* [PATCH] c++: sizeof(expr) in non-templated requires-expr [PR108563]
@ 2023-02-09 18:17 Patrick Palka
2023-02-09 18:22 ` Patrick Palka
0 siblings, 1 reply; 3+ messages in thread
From: Patrick Palka @ 2023-02-09 18:17 UTC (permalink / raw)
To: gcc-patches; +Cc: jason, Patrick Palka
When substituting into sizeof(expr), tsubst_copy_and_build elides
substitution into the operand if args is NULL_TREE, and instead
considers the TREE_TYPE of the operand. But here the (templated)
operand is a TEMPLATE_ID_EXPR with empty TREE_TYPE, so we can't elide
substitution in this case.
Contrary to the associated comment (dating back to r69130) substituting
args=NULL_TREE should generally work since we do exactly that in e.g.
fold_non_dependent_expr, and I don't see why the operand of sizeof would
be an exception. So this patch just removes this special case.
Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
trunk? Diff generated with -w to ignore noisy whitespace changes.
PR c++/108563
gcc/cp/ChangeLog:
* pt.cc (tsubst_copy_and_build) <case SIZEOF_EXPR>: Remove
special case for empty args.
gcc/testsuite/ChangeLog:
* g++.dg/cpp2a/concepts-requires35.C: New test.
---
gcc/cp/pt.cc | 25 ++++++-------------
.../g++.dg/cpp2a/concepts-requires35.C | 14 +++++++++++
2 files changed, 21 insertions(+), 18 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
index 9f3fc1fa089..f21d28263d1 100644
--- a/gcc/cp/pt.cc
+++ b/gcc/cp/pt.cc
@@ -20652,25 +20652,14 @@ tsubst_copy_and_build (tree t,
op1 = TREE_TYPE (op1);
bool std_alignof = (TREE_CODE (t) == ALIGNOF_EXPR
&& ALIGNOF_EXPR_STD_P (t));
- if (!args)
- {
- /* When there are no ARGS, we are trying to evaluate a
- non-dependent expression from the parser. Trying to do
- the substitutions may not work. */
- if (!TYPE_P (op1))
- op1 = TREE_TYPE (op1);
- }
+ ++cp_unevaluated_operand;
+ ++c_inhibit_evaluation_warnings;
+ if (TYPE_P (op1))
+ op1 = tsubst (op1, args, complain, in_decl);
else
- {
- ++cp_unevaluated_operand;
- ++c_inhibit_evaluation_warnings;
- if (TYPE_P (op1))
- op1 = tsubst (op1, args, complain, in_decl);
- else
- op1 = tsubst_copy_and_build (op1, args, complain, in_decl);
- --cp_unevaluated_operand;
- --c_inhibit_evaluation_warnings;
- }
+ op1 = tsubst_copy_and_build (op1, args, complain, in_decl);
+ --cp_unevaluated_operand;
+ --c_inhibit_evaluation_warnings;
if (TYPE_P (op1))
r = cxx_sizeof_or_alignof_type (input_location,
op1, TREE_CODE (t), std_alignof,
diff --git a/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C b/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
new file mode 100644
index 00000000000..2bb4b2b0b5d
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
@@ -0,0 +1,14 @@
+// PR c++/108563
+// { dg-do compile { target c++20 } }
+
+template<class T>
+struct foo {
+ static constexpr T value = 0;
+};
+
+template<class T>
+inline constexpr T foo_v = foo<T>::value;
+
+static_assert(requires { sizeof(foo_v<int>); });
+static_assert(requires { requires sizeof(foo_v<int*>) == sizeof(int*); });
+static_assert(requires { requires sizeof(foo_v<char>) == sizeof(char); });
--
2.39.1.418.g7876265d61
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] c++: sizeof(expr) in non-templated requires-expr [PR108563]
2023-02-09 18:17 [PATCH] c++: sizeof(expr) in non-templated requires-expr [PR108563] Patrick Palka
@ 2023-02-09 18:22 ` Patrick Palka
2023-02-14 23:02 ` Jason Merrill
0 siblings, 1 reply; 3+ messages in thread
From: Patrick Palka @ 2023-02-09 18:22 UTC (permalink / raw)
To: Patrick Palka; +Cc: gcc-patches, jason
On Thu, 9 Feb 2023, Patrick Palka wrote:
> When substituting into sizeof(expr), tsubst_copy_and_build elides
> substitution into the operand if args is NULL_TREE, and instead
> considers the TREE_TYPE of the operand. But here the (templated)
> operand is a TEMPLATE_ID_EXPR with empty TREE_TYPE, so we can't elide
> substitution in this case.
>
> Contrary to the associated comment (dating back to r69130) substituting
> args=NULL_TREE should generally work since we do exactly that in e.g.
> fold_non_dependent_expr, and I don't see why the operand of sizeof would
> be an exception. So this patch just removes this special case.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
> trunk? Diff generated with -w to ignore noisy whitespace changes.
This time with -w actually passed to format-patch:
-- >8 --
PR c++/108563
gcc/cp/ChangeLog:
* pt.cc (tsubst_copy_and_build) <case SIZEOF_EXPR>: Remove
special case for empty args.
gcc/testsuite/ChangeLog:
* g++.dg/cpp2a/concepts-requires35.C: New test.
---
gcc/cp/pt.cc | 11 -----------
gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C | 14 ++++++++++++++
2 files changed, 14 insertions(+), 11 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
index 9f3fc1fa089..f21d28263d1 100644
--- a/gcc/cp/pt.cc
+++ b/gcc/cp/pt.cc
@@ -20652,16 +20652,6 @@ tsubst_copy_and_build (tree t,
op1 = TREE_TYPE (op1);
bool std_alignof = (TREE_CODE (t) == ALIGNOF_EXPR
&& ALIGNOF_EXPR_STD_P (t));
- if (!args)
- {
- /* When there are no ARGS, we are trying to evaluate a
- non-dependent expression from the parser. Trying to do
- the substitutions may not work. */
- if (!TYPE_P (op1))
- op1 = TREE_TYPE (op1);
- }
- else
- {
++cp_unevaluated_operand;
++c_inhibit_evaluation_warnings;
if (TYPE_P (op1))
@@ -20670,7 +20660,6 @@ tsubst_copy_and_build (tree t,
op1 = tsubst_copy_and_build (op1, args, complain, in_decl);
--cp_unevaluated_operand;
--c_inhibit_evaluation_warnings;
- }
if (TYPE_P (op1))
r = cxx_sizeof_or_alignof_type (input_location,
op1, TREE_CODE (t), std_alignof,
diff --git a/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C b/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
new file mode 100644
index 00000000000..2bb4b2b0b5d
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
@@ -0,0 +1,14 @@
+// PR c++/108563
+// { dg-do compile { target c++20 } }
+
+template<class T>
+struct foo {
+ static constexpr T value = 0;
+};
+
+template<class T>
+inline constexpr T foo_v = foo<T>::value;
+
+static_assert(requires { sizeof(foo_v<int>); });
+static_assert(requires { requires sizeof(foo_v<int*>) == sizeof(int*); });
+static_assert(requires { requires sizeof(foo_v<char>) == sizeof(char); });
--
2.39.1.418.g7876265d61
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] c++: sizeof(expr) in non-templated requires-expr [PR108563]
2023-02-09 18:22 ` Patrick Palka
@ 2023-02-14 23:02 ` Jason Merrill
0 siblings, 0 replies; 3+ messages in thread
From: Jason Merrill @ 2023-02-14 23:02 UTC (permalink / raw)
To: Patrick Palka; +Cc: gcc-patches
On 2/9/23 10:22, Patrick Palka wrote:
> On Thu, 9 Feb 2023, Patrick Palka wrote:
>
>> When substituting into sizeof(expr), tsubst_copy_and_build elides
>> substitution into the operand if args is NULL_TREE, and instead
>> considers the TREE_TYPE of the operand. But here the (templated)
>> operand is a TEMPLATE_ID_EXPR with empty TREE_TYPE, so we can't elide
>> substitution in this case.
>>
>> Contrary to the associated comment (dating back to r69130) substituting
>> args=NULL_TREE should generally work since we do exactly that in e.g.
>> fold_non_dependent_expr, and I don't see why the operand of sizeof would
>> be an exception. So this patch just removes this special case.
>>
>> Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
>> trunk? Diff generated with -w to ignore noisy whitespace changes.
>
> This time with -w actually passed to format-patch:
OK.
> -- >8 --
>
> PR c++/108563
>
> gcc/cp/ChangeLog:
>
> * pt.cc (tsubst_copy_and_build) <case SIZEOF_EXPR>: Remove
> special case for empty args.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/cpp2a/concepts-requires35.C: New test.
> ---
> gcc/cp/pt.cc | 11 -----------
> gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C | 14 ++++++++++++++
> 2 files changed, 14 insertions(+), 11 deletions(-)
> create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
>
> diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
> index 9f3fc1fa089..f21d28263d1 100644
> --- a/gcc/cp/pt.cc
> +++ b/gcc/cp/pt.cc
> @@ -20652,16 +20652,6 @@ tsubst_copy_and_build (tree t,
> op1 = TREE_TYPE (op1);
> bool std_alignof = (TREE_CODE (t) == ALIGNOF_EXPR
> && ALIGNOF_EXPR_STD_P (t));
> - if (!args)
> - {
> - /* When there are no ARGS, we are trying to evaluate a
> - non-dependent expression from the parser. Trying to do
> - the substitutions may not work. */
> - if (!TYPE_P (op1))
> - op1 = TREE_TYPE (op1);
> - }
> - else
> - {
> ++cp_unevaluated_operand;
> ++c_inhibit_evaluation_warnings;
> if (TYPE_P (op1))
> @@ -20670,7 +20660,6 @@ tsubst_copy_and_build (tree t,
> op1 = tsubst_copy_and_build (op1, args, complain, in_decl);
> --cp_unevaluated_operand;
> --c_inhibit_evaluation_warnings;
> - }
> if (TYPE_P (op1))
> r = cxx_sizeof_or_alignof_type (input_location,
> op1, TREE_CODE (t), std_alignof,
> diff --git a/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C b/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
> new file mode 100644
> index 00000000000..2bb4b2b0b5d
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/concepts-requires35.C
> @@ -0,0 +1,14 @@
> +// PR c++/108563
> +// { dg-do compile { target c++20 } }
> +
> +template<class T>
> +struct foo {
> + static constexpr T value = 0;
> +};
> +
> +template<class T>
> +inline constexpr T foo_v = foo<T>::value;
> +
> +static_assert(requires { sizeof(foo_v<int>); });
> +static_assert(requires { requires sizeof(foo_v<int*>) == sizeof(int*); });
> +static_assert(requires { requires sizeof(foo_v<char>) == sizeof(char); });
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2023-02-14 23:02 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-09 18:17 [PATCH] c++: sizeof(expr) in non-templated requires-expr [PR108563] Patrick Palka
2023-02-09 18:22 ` Patrick Palka
2023-02-14 23:02 ` Jason Merrill
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).