c++: Allow mixing GNU/std-style attributes [PR69585]

cp_parser_attributes_opt doesn't accept GNU attributes followed by
[[]] attributes and vice versa; only a sequence of attributes of the
same kind.  That causes grief for code like:

  struct __attribute__ ((may_alias)) alignas (2) struct S { };

or

  #define EXPORT __attribute__((visibility("default")))
  struct [[nodiscard]] EXPORT F { };

It doesn't seem to a documented restriction, so this patch fixes the
problem.

However, the patch does not touch the C FE.  The C FE doesn't have
a counterpart to C++'s cp_parser_attributes_opt -- it only has
c_parser_transaction_attributes (which parses both GNU and [[]]
attributes), but that's TM-specific.  The C FE seems to use either
c_parser_gnu_attributes or c_parser_std_attribute_specifier_sequence.
As a consequence, this works:

  [[maybe_unused]] __attribute__((deprecated)) void f2 ();

but this doesn't:

  __attribute__((deprecated)) [[maybe_unused]] void f1 ();

I'm not sure what, if anything, should be done about this.

Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?

	PR c++/102399
	PR c++/69585

gcc/cp/ChangeLog:

	* parser.cc (cp_parser_attributes_opt): Accept GNU attributes
	followed by [[]] attributes and vice versa.

gcc/testsuite/ChangeLog:

	* g++.dg/ext/attrib65.C: New test.
	* g++.dg/ext/attrib66.C: New test.
	* g++.dg/ext/attrib67.C: New test.
---
 gcc/cp/parser.cc                    | 14 +++++++++++---
 gcc/testsuite/g++.dg/ext/attrib65.C |  7 +++++++
 gcc/testsuite/g++.dg/ext/attrib66.C | 27 +++++++++++++++++++++++++++
 gcc/testsuite/g++.dg/ext/attrib67.C | 27 +++++++++++++++++++++++++++
 4 files changed, 72 insertions(+), 3 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/ext/attrib65.C
 create mode 100644 gcc/testsuite/g++.dg/ext/attrib66.C
 create mode 100644 gcc/testsuite/g++.dg/ext/attrib67.C


base-commit: 891d64721626f45fb95fa47a57a3f396b80f31e9

On 6/3/22 20:05, Marek Polacek wrote:
> cp_parser_attributes_opt doesn't accept GNU attributes followed by
> [[]] attributes and vice versa; only a sequence of attributes of the
> same kind.  That causes grief for code like:
> 
>    struct __attribute__ ((may_alias)) alignas (2) struct S { };
> 
> or
> 
>    #define EXPORT __attribute__((visibility("default")))
>    struct [[nodiscard]] EXPORT F { };
> 
> It doesn't seem to a documented restriction, so this patch fixes the
> problem.
> 
> However, the patch does not touch the C FE.  The C FE doesn't have
> a counterpart to C++'s cp_parser_attributes_opt -- it only has
> c_parser_transaction_attributes (which parses both GNU and [[]]
> attributes), but that's TM-specific.  The C FE seems to use either
> c_parser_gnu_attributes or c_parser_std_attribute_specifier_sequence.
> As a consequence, this works:
> 
>    [[maybe_unused]] __attribute__((deprecated)) void f2 ();
> 
> but this doesn't:
> 
>    __attribute__((deprecated)) [[maybe_unused]] void f1 ();
> 
> I'm not sure what, if anything, should be done about this.
> 
> Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?

OK.

> 	PR c++/102399
> 	PR c++/69585
> 
> gcc/cp/ChangeLog:
> 
> 	* parser.cc (cp_parser_attributes_opt): Accept GNU attributes
> 	followed by [[]] attributes and vice versa.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/ext/attrib65.C: New test.
> 	* g++.dg/ext/attrib66.C: New test.
> 	* g++.dg/ext/attrib67.C: New test.
> ---
>   gcc/cp/parser.cc                    | 14 +++++++++++---
>   gcc/testsuite/g++.dg/ext/attrib65.C |  7 +++++++
>   gcc/testsuite/g++.dg/ext/attrib66.C | 27 +++++++++++++++++++++++++++
>   gcc/testsuite/g++.dg/ext/attrib67.C | 27 +++++++++++++++++++++++++++
>   4 files changed, 72 insertions(+), 3 deletions(-)
>   create mode 100644 gcc/testsuite/g++.dg/ext/attrib65.C
>   create mode 100644 gcc/testsuite/g++.dg/ext/attrib66.C
>   create mode 100644 gcc/testsuite/g++.dg/ext/attrib67.C
> 
> diff --git a/gcc/cp/parser.cc b/gcc/cp/parser.cc
> index 3fc73442da5..535bf7eedbb 100644
> --- a/gcc/cp/parser.cc
> +++ b/gcc/cp/parser.cc
> @@ -28727,9 +28727,17 @@ cp_nth_tokens_can_be_attribute_p (cp_parser *parser, size_t n)
>   static tree
>   cp_parser_attributes_opt (cp_parser *parser)
>   {
> -  if (cp_next_tokens_can_be_gnu_attribute_p (parser))
> -    return cp_parser_gnu_attributes_opt (parser);
> -  return cp_parser_std_attribute_spec_seq (parser);
> +  tree attrs = NULL_TREE;
> +  while (true)
> +    {
> +      if (cp_next_tokens_can_be_gnu_attribute_p (parser))
> +	attrs = attr_chainon (attrs, cp_parser_gnu_attributes_opt (parser));
> +      else if (cp_next_tokens_can_be_std_attribute_p (parser))
> +	attrs = attr_chainon (attrs, cp_parser_std_attribute_spec_seq (parser));
> +      else
> +	break;
> +    }
> +  return attrs;
>   }
>   
>   /* Parse an (optional) series of attributes.
> diff --git a/gcc/testsuite/g++.dg/ext/attrib65.C b/gcc/testsuite/g++.dg/ext/attrib65.C
> new file mode 100644
> index 00000000000..0af138700f6
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/attrib65.C
> @@ -0,0 +1,7 @@
> +// PR c++/102399
> +// { dg-do compile { target c++11 } }
> +// Test mixing the GNU and standard forms of attributes.
> +
> +#define EXPORT __attribute__((visibility("default")))
> +
> +struct [[nodiscard]] EXPORT Foo { Foo(); };
> diff --git a/gcc/testsuite/g++.dg/ext/attrib66.C b/gcc/testsuite/g++.dg/ext/attrib66.C
> new file mode 100644
> index 00000000000..102ed709b1d
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/attrib66.C
> @@ -0,0 +1,27 @@
> +// PR c++/69585
> +// { dg-do compile { target c++11 } }
> +
> +struct __attribute__ ((aligned (2))) __attribute__ ((may_alias))
> +S1 { };
> +
> +struct __attribute__ ((aligned (2))) [[gnu::may_alias]]
> +S2 { };
> +
> +struct alignas (2) __attribute__ ((may_alias))
> +S3 { };
> +
> +struct alignas (2) [[gnu::may_alias]]
> +S4 { };
> +
> +
> +struct __attribute__ ((may_alias)) __attribute__ ((aligned (2)))
> +S1_2 { };
> +
> +struct [[gnu::may_alias]] __attribute__ ((aligned (2)))
> +S2_2 { };
> +
> +struct __attribute__ ((may_alias)) alignas (2)
> +S3_2 { };
> +
> +struct [[gnu::may_alias]] alignas (2)
> +S4_2 { };
> diff --git a/gcc/testsuite/g++.dg/ext/attrib67.C b/gcc/testsuite/g++.dg/ext/attrib67.C
> new file mode 100644
> index 00000000000..a5107665077
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ext/attrib67.C
> @@ -0,0 +1,27 @@
> +// PR c++/69585
> +// { dg-do compile { target c++11 } }
> +// Test mixing the GNU and standard forms of attributes.
> +
> +__attribute__((deprecated)) [[maybe_unused]] void f1 ();
> +[[maybe_unused]] __attribute__((deprecated)) void f2 ();
> +[[maybe_unused]] __attribute__((deprecated)) [[nodiscard]] int f3 ();
> +__attribute__((unused)) [[nodiscard]] __attribute__((deprecated)) int f4 ();
> +
> +struct [[maybe_unused]] __attribute__((aligned)) S1 { double d; };
> +struct __attribute__((aligned)) [[maybe_unused]] S2 { double d; };
> +
> +enum E {
> +  X [[maybe_unused]] __attribute__((unavailable)),
> +  Y  __attribute__((unavailable)) [[maybe_unused]],
> +};
> +
> +void
> +g ([[maybe_unused]] __attribute__((unavailable)) int i1,
> +   __attribute__((unavailable)) [[maybe_unused]] int i2)
> +{
> +  [[maybe_unused]] __attribute__((aligned)) int i3;
> +  __attribute__((aligned)) [[maybe_unused]] int i4;
> +
> +[[maybe_unused]]
> +lab:  __attribute__((cold));
> +}
> 
> base-commit: 891d64721626f45fb95fa47a57a3f396b80f31e9

On Fri, 3 Jun 2022, Marek Polacek via Gcc-patches wrote:

> However, the patch does not touch the C FE.  The C FE doesn't have
> a counterpart to C++'s cp_parser_attributes_opt -- it only has
> c_parser_transaction_attributes (which parses both GNU and [[]]
> attributes), but that's TM-specific.  The C FE seems to use either
> c_parser_gnu_attributes or c_parser_std_attribute_specifier_sequence.
> As a consequence, this works:
> 
>   [[maybe_unused]] __attribute__((deprecated)) void f2 ();
> 
> but this doesn't:
> 
>   __attribute__((deprecated)) [[maybe_unused]] void f1 ();
> 
> I'm not sure what, if anything, should be done about this.

In many places, both GNU and [[]] attributes are valid, but appertain to 
different entities, so it would be hard to define semantics for mixing 
them.  At the start of a declaration like that, for example, GNU 
attributes can be mixed arbitrarily among the declaration specifiers (and 
appertain to the declaration) - but [[]] attributes can appear only before 
the declaration specifiers (appertaining to the declaration) or after them 
(appertaining to the type determined by those declaration specifiers), not 
in the middle.