>>>>>>> - sure it's odr-used: at worst this is a no-op, but if
>>>>>>>> we
>>>>>>>> - obtained this FUNCTION_DECL as part of ahead-of-time
>>>>>>>> overload
>>>>>>>> - resolution then that call to mark_used wouldn't have
>>>>>>>> marked
>>>>>>>> it
>>>>>>>> - odr-used yet (53164). */
>>>>>>>> - inner = TREE_OPERAND (inner, 0);
>>>>>>>> - if (DECL_P (inner)
>>>>>>>> - && !mark_used (inner, complain) && !(complain &
>>>>>>>> tf_error))
>>>>>>>> - RETURN (error_mark_node);
>>>>>>>> - }
>>>>>>>> + if (function != NULL_TREE
>>>>>>>> + && DECL_P (function)
>>>>>>>> + && !mark_used (function, complain) && !(complain &
>>>>>>>> tf_error))
>>>>>>>> + RETURN (error_mark_node);
>>>>>>>> if (!maybe_fold_fn_template_args (function, complain))
>>>>>>>> return error_mark_node;
>>>>>>>> @@ -21902,6 +21894,37 @@ check_instantiated_args (tree tmpl, tree
>>>>>>>> args,
>>>>>>>> tsubst_flags_t complain)
>>>>>>>> return result;
>>>>>>>> }
>>>>>>>> +/* Call mark_used on each entity within the non-type template
>>>>>>>> arguments
>>>>>>>> in
>>>>>>>> + ARGS for a specialization of TMPL, to ensure that each such
>>>>>>>> entity
>>>>>>>> is
>>>>>>>> + considered odr-used regardless of whether the specialization
>>>>>>>> was
>>>>>>>> first
>>>>>>>> + formed in a template context.
>>>>>>>> +
>>>>>>>> + This function assumes push_to_top_level has been called
>>>>>>>> beforehand,
>>>>>>>> and
>>>>>>>> + that processing_template_decl is set iff the template
>>>>>>>> arguments are
>>>>>>>> + dependent. */
>>>>>>>> +
>>>>>>>> +static void
>>>>>>>> +mark_template_arguments_used (tree tmpl, tree args)
>>>>>>>> +{
>>>>>>>> + /* It suffices to do this only when fully specializing a
>>>>>>>> primary
>>>>>>>> template. */
>>>>>>>> + if (processing_template_decl || !PRIMARY_TEMPLATE_P (tmpl))
>>>>>>>> + return;
>>>>>>>> +
>>>>>>>> + /* We already marked outer arguments when specializing the
>>>>>>>> context.
>>>>>>>> */
>>>>>>>> + args = INNERMOST_TEMPLATE_ARGS (args);
>>>>>>>> +
>>>>>>>> + for (tree arg : tree_vec_range (args))
>>>>>>>> + if (!TYPE_P (arg))
>>>>>>>> + {
>>>>>>>> + auto mark_used_r = [](tree *tp, int *, void *) {
>>>>>>>> + if (DECL_P (*tp))
>>>>>>>> + mark_used (*tp, tf_none);
>>>>>>>> + return NULL_TREE;
>>>>>>>> + };
>>>>>>>> + cp_walk_tree_without_duplicates (&args, mark_used_r, nullptr);
>>>>>>>
>>>>>>> Do we really need to walk into the args?
>>>>>>
>>>>>> Hmm, I think we might need to some sort for walking for class NTTP
>>>>>> arguments since they could be nested CONSTRUCTORs. But for non-class
>>>>>> arguments we know that an entity will only appear as a pointer- or
>>>>>> reference-to-function/variable, so we could pattern match for such a
>>>>>> structure directly. And doing so should be preferable because
>>>>>> cp_walk_tree is relatively expensive and this is a relatively hot code
>>>>>> path. So maybe the following would be preferable?
>>>>>>
>>>>>> -- >8 --
>>>>>>
>>>>>> Subject: [PATCH] c++: improve "NTTP argument considered unused" fix
>>>>>> [PR53164,
>>>>>> PR105848]
>>>>>>
>>>>>> r13-995-g733a792a2b2e16 worked around the problem of (pointer to)
>>>>>> function NTTP arguments not always getting marked as odr-used by
>>>>>> redundantly calling mark_used on the substituted ADDR_EXPR callee of a
>>>>>> CALL_EXPR. That is just a narrow workaround however, since it assumes
>>>>>> the function is later called. But the use as a template argument
>>>>>> alone
>>>>>> should constitute an odr-use of the function (since template arguments
>>>>>> are an evaluated context, and we need its address); we shouldn't need
>>>>>> to
>>>>>> subsequently call or otherwise use the NTTP argument.
>>>>>>
>>>>>> This patch fixes this in a more general way at template specialization
>>>>>> time by walking the template arguments of the specialization and
>>>>>> calling
>>>>>> mark_used on all entities used within. As before, the call to
>>>>>> mark_used
>>>>>> as it worst a no-op, but it compensates for the situation where we end
>>>>>> up forming a specialization in a template context in which mark_used
>>>>>> is
>>>>>> inhibited.
>>>>>>
>>>>>> Another approach would be to call mark_used whenever we substitute a
>>>>>> TEMPLATE_PARM_INDEX, but that would result in many more redundant
>>>>>> calls
>>>>>> to mark_used compared to this approach. And as the second testcase
>>>>>> below illustrates, we also need to walk C++20 class NTTP arguments
>>>>>> which
>>>>>> can in theory be large and thus expensive to walk repeatedly. The
>>>>>> change to invalid_tparm_referent_p is needed to avoid bogusly
>>>>>> rejecting
>>>>>> the testcase's class NTTP arguments containing function pointers.
>>>>>>
>>>>>> (The third testcase is unrelated to this fix, but it helped rule out
>>>>>> an
>>>>>> earlier approach I was considering and it seems we don't have existing
>>>>>> test coverage for this situation.)
>>>>>>
>>>>>> PR c++/53164
>>>>>> PR c++/105848
>>>>>>
>>>>>> gcc/cp/ChangeLog:
>>>>>>
>>>>>> * pt.cc (invalid_tparm_referent_p): Accept ADDR_EXPR of
>>>>>> FUNCTION_DECL.
>>>>>> (instantiate_class_template): Call
>>>>>> mark_template_arguments_used.
>>>>>> (tsubst_copy_and_build) <case CALL_EXPR>: Revert r13-995
>>>>>> change.
>>>>>> (mark_template_arguments_used): Define.
>>>>>> (instantiate_template): Call mark_template_arguments_used.
>>>>>>
>>>>>> gcc/testsuite/ChangeLog:
>>>>>>
>>>>>> * g++.dg/template/fn-ptr3a.C: New test.
>>>>>> * g++.dg/template/fn-ptr3b.C: New test.
>>>>>> * g++.dg/template/fn-ptr4.C: New test.
>>>>>> ---
>>>>>> gcc/cp/pt.cc | 78
>>>>>> ++++++++++++++++++------
>>>>>> gcc/testsuite/g++.dg/template/fn-ptr3a.C | 25 ++++++++
>>>>>> gcc/testsuite/g++.dg/template/fn-ptr3b.C | 28 +++++++++
>>>>>> gcc/testsuite/g++.dg/template/fn-ptr4.C | 14 +++++
>>>>>> 4 files changed, 127 insertions(+), 18 deletions(-)
>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr3a.C
>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr3b.C
>>>>>> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr4.C
>>>>>>
>>>>>> diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
>>>>>> index e514a277872..01ab220320c 100644
>>>>>> --- a/gcc/cp/pt.cc
>>>>>> +++ b/gcc/cp/pt.cc
>>>>>> @@ -220,6 +220,7 @@ static tree make_argument_pack (tree);
>>>>>> static tree enclosing_instantiation_of (tree tctx);
>>>>>> static void instantiate_body (tree pattern, tree args, tree d, bool
>>>>>> nested);
>>>>>> static tree maybe_dependent_member_ref (tree, tree, tsubst_flags_t,
>>>>>> tree);
>>>>>> +static void mark_template_arguments_used (tree, tree);
>>>>>> /* Make the current scope suitable for access checking when we
>>>>>> are
>>>>>> processing T. T can be FUNCTION_DECL for instantiated function
>>>>>> @@ -7142,12 +7143,13 @@ invalid_tparm_referent_p (tree type, tree
>>>>>> expr,
>>>>>> tsubst_flags_t complain)
>>>>>> decl = TREE_OPERAND (decl, 0);
>>>>>> }
>>>>>> - if (!VAR_P (decl))
>>>>>> + if (!VAR_OR_FUNCTION_DECL_P (decl))
>>>>>> {
>>>>>> if (complain & tf_error)
>>>>>> error_at (cp_expr_loc_or_input_loc (expr),
>>>>>> "%qE is not a valid template argument of type
>>>>>> %qT "
>>>>>> - "because %qE is not a variable", expr, type,
>>>>>> decl);
>>>>>> + "because %qE is not a variable or function",
>>>>>> + expr, type, decl);
>>>>>> return true;
>>>>>> }
>>>>>> else if (cxx_dialect < cxx11 && !DECL_EXTERNAL_LINKAGE_P
>>>>>> (decl))
>>>>>> @@ -12152,6 +12154,9 @@ instantiate_class_template (tree type)
>>>>>> cp_unevaluated_operand = 0;
>>>>>> c_inhibit_evaluation_warnings = 0;
>>>>>> }
>>>>>> +
>>>>>> + mark_template_arguments_used (templ, CLASSTYPE_TI_ARGS (type));
>>>>>> +
>>>>>> /* Use #pragma pack from the template context. */
>>>>>> saved_maximum_field_alignment = maximum_field_alignment;
>>>>>> maximum_field_alignment = TYPE_PRECISION (pattern);
>>>>>> @@ -21192,22 +21197,10 @@ tsubst_copy_and_build (tree t,
>>>>>> }
>>>>>> /* Remember that there was a reference to this entity. */
>>>>>> - if (function != NULL_TREE)
>>>>>> - {
>>>>>> - tree inner = function;
>>>>>> - if (TREE_CODE (inner) == ADDR_EXPR
>>>>>> - && TREE_CODE (TREE_OPERAND (inner, 0)) ==
>>>>>> FUNCTION_DECL)
>>>>>> - /* We should already have called mark_used when taking
>>>>>> the
>>>>>> - address of this function, but do so again anyway to
>>>>>> make
>>>>>> - sure it's odr-used: at worst this is a no-op, but if
>>>>>> we
>>>>>> - obtained this FUNCTION_DECL as part of ahead-of-time
>>>>>> overload
>>>>>> - resolution then that call to mark_used wouldn't have
>>>>>> marked
>>>>>> it
>>>>>> - odr-used yet (53164). */
>>>>>> - inner = TREE_OPERAND (inner, 0);
>>>>>> - if (DECL_P (inner)
>>>>>> - && !mark_used (inner, complain) && !(complain &
>>>>>> tf_error))
>>>>>> - RETURN (error_mark_node);
>>>>>> - }
>>>>>> + if (function != NULL_TREE
>>>>>> + && DECL_P (function)
>>>>>> + && !mark_used (function, complain) && !(complain &
>>>>>> tf_error))
>>>>>> + RETURN (error_mark_node);
>>>>>> if (!maybe_fold_fn_template_args (function, complain))
>>>>>> return error_mark_node;
>>>>>> @@ -21902,6 +21895,53 @@ check_instantiated_args (tree tmpl, tree
>>>>>> args,
>>>>>> tsubst_flags_t complain)
>>>>>> return result;
>>>>>> }
>>>>>> +/* Call mark_used on each entity within the non-type template
>>>>>> arguments
>>>>>> in
>>>>>> + ARGS for a specialization of TMPL, to ensure that each such entity
>>>>>> is
>>>>>> + considered odr-used regardless of whether the specialization was
>>>>>> first
>>>>>> + formed in a template context.
>>>>>> +
>>>>>> + This function assumes push_to_top_level has been called
>>>>>> beforehand, and
>>>>>> + that processing_template_decl is set iff the template arguments
>>>>>> are
>>>>>> + dependent. */
>>>>>> +
>>>>>> +static void
>>>>>> +mark_template_arguments_used (tree tmpl, tree args)
>>>>>> +{
>>>>>> + /* It suffices to do this only when fully specializing a primary
>>>>>> template. */
>>>>>> + if (processing_template_decl || !PRIMARY_TEMPLATE_P (tmpl))
>>>>>> + return;
>>>>>> +
>>>>>> + /* We already marked outer arguments when specializing the context.
>>>>>> */
>>>>>> + args = INNERMOST_TEMPLATE_ARGS (args);
>>>>>> +
>>>>>> + for (tree arg : tree_vec_range (args))
>>>>>> + {
>>>>>> + /* A (pointer/reference to) function or variable NTTP argument.
>>>>>> */
>>>>>> + if (TREE_CODE (arg) == ADDR_EXPR
>>>>>> + || TREE_CODE (arg) == INDIRECT_REF)
>>>>>> + {
>>>>>> + while (TREE_CODE (arg) == ADDR_EXPR
>>>>>> + || REFERENCE_REF_P (arg)
>>>>>> + || CONVERT_EXPR_P (arg))
>>>>>> + arg = TREE_OPERAND (arg, 0);
>>>>>> + if (DECL_P (arg))
>>>>>> + mark_used (arg, tf_none);
>>>>>
>>>>> Passing tf_none and also ignoring the return value needs a comment
>>>>> justifying
>>>>> why you assume it can't fail.
>>>>
>>>> *nod* Since these calls to mark_used ought to be at worst redundant, it
>>>> should be safe to even assert
>>>>
>>>> bool ok = mark_used (arg, tf_none);
>>>> gcc_checking_assert (ok || seen_error ());
>>>>
>>>>>
>>>>>> + }
>>>>>> + /* A class NTTP argument. */
>>>>>> + else if (VAR_P (arg)
>>>>>> + && DECL_NTTP_OBJECT_P (arg))
>>>>>
>>>>> Maybe avoid doing this multiple times for the same NTTP object by gating
>>>>> it on
>>>>> DECL_ODR_USED (arg)?
>>>>
>>>> Looks like these VAR_DECLs don't have DECL_LANG_SPECIFIC, so we can't
>>>> inspect DECL_ODR_USED for them..
>>>>
>>>> It occurred to me that instantiate_class_template is perhaps the wrong
>>>> place to do this marking; it'd be more appropriate to do it in
>>>> lookup_template_class when forming the type specialization rather than
>>>> when instantiating its definition. So in order to handle all template
>>>> kinds consistently, I think we should do the marking in either
>>>> instantiate_class_template and instantiate_body (i.e. at instantiation
>>>> time), or in lookup_template_class and instantiate_template (i.e. at
>>>> specialization time), but currently the patch does it in
>>>> instantiate_class_template and instantiate_template which sounds
>>>> consistent but it's sadly not.
>>>> However, lookup_template_class doesn't seem to call push_to_top_level at
>>>> all, so if we want to call mark_used uninhibited on the template
>>>> arguments from there I think we need to add a flag to mark_used that
>>>> simply disables its
>>>>
>>>> if (processing_template_decl || in_template_function ())
>>>> return true;
>>>>
>>>> early exit. IIUC the purpose of this early exit is to avoid
>>>> instantiating a template due to a use from an uninstantiated template,
>>>> under the assumption that we'll call mark_used again at instantiation
>>>> time for that use, but in this case it seems we need to allow such
>>>> ahead-of-time instantiation since we don't reprocess the template
>>>> arguments of a non-dependent specialization again at instantiation time.
>>>>
>>>> This is quite the can of worms :/
>>
>> Yes, I think I'd prefer to move toward the
>> instantiate_class_template/instantiate_body consistency, to avoid unnecessary
>> marking in uninstantiated templates.
> Aha, makes sense, that approach seems to align best with our existing
> efforts to avoid instantiation from uninstantiated templates. Like so?
> Bootstrapped and regtested on x86_64-pc-linux-gnu.
OK.
> -- >8 --
>
> Subject: [PATCH] c++: improve "NTTP argument considered unused" fix [PR53164,
> PR105848]
>
> r13-995-g733a792a2b2e16 worked around the problem of (pointer to)
> function NTTP arguments not always getting marked as odr-used, by
> redundantly calling mark_used on the substituted ADDR_EXPR callee of a
> CALL_EXPR. That is just a narrow workaround however, since it assumes
> the function is later called. But the use as a template argument alone
> should constitute an odr-use of the function (since template arguments
> are an evaluated context, and we're really passing its address); we
> shouldn't need to subsequently call or otherwise use the NTTP argument.
>
> This patch fixes this in a more general way at template instantiation
> time by walking the template arguments of the specialization and calling
> mark_used on all entities used within. As before, the call to mark_used
> as it worst a no-op, but it compensates for the situation where we end
> up forming a specialization in a template context in which mark_used is
> inhibited.
>
> Another approach would be to call mark_used whenever we substitute a
> TEMPLATE_PARM_INDEX, but that would result in many more redundant calls
> to mark_used compared to this approach. And as the second testcase
> below illustrates, we also need to walk C++20 class NTTP arguments which
> can be large and thus expensive to walk repeatedly. The change to
> invalid_tparm_referent_p is needed to avoid incorrectly rejecting class
> NTTP arguments containing function pointers as in the testcase.
>
> (The third testcase is unrelated to this fix, but it helped rule out an
> earlier approach I was considering and it seems we don't have existing
> test coverage for this situation.)
>
> PR c++/53164
> PR c++/105848
>
> gcc/cp/ChangeLog:
>
> * pt.cc (invalid_tparm_referent_p): Accept ADDR_EXPR of
> FUNCTION_DECL.
> (instantiate_class_template): Call mark_template_arguments_used.
> (tsubst_copy_and_build) <case CALL_EXPR>: Revert r13-995 change.
> (mark_template_arguments_used): Define.
> (instantiate_body): Call mark_template_arguments_used.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/template/fn-ptr3a.C: New test.
> * g++.dg/template/fn-ptr3b.C: New test.
> * g++.dg/template/fn-ptr4.C: New test.
> ---
> gcc/cp/pt.cc | 86 +++++++++++++++++++-----
> gcc/testsuite/g++.dg/template/fn-ptr3a.C | 27 ++++++++
> gcc/testsuite/g++.dg/template/fn-ptr3b.C | 30 +++++++++
> gcc/testsuite/g++.dg/template/fn-ptr4.C | 14 ++++
> 4 files changed, 139 insertions(+), 18 deletions(-)
> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr3a.C
> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr3b.C
> create mode 100644 gcc/testsuite/g++.dg/template/fn-ptr4.C
>
> diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
> index dd7f0db9658..022f295b36f 100644
> --- a/gcc/cp/pt.cc
> +++ b/gcc/cp/pt.cc
> @@ -220,6 +220,7 @@ static tree make_argument_pack (tree);
> static tree enclosing_instantiation_of (tree tctx);
> static void instantiate_body (tree pattern, tree args, tree d, bool nested);
> static tree maybe_dependent_member_ref (tree, tree, tsubst_flags_t, tree);
> +static void mark_template_arguments_used (tree, tree);
>
> /* Make the current scope suitable for access checking when we are
> processing T. T can be FUNCTION_DECL for instantiated function
> @@ -7142,12 +7143,13 @@ invalid_tparm_referent_p (tree type, tree expr, tsubst_flags_t complain)
> decl = TREE_OPERAND (decl, 0);
> }
>
> - if (!VAR_P (decl))
> + if (!VAR_OR_FUNCTION_DECL_P (decl))
> {
> if (complain & tf_error)
> error_at (cp_expr_loc_or_input_loc (expr),
> "%qE is not a valid template argument of type %qT "
> - "because %qE is not a variable", expr, type, decl);
> + "because %qE is not a variable or function",
> + expr, type, decl);
> return true;
> }
> else if (cxx_dialect < cxx11 && !DECL_EXTERNAL_LINKAGE_P (decl))
> @@ -12152,6 +12154,9 @@ instantiate_class_template (tree type)
> cp_unevaluated_operand = 0;
> c_inhibit_evaluation_warnings = 0;
> }
> +
> + mark_template_arguments_used (templ, CLASSTYPE_TI_ARGS (type));
> +
> /* Use #pragma pack from the template context. */
> saved_maximum_field_alignment = maximum_field_alignment;
> maximum_field_alignment = TYPE_PRECISION (pattern);
> @@ -21192,22 +21197,10 @@ tsubst_copy_and_build (tree t,
> }
>
> /* Remember that there was a reference to this entity. */
> - if (function != NULL_TREE)
> - {
> - tree inner = function;
> - if (TREE_CODE (inner) == ADDR_EXPR
> - && TREE_CODE (TREE_OPERAND (inner, 0)) == FUNCTION_DECL)
> - /* We should already have called mark_used when taking the
> - address of this function, but do so again anyway to make
> - sure it's odr-used: at worst this is a no-op, but if we
> - obtained this FUNCTION_DECL as part of ahead-of-time overload
> - resolution then that call to mark_used wouldn't have marked it
> - odr-used yet (53164). */
> - inner = TREE_OPERAND (inner, 0);
> - if (DECL_P (inner)
> - && !mark_used (inner, complain) && !(complain & tf_error))
> - RETURN (error_mark_node);
> - }
> + if (function != NULL_TREE
> + && DECL_P (function)
> + && !mark_used (function, complain) && !(complain & tf_error))
> + RETURN (error_mark_node);
>
> if (!maybe_fold_fn_template_args (function, complain))
> return error_mark_node;
> @@ -21902,6 +21895,61 @@ check_instantiated_args (tree tmpl, tree args, tsubst_flags_t complain)
> return result;
> }
>
> +/* Call mark_used on each entity within the non-type template arguments in
> + ARGS for an instantiation of TMPL, to ensure that each such entity is
> + considered odr-used (and therefore marked for instantiation) regardless of
> + whether the specialization was first formed in a template context (which
> + inhibits mark_used).
> +
> + This function assumes push_to_top_level has been called beforehand. */
> +
> +static void
> +mark_template_arguments_used (tree tmpl, tree args)
> +{
> + /* It suffices to do this only when instantiating a primary template. */
> + if (TREE_CODE (tmpl) != TEMPLATE_DECL || !PRIMARY_TEMPLATE_P (tmpl))
> + return;
> +
> + /* We already marked outer arguments when specializing the context. */
> + args = INNERMOST_TEMPLATE_ARGS (args);
> +
> + for (tree arg : tree_vec_range (args))
> + {
> + /* A (pointer/reference to) function or variable NTTP argument. */
> + if (TREE_CODE (arg) == ADDR_EXPR
> + || TREE_CODE (arg) == INDIRECT_REF)
> + {
> + while (TREE_CODE (arg) == ADDR_EXPR
> + || REFERENCE_REF_P (arg)
> + || CONVERT_EXPR_P (arg))
> + arg = TREE_OPERAND (arg, 0);
> + if (VAR_OR_FUNCTION_DECL_P (arg))
> + {
> + /* Pass tf_none to avoid duplicate diagnostics: if this call
> + fails then an earlier call to mark_used for this argument
> + must have also failed and emitted a diagnostic. */
> + bool ok = mark_used (arg, tf_none);
> + gcc_checking_assert (ok || seen_error ());
> + }
> + }
> + /* A class NTTP argument. */
> + else if (VAR_P (arg)
> + && DECL_NTTP_OBJECT_P (arg))
> + {
> + auto mark_used_r = [](tree *tp, int *, void *) {
> + if (VAR_OR_FUNCTION_DECL_P (*tp))
> + {
> + bool ok = mark_used (*tp, tf_none);
> + gcc_checking_assert (ok || seen_error ());
> + }
> + return NULL_TREE;
> + };
> + cp_walk_tree_without_duplicates (&DECL_INITIAL (arg),
> + mark_used_r, nullptr);
> + }
> + }
> +}
> +
> /* We're out of SFINAE context now, so generate diagnostics for the access
> errors we saw earlier when instantiating D from TMPL and ARGS. */
>
> @@ -26775,6 +26823,8 @@ instantiate_body (tree pattern, tree args, tree d, bool nested_p)
> c_inhibit_evaluation_warnings = 0;
> }
>
> + mark_template_arguments_used (pattern, args);
> +
> if (VAR_P (d))
> {
> /* The variable might be a lambda's extra scope, and that
> diff --git a/gcc/testsuite/g++.dg/template/fn-ptr3a.C b/gcc/testsuite/g++.dg/template/fn-ptr3a.C
> new file mode 100644
> index 00000000000..345c621a0c6
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/template/fn-ptr3a.C
> @@ -0,0 +1,27 @@
> +// PR c++/53164
> +// PR c++/105848
> +// A stricter version of fn-ptr3.C that verifies using f as a template
> +// argument alone constitutes an odr-use.
> +
> +template<class T>
> +void f(T) { T::fail; } // { dg-error "fail" }
> +
> +template<void (*P)(int)>
> +struct A {
> + // P not called
> +};
> +
> +template<void (&P)(char)>
> +void wrap() {
> + // P not called
> +}
> +
> +template<int>
> +void g() {
> + A<f> a; // { dg-message "required from" }
> + wrap<f>(); // { dg-message "required from" }
> +}
> +
> +int main() {
> + g<0>();
> +}
> diff --git a/gcc/testsuite/g++.dg/template/fn-ptr3b.C b/gcc/testsuite/g++.dg/template/fn-ptr3b.C
> new file mode 100644
> index 00000000000..899c355fb38
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/template/fn-ptr3b.C
> @@ -0,0 +1,30 @@
> +// PR c++/53164
> +// PR c++/105848
> +// A C++20 version of fn-ptr3a.C using class NTTPs.
> +// { dg-do compile { target c++20 } }
> +
> +template<class T>
> +void f(T) { T::fail; } // { dg-error "fail" }
> +
> +struct B_int { void (*P)(int); };
> +struct B_char { void (&P)(char); };
> +
> +template<B_int B>
> +struct A {
> + // B.P not called
> +};
> +
> +template<B_char B>
> +void wrap() {
> + // B.P not called
> +}
> +
> +template<int>
> +void g() {
> + A<B_int{f}> a; // { dg-message "required from" }
> + wrap<B_char{f}>(); // { dg-message "required from" }
> +}
> +
> +int main() {
> + g<0>();
> +}
> diff --git a/gcc/testsuite/g++.dg/template/fn-ptr4.C b/gcc/testsuite/g++.dg/template/fn-ptr4.C
> new file mode 100644
> index 00000000000..36551ec5d7f
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/template/fn-ptr4.C
> @@ -0,0 +1,14 @@
> +// { dg-do compile }
> +
> +template<void (*P)()>
> +void wrap() {
> + P(); // OK, despite A::g not being accessible from here.
> +}
> +
> +struct A {
> + static void f() {
> + wrap<A::g>();
> + }
> +private:
> + static void g();
> +};
prev parent reply other threads:[~2023-04-01 3:50 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-23 21:18 Patrick Palka
2023-03-23 21:18 ` [PATCH 2/2] c++: duplicate "use of deleted fn" diagnostic [PR106880] Patrick Palka
2023-03-30 2:58 ` Jason Merrill
2023-03-23 21:26 ` [PATCH 1/2] c++: improve "NTTP argument considered unused" fix [PR53164, PR105848] Patrick Palka
2023-03-27 13:30 ` Patrick Palka
2023-03-29 21:36 ` Jason Merrill
2023-03-30 18:53 ` Patrick Palka
2023-03-30 21:39 ` Jason Merrill
2023-03-31 15:25 ` Patrick Palka
2023-03-31 15:55 ` Patrick Palka
2023-03-31 17:28 ` Jason Merrill
2023-03-31 19:06 ` Patrick Palka
2023-04-01 3:50 ` Jason Merrill [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=74ccb540-b964-fc2f-d52d-c5832b5df563@redhat.com \
--to=jason@redhat.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=ppalka@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).