public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Richard Sandiford <richard.sandiford@arm.com>
To: Joseph Myers <joseph@codesourcery.com>
Cc: gcc-patches@gcc.gnu.org
Subject: [PATCH] c: Add support for [[__extension__ ...]]
Date: Thu, 17 Aug 2023 12:24:48 +0100	[thread overview]
Message-ID: <mptlee9j4hr.fsf_-_@arm.com> (raw)
In-Reply-To: <50a9b1a1-9efb-a170-34a1-c1e98f69afa@codesourcery.com> (Joseph Myers's message of "Wed, 16 Aug 2023 13:22:18 +0000")

Joseph Myers <joseph@codesourcery.com> writes:
> On Wed, 16 Aug 2023, Richard Sandiford via Gcc-patches wrote:
>
>> Would it be OK to add support for:
>> 
>>   [[__extension__ ...]]
>> 
>> to suppress the pedwarn about using [[]] prior to C2X?  Then we can
>
> That seems like a plausible feature to add.

Thanks.  Of course, once I actually tried it, I hit a snag:
:: isn't a single lexing token prior to C2X, and so something like:

  [[__extension__ arm::streaming]]

would not be interpreted as a scoped attribute in C11.  The patch
gets around that by allowing two colons in place of :: when
__extension__ is used.  I realise that's pushing the bounds of
acceptability though...

I wondered about trying to require the two colons to be immediately
adjacent.  But:

(a) There didn't appear to be an existing API to check that, which seemed
    like a red flag.  The closest I could find was get_source_text_between.

    Similarly to that, it would in principle be possible to compare
    two expanded locations.  But...

(b) I had a vague impression that locations were allowed to drop column
    information for very large inputs (maybe I'm wrong).

(c) It wouldn't cope with token pasting.

So in the end I just used a simple two-token test, like for [[ and ]].

Bootstrapped & regression-tested on aarch64-linux-gnu.

Richard

----

[[]] attributes are a recent addition to C, but as a GNU extension,
GCC allows them to be used in C11 and earlier.  Normally this use
would trigger a pedwarn (for -pedantic, -Wc11-c2x-compat, etc.).

This patch allows the pedwarn to be suppressed by starting the
attribute-list with __extension__.

Also, :: is not a single lexing token prior to C2X, so it wasn't
possible to use scoped attributes in C11, even as a GNU extension.
The patch allows two colons to be used in place of :: when
__extension__ is used.  No attempt is made to check whether the
two colons are immediately adjacent.

gcc/
	* doc/extend.texi: Document the C [[__extension__ ...]] construct.

gcc/c/
	* c-parser.cc (c_parser_std_attribute): Conditionally allow
	two colons to be used in place of ::.
	(c_parser_std_attribute_list): New function, split out from...
	(c_parser_std_attribute_specifier): ...here.  Allow the attribute-list
	to start with __extension__.  When it does, also allow two colons
	to be used in place of ::.

gcc/testsuite/
	* gcc.dg/c2x-attr-syntax-6.c: New test.
	* gcc.dg/c2x-attr-syntax-7.c: Likewise.
---
 gcc/c/c-parser.cc                        | 68 ++++++++++++++++++------
 gcc/doc/extend.texi                      | 27 ++++++++--
 gcc/testsuite/gcc.dg/c2x-attr-syntax-6.c | 50 +++++++++++++++++
 gcc/testsuite/gcc.dg/c2x-attr-syntax-7.c | 48 +++++++++++++++++
 4 files changed, 173 insertions(+), 20 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/c2x-attr-syntax-6.c
 create mode 100644 gcc/testsuite/gcc.dg/c2x-attr-syntax-7.c

diff --git a/gcc/c/c-parser.cc b/gcc/c/c-parser.cc
index 33fe7b115ff..82e56b28446 100644
--- a/gcc/c/c-parser.cc
+++ b/gcc/c/c-parser.cc
@@ -5390,10 +5390,18 @@ c_parser_balanced_token_sequence (c_parser *parser)
      ( balanced-token-sequence[opt] )
 
    Keywords are accepted as identifiers for this purpose.
-*/
+
+   As an extension, we permit an attribute-specifier to be:
+
+     [ [ __extension__ attribute-list ] ]
+
+   Two colons are then accepted as a synonym for ::.  No attempt is made
+   to check whether the colons are immediately adjacent.  LOOSE_SCOPE_P
+   indicates whether this relaxation is in effect.  */
 
 static tree
-c_parser_std_attribute (c_parser *parser, bool for_tm)
+c_parser_std_attribute (c_parser *parser, bool for_tm,
+			bool loose_scope_p = false)
 {
   c_token *token = c_parser_peek_token (parser);
   tree ns, name, attribute;
@@ -5406,9 +5414,18 @@ c_parser_std_attribute (c_parser *parser, bool for_tm)
     }
   name = canonicalize_attr_name (token->value);
   c_parser_consume_token (parser);
-  if (c_parser_next_token_is (parser, CPP_SCOPE))
+  if (c_parser_next_token_is (parser, CPP_SCOPE)
+      || (loose_scope_p
+	  && c_parser_next_token_is (parser, CPP_COLON)
+	  && c_parser_peek_token (parser)->type == CPP_COLON))
     {
       ns = name;
+      if (c_parser_next_token_is (parser, CPP_COLON))
+	{
+	  c_parser_consume_token (parser);
+	  if (!c_parser_next_token_is (parser, CPP_COLON))
+	    gcc_unreachable ();
+	}
       c_parser_consume_token (parser);
       token = c_parser_peek_token (parser);
       if (token->type != CPP_NAME && token->type != CPP_KEYWORD)
@@ -5481,19 +5498,9 @@ c_parser_std_attribute (c_parser *parser, bool for_tm)
 }
 
 static tree
-c_parser_std_attribute_specifier (c_parser *parser, bool for_tm)
+c_parser_std_attribute_list (c_parser *parser, bool for_tm,
+			     bool loose_scope_p = false)
 {
-  location_t loc = c_parser_peek_token (parser)->location;
-  if (!c_parser_require (parser, CPP_OPEN_SQUARE, "expected %<[%>"))
-    return NULL_TREE;
-  if (!c_parser_require (parser, CPP_OPEN_SQUARE, "expected %<[%>"))
-    {
-      c_parser_skip_until_found (parser, CPP_CLOSE_SQUARE, "expected %<]%>");
-      return NULL_TREE;
-    }
-  if (!for_tm)
-    pedwarn_c11 (loc, OPT_Wpedantic,
-		 "ISO C does not support %<[[]]%> attributes before C2X");
   tree attributes = NULL_TREE;
   while (true)
     {
@@ -5505,7 +5512,7 @@ c_parser_std_attribute_specifier (c_parser *parser, bool for_tm)
 	  c_parser_consume_token (parser);
 	  continue;
 	}
-      tree attribute = c_parser_std_attribute (parser, for_tm);
+      tree attribute = c_parser_std_attribute (parser, for_tm, loose_scope_p);
       if (attribute != error_mark_node)
 	{
 	  TREE_CHAIN (attribute) = attributes;
@@ -5514,6 +5521,35 @@ c_parser_std_attribute_specifier (c_parser *parser, bool for_tm)
       if (c_parser_next_token_is_not (parser, CPP_COMMA))
 	break;
     }
+  return attributes;
+}
+
+static tree
+c_parser_std_attribute_specifier (c_parser *parser, bool for_tm)
+{
+  location_t loc = c_parser_peek_token (parser)->location;
+  if (!c_parser_require (parser, CPP_OPEN_SQUARE, "expected %<[%>"))
+    return NULL_TREE;
+  if (!c_parser_require (parser, CPP_OPEN_SQUARE, "expected %<[%>"))
+    {
+      c_parser_skip_until_found (parser, CPP_CLOSE_SQUARE, "expected %<]%>");
+      return NULL_TREE;
+    }
+  tree attributes;
+  if (c_parser_next_token_is_keyword (parser, RID_EXTENSION))
+    {
+      auto ext = disable_extension_diagnostics ();
+      c_parser_consume_token (parser);
+      attributes = c_parser_std_attribute_list (parser, for_tm, true);
+      restore_extension_diagnostics (ext);
+    }
+  else
+    {
+      if (!for_tm)
+	pedwarn_c11 (loc, OPT_Wpedantic,
+		     "ISO C does not support %<[[]]%> attributes before C2X");
+      attributes = c_parser_std_attribute_list (parser, for_tm);
+    }
   c_parser_skip_until_found (parser, CPP_CLOSE_SQUARE, "expected %<]%>");
   c_parser_skip_until_found (parser, CPP_CLOSE_SQUARE, "expected %<]%>");
   return nreverse (attributes);
diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
index 73a997276cb..d764e44cbce 100644
--- a/gcc/doc/extend.texi
+++ b/gcc/doc/extend.texi
@@ -11914,10 +11914,29 @@ macros to replace them with the customary keywords.  It looks like this:
 @findex __extension__
 @opindex pedantic
 @option{-pedantic} and other options cause warnings for many GNU C extensions.
-You can
-prevent such warnings within one expression by writing
-@code{__extension__} before the expression.  @code{__extension__} has no
-effect aside from this.
+You can suppress such warnings using the keyword @code{__extension__}.
+Specifically:
+
+@itemize @bullet
+@item
+Writing @code{__extension__} before an expression prevents warnings
+about extensions within that expression.
+
+@item
+In C, writing:
+
+@smallexample
+[[__extension__ @dots{}]]
+@end smallexample
+
+suppresses warnings about using @samp{[[]]} attributes in C versions
+that predate C2X@.  Since the scope token @samp{::} is not a single
+lexing token in earlier versions of C, this construct also allows two colons
+to be used in place of @code{::}.  GCC does not check whether the two
+colons are immediately adjacent.
+@end itemize
+
+@code{__extension__} has no effect aside from this.
 
 @node Incomplete Enums
 @section Incomplete @code{enum} Types
diff --git a/gcc/testsuite/gcc.dg/c2x-attr-syntax-6.c b/gcc/testsuite/gcc.dg/c2x-attr-syntax-6.c
new file mode 100644
index 00000000000..01eff1f32df
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/c2x-attr-syntax-6.c
@@ -0,0 +1,50 @@
+/* Test C2x attribute syntax: use of __extension__ in C11 mode.  */
+/* { dg-do compile } */
+/* { dg-options "-std=c11 -pedantic-errors" } */
+
+#define FOO ::
+#define BAR :
+
+typedef int [[__extension__ gnu :: vector_size (4)]] g1;
+typedef int [[__extension__ gnu : : vector_size (4)]] g2;
+typedef int [[__extension__ gnu FOO vector_size (4)]] g3;
+typedef int [[__extension__ gnu BAR BAR vector_size (4)]] g4;
+typedef int [[__extension__ gnu :: vector_size (sizeof (void (*)(...)))]] g5;
+typedef int [[__extension__]] g6;
+typedef int [[__extension__,]] g7;
+typedef int [[__extension__, ,,,, ,, ,]] g8;
+[[__extension__ deprecated]] int g9 ();
+[[__extension__ nodiscard]] int g10 ();
+[[__extension__ noreturn]] void g11 ();
+
+int
+cases (int x)
+{
+  switch (x)
+    {
+    case 1:
+    case 2:
+    case 4:
+      x += 1;
+      [[__extension__ fallthrough]];
+    case 19:
+    case 33:
+      x *= 2;
+      [[fallthrough]];  /* { dg-error {attributes before C2X} } */
+    case 99:
+      return x;
+
+    default:
+      return 0;
+    }
+}
+
+typedef int [[__extension__ vector_size (4)]] b1; /* { dg-error {'vector_size' attribute ignored} } */
+typedef int [[__extension__ __extension__]] b2; /* { dg-error {'extension' attribute ignored} } */
+typedef int [[__extension__ unknown_attribute]] b3; /* { dg-error {'unknown_attribute' attribute ignored} } */
+typedef int [[gnu::vector_size(4)]] b4; /* { dg-error {expected '\]' before ':'} } */
+/* { dg-error {'gnu' attribute ignored} "" { target *-*-* } .-1 } */
+/* { dg-error {attributes before C2X} "" { target *-*-* } .-2 } */
+typedef int [[gnu : : vector_size(4)]] b5; /* { dg-error {expected '\]' before ':'} } */
+/* { dg-error {'gnu' attribute ignored} "" { target *-*-* } .-1 } */
+/* { dg-error {attributes before C2X} "" { target *-*-* } .-2 } */
diff --git a/gcc/testsuite/gcc.dg/c2x-attr-syntax-7.c b/gcc/testsuite/gcc.dg/c2x-attr-syntax-7.c
new file mode 100644
index 00000000000..cfa9b00cfc9
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/c2x-attr-syntax-7.c
@@ -0,0 +1,48 @@
+/* Test C2x attribute syntax: use of __extension__ in C11 mode.  */
+/* { dg-do compile } */
+/* { dg-options "-std=c2x -pedantic-errors -Wc11-c2x-compat" } */
+
+#define FOO ::
+#define BAR :
+
+typedef int [[__extension__ gnu :: vector_size (4)]] g1;
+typedef int [[__extension__ gnu : : vector_size (4)]] g2;
+typedef int [[__extension__ gnu FOO vector_size (4)]] g3;
+typedef int [[__extension__ gnu BAR BAR vector_size (4)]] g4;
+typedef int [[__extension__ gnu :: vector_size (sizeof (void (*)(...)))]] g5;
+typedef int [[__extension__]] g6;
+typedef int [[__extension__,]] g7;
+typedef int [[__extension__, ,,,, ,, ,]] g8;
+[[__extension__ deprecated]] int g9 ();
+[[__extension__ nodiscard]] int g10 ();
+[[__extension__ noreturn]] void g11 ();
+
+int
+cases (int x)
+{
+  switch (x)
+    {
+    case 1:
+    case 2:
+    case 4:
+      x += 1;
+      [[__extension__ fallthrough]];
+    case 19:
+    case 33:
+      x *= 2;
+      [[fallthrough]];  /* { dg-warning {attributes before C2X} } */
+    case 99:
+      return x;
+
+    default:
+      return 0;
+    }
+}
+
+typedef int [[__extension__ vector_size (4)]] b1; /* { dg-error {'vector_size' attribute ignored} } */
+typedef int [[__extension__ __extension__]] b2; /* { dg-error {'extension' attribute ignored} } */
+typedef int [[__extension__ unknown_attribute]] b3; /* { dg-error {'unknown_attribute' attribute ignored} } */
+typedef int [[gnu::vector_size(4)]] b4; /* { dg-warning {attributes before C2X} } */
+typedef int [[gnu : : vector_size(4)]] b5; /* { dg-error {expected '\]' before ':'} } */
+/* { dg-error {'gnu' attribute ignored} "" { target *-*-* } .-1 } */
+/* { dg-warning {attributes before C2X} "" { target *-*-* } .-2 } */
-- 
2.25.1


  reply	other threads:[~2023-08-17 11:24 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-07-14 15:56 [WIP RFC] Add support for keyword-based attributes Richard Sandiford
2023-07-14 17:17 ` Jakub Jelinek
2023-07-16 10:50   ` Richard Sandiford
2023-07-17 13:39     ` Jason Merrill
2023-07-17 14:06       ` Richard Sandiford
2023-07-14 21:14 ` Nathan Sidwell
2023-07-16 10:18   ` Richard Sandiford
2023-07-17  6:38 ` Richard Biener
2023-07-17  8:21   ` Richard Sandiford
2023-07-17  9:05     ` Richard Biener
2023-07-17 13:53     ` Michael Matz
2023-07-21 23:25       ` Joseph Myers
2023-08-16 10:36         ` Richard Sandiford
2023-08-16 13:22           ` Joseph Myers
2023-08-17 11:24             ` Richard Sandiford [this message]
2023-08-17 17:07               ` [PATCH] c: Add support for [[__extension__ ...]] Richard Biener
2023-08-17 18:36                 ` Richard Sandiford
2023-08-18  9:51               ` Richard Sandiford
2023-08-18 19:51                 ` Joseph Myers

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=mptlee9j4hr.fsf_-_@arm.com \
    --to=richard.sandiford@arm.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=joseph@codesourcery.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).