This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

-Wmissing-attributes: avoid duplicates and false positives


Hello, Martin,

The initial patch for PR 81824 fixed one of the possibilities of
-Wmissing-attributes reporting duplicates, namely, if TMPL had an
attribute in ATTRLIST that was missing from DECL's decl and type
attribute lists, both being non-empty.

Another possibility of duplicate reporting remained: when an attribute
in ATTRLIST is present in both decl and type attribute lists of TMPL,
and absent from DECL's attribute lists, it is reported once for each
of TMPL's lists.

The implementation still allowed for false positives: an attribute in
ATTRLIST that is present in TMPL will be regarded as missing as soon
as it is not found in DECL's decl attribute list, even if it is later
found in DECL's type attribute list.

This patch fixes both problems, so that an attribute from ATTRLIST
that is present in any of TMPL's lists will be reported at most once,
and only if it is missing from both DECL's lists.


Now, I realize there are some attributes that are only acceptable for
types, and some that are only acceptable for function declarations.
ISTM that some even have different meanings depending on whether
they're associated with types or declarations.  There's room for an
argument for checking only corresponding lists, for at least some of
the attributes.  AFAICT it doesn't apply to -Wmissing-attributes, that
are either acceptable in either list, or only in the FUNCTION_DECL
list, so I'm leaving it at that in the hope that it doesn't apply to
any other users of decls_mismatched_attributes either.


Regstrapping on x86_64-linux-gnu.  Ok to install if it passes?


for  gcc/ChangeLog

	PR middle-end/81824
	* attribs.c (decls_mismatched_attributes): Avoid duplicates
	and false positives.

for  gcc/testsuite/ChangeLog

	PR middle-end/81824
	* g++.dg/Wmissing-attributes-1.C: New.
---
 gcc/attribs.c                                |   14 +++++--
 gcc/testsuite/g++.dg/Wmissing-attributes-1.C |   55 ++++++++++++++++++++++++++
 2 files changed, 65 insertions(+), 4 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/Wmissing-attributes-1.C

diff --git a/gcc/attribs.c b/gcc/attribs.c
index 8e5401655972..f4777c6a8233 100644
--- a/gcc/attribs.c
+++ b/gcc/attribs.c
@@ -1931,15 +1931,19 @@ decls_mismatched_attributes (tree tmpl, tree decl, tree attrlist,
 	  if (!has_attribute (tmpls[j], tmpl_attrs[j], blacklist[i]))
 	    continue;
 
+	  bool found = false;
 	  unsigned kmax = 1 + !!decl_attrs[1];
 	  for (unsigned k = 0; k != kmax; ++k)
 	    {
 	      if (has_attribute (decls[k], decl_attrs[k], blacklist[i]))
-		break;
-
-	      if (!k && kmax > 1)
-		continue;
+		{
+		  found = true;
+		  break;
+		}
+	    }
 
+	  if (!found)
+	    {
 	      if (nattrs)
 		pp_string (attrstr, ", ");
 	      pp_begin_quote (attrstr, pp_show_color (global_dc->printer));
@@ -1947,6 +1951,8 @@ decls_mismatched_attributes (tree tmpl, tree decl, tree attrlist,
 	      pp_end_quote (attrstr, pp_show_color (global_dc->printer));
 	      ++nattrs;
 	    }
+
+	  break;
 	}
     }
 
diff --git a/gcc/testsuite/g++.dg/Wmissing-attributes-1.C b/gcc/testsuite/g++.dg/Wmissing-attributes-1.C
new file mode 100644
index 000000000000..56d28b339e17
--- /dev/null
+++ b/gcc/testsuite/g++.dg/Wmissing-attributes-1.C
@@ -0,0 +1,55 @@
+// { dg-do compile }
+// { dg-options "-Wmissing-attributes" }
+
+#define ATTR(list)   __attribute__ (list)
+
+/* Type attributes are normally absent in template functions, and the
+   mere presence of any such attribute used to cause the
+   -Wmissing-attributes checks, that checked for attributes typically
+   associated with functions rather than types, to report any missing
+   attributes twice: once for the specialization attribute list, once
+   for its type attribute list.
+
+   If we force any of the checked-for attributes to be associated with
+   types rather than functions, even later implementations that fixed
+   the duplicate reporting problem above would still report them as
+   missing (in the function attribute list) even when present (in the
+   type attribute list).  */
+typedef void* ATTR ((alloc_size (1))) f_type (int);
+
+template <class T>
+f_type
+ATTR ((malloc))
+missing_malloc;            // { dg-message "missing primary template attribute .malloc." }
+
+template <>
+f_type
+missing_malloc<char>;      // { dg-warning "explicit specialization .\[^\n\r\]+. may be missing attributes" }
+
+
+/* Check that even an attribute that appears in both lists (decl and
+   type) in a template declaration is reported as missing only
+   once.  */
+
+template <class T>
+f_type
+ATTR ((alloc_size (1))) // In both attr lists, decl's and type's.
+missing_alloc_size;            // { dg-message "missing primary template attribute .alloc_size." }
+
+template <>
+void *
+missing_alloc_size<char>(int); // { dg-warning "explicit specialization .\[^\n\r\]+. may be missing attributes" }
+
+
+/* Check that even an attribute that appears in both lists (decl and
+   type) is not report as missing if it's present only in the type
+   list.  */
+
+template <class T>
+f_type
+ATTR ((alloc_size (1))) // In both attr lists, decl's and type's.
+missing_nothing;
+
+template <>
+f_type
+missing_nothing<char>;

-- 
Alexandre Oliva, freedom fighter  he/him   https://FSFLA.org/blogs/lxo
Be the change, be Free!                 FSF Latin America board member
GNU Toolchain Engineer                        Free Software Evangelist
Hay que enGNUrecerse, pero sin perder la terGNUra jamás - Che GNUevara


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]