public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Patrick Palka <ppalka@redhat.com>
To: Patrick Palka <ppalka@redhat.com>
Cc: gcc-patches@gcc.gnu.org, jason@redhat.com
Subject: Re: [PATCH 1/2] c++: refine dependent_alias_template_spec_p [PR90679]
Date: Fri, 2 Jun 2023 12:20:44 -0400 (EDT)	[thread overview]
Message-ID: <bcc39c6d-e85b-af21-7e81-06190c79317e@idea> (raw)
In-Reply-To: <20230601144938.765175-1-ppalka@redhat.com>

On Thu, 1 Jun 2023, Patrick Palka wrote:

> For a complex alias template-id, dependent_alias_template_spec_p returns
> true if any template argument of the template-id is dependent.  This
> predicate indicates that substitution into the template-id may behave
> differently with respect to SFINAE than substitution into the expanded
> alias, and so the alias is in a way non-transparent.  For example
> 'first_t<T, T&>' in
> 
>   template<class T, class...> using first_t = T;
>   template<class T> first_t<T, T&> f();
> 
> is such an alias template-id since first_t doesn't use its second
> template parameter and so the substitution into the expanded alias would
> discard the SFINAE effects of the corresponding (dependent) argument 'T&'.
> 
> But this predicate is overly conservative since what really matters for
> sake of SFINAE equivalence is whether a template argument corresponding
> to an _unused_ template parameter is dependent.  So the predicate should
> return false for e.g. 'first_t<T, int>' or 'first_t<T>'.
> 
> This patch refines the predicate appropriately.  We need to be able to
> efficiently determine which template parameters of a complex alias
> template are unused, so to that end we add a new out parameter to
> complex_alias_template_p and cache its result in an on-the-side
> hash_map that replaces the existing TEMPLATE_DECL_COMPLEX_ALIAS_P
> flag.  And in doing so, we fix a latent bug that this flag wasn't
> being propagated during partial instantiation, and so we were treating
> all partially instantiated member alias templates as non-complex.

Whoops, this last sentence is wrong I think.  The flag propagation would
have happened via the call to copy_decl from tsubst_template_decl, so
there was no latent bug.

> 
> 	PR c++/90679
> 
> gcc/cp/ChangeLog:
> 
> 	* cp-tree.h (TEMPLATE_DECL_COMPLEX_ALIAS_P): Remove.
> 	(most_general_template): Constify parameter.
> 	* pt.cc (push_template_decl): Adjust after removing
> 	TEMPLATE_DECL_COMPLEX_ALIAS_P.
> 	(complex_alias_tmpl_info): New hash_map.
> 	(uses_all_template_parms_data::seen): Change type to
> 	tree* from bool*.
> 	(complex_alias_template_r): Adjust accordingly.
> 	(complex_alias_template_p): Add 'seen_out' out parameter.
> 	Call most_general_template and check PRIMARY_TEMPLATE_P.
> 	Use complex_alias_tmpl_info to cache the result and set
> 	'*seen_out' accordigly.
> 	(dependent_alias_template_spec_p): Add !processing_template_decl
> 	early exit test.  Consider dependence of only template arguments
> 	corresponding to seen template parameters as per
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/cpp0x/alias-decl-75.C: New test.
> ---
>  gcc/cp/cp-tree.h                           |   7 +-
>  gcc/cp/pt.cc                               | 101 +++++++++++++++------
>  gcc/testsuite/g++.dg/cpp0x/alias-decl-75.C |  24 +++++
>  3 files changed, 100 insertions(+), 32 deletions(-)
>  create mode 100644 gcc/testsuite/g++.dg/cpp0x/alias-decl-75.C
> 
> diff --git a/gcc/cp/cp-tree.h b/gcc/cp/cp-tree.h
> index a1b882f11fe..5330d1e1f62 100644
> --- a/gcc/cp/cp-tree.h
> +++ b/gcc/cp/cp-tree.h
> @@ -543,7 +543,6 @@ extern GTY(()) tree cp_global_trees[CPTI_MAX];
>     2: DECL_THIS_EXTERN (in VAR_DECL, FUNCTION_DECL or PARM_DECL)
>        DECL_IMPLICIT_TYPEDEF_P (in a TYPE_DECL)
>        DECL_CONSTRAINT_VAR_P (in a PARM_DECL)
> -      TEMPLATE_DECL_COMPLEX_ALIAS_P (in TEMPLATE_DECL)
>        DECL_INSTANTIATING_NSDMI_P (in a FIELD_DECL)
>        USING_DECL_UNRELATED_P (in USING_DECL)
>     3: DECL_IN_AGGR_P.
> @@ -3655,10 +3654,6 @@ struct GTY(()) lang_decl {
>  #define TYPE_DECL_ALIAS_P(NODE) \
>    DECL_LANG_FLAG_6 (TYPE_DECL_CHECK (NODE))
>  
> -/* Nonzero for TEMPLATE_DECL means that it is a 'complex' alias template.  */
> -#define TEMPLATE_DECL_COMPLEX_ALIAS_P(NODE) \
> -  DECL_LANG_FLAG_2 (TEMPLATE_DECL_CHECK (NODE))
> -
>  /* Nonzero for a type which is an alias for another type; i.e, a type
>     which declaration was written 'using name-of-type =
>     another-type'.  */
> @@ -7403,7 +7398,7 @@ extern tree tsubst_argument_pack		(tree, tree, tsubst_flags_t, tree);
>  extern tree tsubst_template_args		(tree, tree, tsubst_flags_t, tree);
>  extern tree tsubst_template_arg			(tree, tree, tsubst_flags_t, tree);
>  extern tree tsubst_function_parms		(tree, tree, tsubst_flags_t, tree);
> -extern tree most_general_template		(tree);
> +extern tree most_general_template		(const_tree);
>  extern tree get_mostly_instantiated_function_type (tree);
>  extern bool problematic_instantiation_changed	(void);
>  extern void record_last_problematic_instantiation (void);
> diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
> index 7fb3e75bceb..1b28195e10d 100644
> --- a/gcc/cp/pt.cc
> +++ b/gcc/cp/pt.cc
> @@ -211,7 +211,6 @@ static tree listify (tree);
>  static tree listify_autos (tree, tree);
>  static tree tsubst_template_parm (tree, tree, tsubst_flags_t);
>  static tree instantiate_alias_template (tree, tree, tsubst_flags_t);
> -static bool complex_alias_template_p (const_tree tmpl);
>  static tree get_underlying_template (tree);
>  static tree tsubst_attributes (tree, tree, tsubst_flags_t, tree);
>  static tree canonicalize_expr_argument (tree, tsubst_flags_t);
> @@ -6233,8 +6232,6 @@ push_template_decl (tree decl, bool is_friend)
>  		  constr = build_constraints (constr, NULL_TREE);
>  		  set_constraints (decl, constr);
>  		}
> -	      if (complex_alias_template_p (tmpl))
> -		TEMPLATE_DECL_COMPLEX_ALIAS_P (tmpl) = true;
>  	    }
>  	}
>  
> @@ -6526,12 +6523,16 @@ alias_template_specialization_p (const_tree t,
>    return NULL_TREE;
>  }
>  
> +/* A cache of the result of complex_alias_template_p.  */
> +
> +static GTY(()) hash_map<const_tree, tree> *complex_alias_tmpl_info;
> +
>  /* Data structure for complex_alias_template_*.  */
>  
>  struct uses_all_template_parms_data
>  {
>    int level;
> -  bool *seen;
> +  tree *seen;
>  };
>  
>  /* walk_tree callback for complex_alias_template_p.  */
> @@ -6551,7 +6552,7 @@ complex_alias_template_r (tree *tp, int *walk_subtrees, void *data_)
>        {
>  	tree idx = get_template_parm_index (t);
>  	if (TEMPLATE_PARM_LEVEL (idx) == data.level)
> -	  data.seen[TEMPLATE_PARM_IDX (idx)] = true;
> +	  data.seen[TEMPLATE_PARM_IDX (idx)] = boolean_true_node;
>        }
>  
>      default:;
> @@ -6576,7 +6577,7 @@ complex_alias_template_r (tree *tp, int *walk_subtrees, void *data_)
>  	return t;
>  
>        /* Consider the expanded packs to be used outside the expansion...  */
> -      data.seen[idx] = true;
> +      data.seen[idx] = boolean_true_node;
>      }
>  
>    /* ...but don't walk into the pattern.  Consider PR104008:
> @@ -6598,12 +6599,17 @@ complex_alias_template_r (tree *tp, int *walk_subtrees, void *data_)
>     using that alias can be ill-formed when the expansion is not, as with
>     the void_t template.
>  
> -   Returns 1 if always complex, 0 if not complex, -1 if complex iff any of the
> -   template arguments are empty packs.  */
> +   If this predicate returns true in the ordinary case, the out parameter
> +   SEEN_OUT is set to a TREE_VEC containing boolean_true_node at element I if
> +   the I'th template parameter of the alias template is used in the alias.  */
>  
>  static bool
> -complex_alias_template_p (const_tree tmpl)
> +complex_alias_template_p (const_tree tmpl, tree *seen_out)
>  {
> +  tmpl = most_general_template (tmpl);
> +  if (!PRIMARY_TEMPLATE_P (tmpl))
> +    return false;
> +
>    /* A renaming alias isn't complex.  */
>    if (get_underlying_template (CONST_CAST_TREE (tmpl)) != tmpl)
>      return false;
> @@ -6612,26 +6618,52 @@ complex_alias_template_p (const_tree tmpl)
>    if (get_constraints (tmpl))
>      return true;
>  
> +  if (tree *slot = hash_map_safe_get (complex_alias_tmpl_info, tmpl))
> +    {
> +      tree result = *slot;
> +      if (result == boolean_false_node)
> +	return false;
> +      if (result == boolean_true_node)
> +	return true;
> +      gcc_assert (TREE_CODE (result) == TREE_VEC);
> +      if (seen_out)
> +	*seen_out = result;
> +      return true;
> +    }
> +
>    struct uses_all_template_parms_data data;
>    tree pat = DECL_ORIGINAL_TYPE (DECL_TEMPLATE_RESULT (tmpl));
>    tree parms = DECL_TEMPLATE_PARMS (tmpl);
>    data.level = TMPL_PARMS_DEPTH (parms);
>    int len = TREE_VEC_LENGTH (INNERMOST_TEMPLATE_PARMS (parms));
> -  data.seen = XALLOCAVEC (bool, len);
> +  tree seen = make_tree_vec (len);
> +  data.seen = TREE_VEC_BEGIN (seen);
>    for (int i = 0; i < len; ++i)
> -    data.seen[i] = false;
> +    data.seen[i] = boolean_false_node;
>  
>    if (cp_walk_tree_without_duplicates (&pat, complex_alias_template_r, &data))
> -    return true;
> -  for (int i = 0; i < len; ++i)
> -    if (!data.seen[i])
> +    {
> +      hash_map_safe_put<hm_ggc> (complex_alias_tmpl_info, tmpl, boolean_true_node);
>        return true;
> +    }
> +
> +  for (int i = 0; i < len; ++i)
> +    if (data.seen[i] != boolean_true_node)
> +      {
> +	hash_map_safe_put<hm_ggc> (complex_alias_tmpl_info, tmpl, seen);
> +	if (seen_out)
> +	  *seen_out = seen;
> +	return true;
> +      }
> +
> +  hash_map_safe_put<hm_ggc> (complex_alias_tmpl_info, tmpl, boolean_false_node);
>    return false;
>  }
>  
> -/* If T is a specialization of a complex alias template with dependent
> -   template-arguments, return it; otherwise return NULL_TREE.  If T is a
> -   typedef to such a specialization, return the specialization.  */
> +/* If T is a specialization of a complex alias template with a dependent
> +   argument for an unused template parameter, return it; otherwise return
> +   NULL_TREE.  If T is a typedef to such a specialization, return the
> +   specialization.  */
>  
>  tree
>  dependent_alias_template_spec_p (const_tree t, bool transparent_typedefs)
> @@ -6640,15 +6672,32 @@ dependent_alias_template_spec_p (const_tree t, bool transparent_typedefs)
>      return NULL_TREE;
>    gcc_assert (TYPE_P (t));
>  
> -  if (!typedef_variant_p (t))
> +  if (!processing_template_decl || !typedef_variant_p (t))
>      return NULL_TREE;
>  
> -  tree tinfo = TYPE_ALIAS_TEMPLATE_INFO (t);
> -  if (tinfo
> -      && TEMPLATE_DECL_COMPLEX_ALIAS_P (TI_TEMPLATE (tinfo))
> -      && (any_dependent_template_arguments_p
> -	  (INNERMOST_TEMPLATE_ARGS (TI_ARGS (tinfo)))))
> -    return CONST_CAST_TREE (t);
> +  if (tree tinfo = TYPE_ALIAS_TEMPLATE_INFO (t))
> +    {
> +      tree seen = NULL_TREE;
> +      if (complex_alias_template_p (TI_TEMPLATE (tinfo), &seen))
> +	{
> +	  tree args = INNERMOST_TEMPLATE_ARGS (TI_ARGS (tinfo));
> +	  if (!seen)
> +	    {
> +	      if (any_dependent_template_arguments_p (args))
> +		return CONST_CAST_TREE (t);
> +	    }
> +	  else
> +	    {
> +	      gcc_assert (TREE_VEC_LENGTH (args) == TREE_VEC_LENGTH (seen));
> +	      for (int i = 0, len = TREE_VEC_LENGTH (args); i < len; ++i)
> +		if (TREE_VEC_ELT (seen, i) != boolean_true_node
> +		    && dependent_template_arg_p (TREE_VEC_ELT (args, i)))
> +		  return CONST_CAST_TREE (t);
> +	    }
> +
> +	  return NULL_TREE;
> +	}
> +    }
>  
>    if (transparent_typedefs)
>      {
> @@ -25958,7 +26007,7 @@ most_specialized_instantiation (tree templates)
>     `template <class T> template <class U> S<T*>::f(U)'.  */
>  
>  tree
> -most_general_template (tree decl)
> +most_general_template (const_tree decl)
>  {
>    if (TREE_CODE (decl) != TEMPLATE_DECL)
>      {
> @@ -25992,7 +26041,7 @@ most_general_template (tree decl)
>        decl = DECL_TI_TEMPLATE (decl);
>      }
>  
> -  return decl;
> +  return CONST_CAST_TREE (decl);
>  }
>  
>  /* Return the most specialized of the template partial specializations
> diff --git a/gcc/testsuite/g++.dg/cpp0x/alias-decl-75.C b/gcc/testsuite/g++.dg/cpp0x/alias-decl-75.C
> new file mode 100644
> index 00000000000..10592f521a0
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp0x/alias-decl-75.C
> @@ -0,0 +1,24 @@
> +// PR c++/90679
> +// { dg-do compile { target c++11 } }
> +
> +template<class T, class...>
> +using first_t = T;
> +
> +template<class T>
> +struct A;
> +
> +template<class T>
> +struct traits;
> +
> +template<class T>
> +struct traits<A<first_t<T>>> {
> +  static constexpr int value = 1;
> +};
> +
> +template<class T>
> +struct traits<A<first_t<const T>>> {
> +  static constexpr int value = 2;
> +};
> +
> +static_assert(traits<A<int>>::value == 1, "");
> +static_assert(traits<A<const int>>::value == 2, ""); // { dg-bogus "ambiguous" }
> -- 
> 2.41.0.rc1.10.g9e49351c30
> 
> 


  parent reply	other threads:[~2023-06-02 16:20 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-06-01 14:49 Patrick Palka
2023-06-01 14:49 ` [PATCH 2/2] c++: partial ordering and dep alias tmpl specs [PR90679] Patrick Palka
2023-09-11 14:19   ` Patrick Palka
2023-12-15 19:07   ` Patrick Palka
2023-12-18 21:57     ` Jason Merrill
2023-06-02 16:20 ` Patrick Palka [this message]
2023-09-11 14:19 ` [PATCH 1/2] c++: refine dependent_alias_template_spec_p [PR90679] Patrick Palka
2023-12-15 19:06   ` Patrick Palka
2023-12-18 21:53     ` Jason Merrill

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=bcc39c6d-e85b-af21-7e81-06190c79317e@idea \
    --to=ppalka@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=jason@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).