[Bug c/13856] New: hidden support broken with builtin functions

[Bug c/13856] New: hidden support broken with builtin functions

[aj at gcc dot gnu dot org]

With current GCC 3.4 I cannot build glibc anymore, it fails with:
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_strncmp'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_strftime'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_fputs_unlocked'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_fwrite_unlocked'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_memmove'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_strncpy'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI__exit'
/builds/glibc/3.4/libc.so.6: undefined reference to `__GI_strstr'
collect2: ld returned 1 exit status

Let's look at fputs_unlocked:

The 3.3 branch compiles the file to use the following symbols:

$ nm libio/iofputs_u.os
                 U _GLOBAL_OFFSET_TABLE_
0000000000000000 T __GI_fputs_unlocked
0000000000000000 T fputs_unlocked
                 U strlen

3.4 compiles to:
$ nm libio/iofputs_u.os
                 U __GI_fputs_unlocked
                 U __GI_strlen
0000000000000000 T fputs_unlocked


Looking further into this, I see the following difference between 3.4
and 3.3 in the assembler files:

3.3 branch outputs:
        .globl fputs_unlocked ; .set fputs_unlocked, __GI_fputs_unlocked
#NO_APP
        .p2align 4,,15
.globl __GI_fputs_unlocked
        .hidden __GI_fputs_unlocked
        .type   __GI_fputs_unlocked, @function
__GI_fputs_unlocked:

But 3.4 does:
        .globl fputs_unlocked ; .set fputs_unlocked, __GI_fputs_unlocked
#NO_APP
        .p2align 4,,15
.globl fputs_unlocked
        .type   fputs_unlocked, @function
fputs_unlocked:

I'm appending a small test program for this.

Looking at the produced assembler files, you'll see the difference directly.

-- 
           Summary: hidden support broken with builtin functions
           Product: gcc
           Version: 3.4.0
            Status: UNCONFIRMED
          Severity: critical
          Priority: P2
         Component: c
        AssignedTo: zack at codesourcery dot com
        ReportedBy: aj at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: x86_64-linux-gnu
  GCC host triplet: x86_64-linux-gnu
GCC target triplet: x86_64-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] hidden support broken with builtin functions

[aj at gcc dot gnu dot org]

------- Additional Comments From aj at gcc dot gnu dot org  2004-01-25 12:29 -------
Created an attachment (id=5572)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=5572&action=view)
Testcase that shows failure


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] hidden support broken with builtin functions

[aj at gcc dot gnu dot org]

------- Additional Comments From aj at gcc dot gnu dot org  2004-01-25 12:30 -------
Note, this shows on x86-64 but I see nothing target specific.  I get the same
results on i686-linux-gnu

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4 Regression] hidden support broken with builtin functions

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-25 15:51 -------
Confirmed.
Here is a shorter testcase:
int fputs_unlocked(__const char *__restrict __s, int *__restrict __stream) __asm__ ("" 
"__GI_fputs_unlocked") __attribute__ ((visibility ("hidden")));
int
fputs_unlocked (str, fp)
      const char *str;
      int *fp;
{
}
__asm__ (".globl fputs_unlocked ; .set fputs_unlocked, __GI_fputs_unlocked");

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2004-01-25 15:51:55
               date|                            |
            Summary|hidden support broken with  |[3.4 Regression] hidden
                   |builtin functions           |support broken with builtin
                   |                            |functions
   Target Milestone|---                         |3.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[aj at gcc dot gnu dot org]

------- Additional Comments From aj at gcc dot gnu dot org  2004-01-26 06:30 -------
(In reply to comment #3)
> Confirmed.
> Here is a shorter testcase:
> int fputs_unlocked(__const char *__restrict __s, int *__restrict __stream)
__asm__ ("" 
> "__GI_fputs_unlocked") __attribute__ ((visibility ("hidden")));
> int
> fputs_unlocked (str, fp)
>       const char *str;
>       int *fp;

Are you sure this works without warnings?  Is the builtin just declared with
a pointer as second argument?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|zack at codesourcery dot com|zack at gcc dot gnu dot org
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[uweigand at gcc dot gnu dot org]

------- Additional Comments From uweigand at gcc dot gnu dot org  2004-02-03 20:19 -------
We're seeing this problem on s390 as well.  (B.t.w. this doesn't 
appear to have anything to do with "builtin functions", does it?) 
 
>From my initial experiments, it would appear that the problem 
is caused by the K&R-style function definition -- when using 
a C89-style definition like 
int fputs_unlocked (const char *str, int *fp) 
everything works as expected. 
 
Maybe something goes wrong when merging attributes between the 
declaration and definition type nodes in case the latter is K&R? 
 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[uweigand at gcc dot gnu dot org]

------- Additional Comments From uweigand at gcc dot gnu dot org  2004-02-03 20:47 -------
Sorry, I was wrong: of course this is related to builtin functions. 
 
In fact, the current behaviour is directly caused by Zack's patch 
to fix PR 18314 (http://gcc.gnu.org/ml/gcc-patches/2004-01/msg02475.html). 
 
What happens is that 
 - first, we have a builtin declaration 
 - second, we get an excplicit declaration which matches the builtin; 
   it provides new attributes that are merged into the builtin decl 
 - third, we get a K&R definition; after Zack's patch K&R definitions 
   are considered to always override the builtin decl, hence the  
   builtin decl is thrown away together with the changed attributes 
 
Is this behaviour wrong?  Is the source code valid in the first place? 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[zack at gcc dot gnu dot org]

------- Additional Comments From zack at gcc dot gnu dot org  2004-02-03 21:25 -------
The source code is fine.  I think what needs to happen is that the various
special cases for builtin functions should only apply if the builtin is still
invisible (C_DECL_INVISIBLE (olddecl)) -- if it is visible, then there was a
declaration in the source code, and following declarations must agree with that one.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-02-08 01:52 -------
Subject: Bug 13856

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_4-branch
Changes by:	zack@gcc.gnu.org	2004-02-08 01:52:52

Modified files:
	gcc            : ChangeLog c-decl.c cgraphunit.c dwarf2out.c 
	                 tree-inline.c tree.h 
	gcc/cp         : ChangeLog decl.c optimize.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: visibility-8.c 

Log message:
	Bug 13856
	* c-decl.c (diagnose_mismatched_decls): Only give special
	treatment when olddecl is DECL_BUILT_IN, if C_DECL_INVISIBLE
	is also true.
	(merge_decls): Don't clear DECL_BUILT_IN_CLASS and
	DECL_FUNCTION_CODE when defining a built-in function.
	Don't update DECL_ESTIMATED_INSNS.
	* dwarf2out.c (dwarf2out_decl): Don't ignore built-in
	FUNCTION_DECLs.
	* tree.h: Delete DECL_ESTIMATED_INSNS.
	* tree-inline.c (struct inline_data): Delete inlined_insns field.
	(expand_call_inline, optimize_inline_calls): Don't update
	DECL_ESTIMATED_INSNS nor inlined_insns.
	* cgraphunit.c (cgraph_analyze_function): Don't update
	DECL_ESTIMATED_INSNS.
	cp:
	* optimize.c (maybe_clone_body): Don't update DECL_ESTIMATED_INSNS.
	* decl.c (duplicate_decls, start_function): Likewise.
	testsuite:
	* gcc.dg/visibility-8.c: New testcase.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=2.2326.2.158&r2=2.2326.2.159
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/c-decl.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.470.4.4&r2=1.470.4.5
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cgraphunit.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.46.2.4&r2=1.46.2.5
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/dwarf2out.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.478.2.3&r2=1.478.2.4
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-inline.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.90.4.1&r2=1.90.4.2
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree.h.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.458.2.3&r2=1.458.2.4
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3892.2.45&r2=1.3892.2.46
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.1174.2.7&r2=1.1174.2.8
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/optimize.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.103.4.1&r2=1.103.4.2
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3389.2.60&r2=1.3389.2.61
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/visibility-8.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[zack at codesourcery dot com]

------- Additional Comments From zack at codesourcery dot com  2004-02-08 01:53 -------
Subject: c-decl.c rewrite: fix bug 13856


Bug 13856 reports that a visibility:hidden attribute gets lost when
applied to the definition of a built-in function (in glibc).  The fix
is regrettably more invasive than it ought to be.  First, don't give
special treatment in duplicate_decls to functions that have been
explicitly declared just because they're built-in.  Second, do not
forget that functions are built-in just because they've been defined
(assuming compatible signatures).  Third, don't ignore builtins in
dwarf2out.c (they will get ignored below if they have no definitions,
anyway).  Last, do not overwrite DECL_FUNCTION_CODE from language-
independent code. DECL_ESTIMATED_INSNS turns out to be write-only, so
this is no hardship.

bootstrapped i686-linux, applied 3.4 branch and shortly to mainline.

zw

        Bug 13856
        * c-decl.c (diagnose_mismatched_decls): Only give special
        treatment when olddecl is DECL_BUILT_IN, if C_DECL_INVISIBLE 
        is also true.
        (merge_decls): Don't clear DECL_BUILT_IN_CLASS and
        DECL_FUNCTION_CODE when defining a built-in function.
        Don't update DECL_ESTIMATED_INSNS.
        * dwarf2out.c (dwarf2out_decl): Don't ignore built-in
        FUNCTION_DECLs.
        * tree.h: Delete DECL_ESTIMATED_INSNS.
        * tree-inline.c (struct inline_data): Delete inlined_insns field.
        (expand_call_inline, optimize_inline_calls): Don't update
        DECL_ESTIMATED_INSNS nor inlined_insns.
        * cgraphunit.c (cgraph_analyze_function): Don't update
        DECL_ESTIMATED_INSNS.
cp:
        * optimize.c (maybe_clone_body): Don't update DECL_ESTIMATED_INSNS.
        * decl.c (duplicate_decls, start_function): Likewise.
testsuite:
        * gcc.dg/visibility-8.c: New testcase.

===================================================================
Index: c-decl.c
--- c-decl.c	8 Feb 2004 00:52:42 -0000
+++ c-decl.c	8 Feb 2004 01:09:37 -0000
@@ -939,7 +939,8 @@
      unless OLDDECL is a builtin.  OLDDECL will be discarded in any case.  */
   if (TREE_CODE (olddecl) != TREE_CODE (newdecl))
     {
-      if (TREE_CODE (olddecl) != FUNCTION_DECL || !DECL_BUILT_IN (olddecl))
+      if (TREE_CODE (olddecl) != FUNCTION_DECL
+          || !DECL_BUILT_IN (olddecl) || !C_DECL_INVISIBLE (olddecl))
 	{
 	  error ("%J'%D' redeclared as different kind of symbol",
 		 newdecl, newdecl);
@@ -956,7 +957,8 @@
 
   if (!comptypes (oldtype, newtype, COMPARE_STRICT))
     {
-      if (TREE_CODE (olddecl) == FUNCTION_DECL && DECL_BUILT_IN (olddecl))
+      if (TREE_CODE (olddecl) == FUNCTION_DECL
+	  && DECL_BUILT_IN (olddecl) && C_DECL_INVISIBLE (olddecl))
 	{
 	  /* Accept harmless mismatch in function types.
 	     This is for the ffs and fprintf builtins.  */
@@ -1034,6 +1036,7 @@
 	 can't validate the argument list) the built-in definition is
 	 overridden, but optionally warn this was a bad choice of name.  */
       if (DECL_BUILT_IN (olddecl)
+	  && C_DECL_INVISIBLE (olddecl)
 	  && (!TREE_PUBLIC (newdecl)
 	      || (DECL_INITIAL (newdecl)
 		  && !TYPE_ARG_TYPES (TREE_TYPE (newdecl)))))
@@ -1428,20 +1431,9 @@
 
       if (DECL_BUILT_IN (olddecl))
 	{
-	  /* Get rid of any built-in function if we have a function
-	     definition.  */
-	  if (new_is_definition)
-	    {
-	      TREE_TYPE (olddecl) = TREE_TYPE (newdecl);
-	      DECL_BUILT_IN_CLASS (olddecl) = NOT_BUILT_IN;
-	    }
-	  else
-	    {
-	      /* If redeclaring a builtin function, and not a definition,
-		 it stays built in.  */
-	      DECL_BUILT_IN_CLASS (newdecl) = DECL_BUILT_IN_CLASS (olddecl);
-	      DECL_FUNCTION_CODE (newdecl) = DECL_FUNCTION_CODE (olddecl);
-	    }
+	  /* If redeclaring a builtin function, it stays built in.  */
+	  DECL_BUILT_IN_CLASS (newdecl) = DECL_BUILT_IN_CLASS (olddecl);
+	  DECL_FUNCTION_CODE (newdecl) = DECL_FUNCTION_CODE (olddecl);
 	}
 
       /* Also preserve various other info from the definition.  */
@@ -1451,7 +1443,6 @@
 	  DECL_INITIAL (newdecl) = DECL_INITIAL (olddecl);
 	  DECL_SAVED_INSNS (newdecl) = DECL_SAVED_INSNS (olddecl);
 	  DECL_SAVED_TREE (newdecl) = DECL_SAVED_TREE (olddecl);
-	  DECL_ESTIMATED_INSNS (newdecl) = DECL_ESTIMATED_INSNS (olddecl);
 	  DECL_ARGUMENTS (newdecl) = DECL_ARGUMENTS (olddecl);
 
 	  /* Set DECL_INLINE on the declaration if we've got a body
===================================================================
Index: dwarf2out.c
--- dwarf2out.c	5 Feb 2004 22:01:33 -0000	1.478.2.3
+++ dwarf2out.c	8 Feb 2004 01:09:38 -0000
@@ -12153,12 +12153,6 @@ dwarf2out_decl (tree decl)
       return;
 
     case FUNCTION_DECL:
-      /* Ignore this FUNCTION_DECL if it refers to a builtin declaration of a
-	 builtin function.  Explicit programmer-supplied declarations of
-	 these same functions should NOT be ignored however.  */
-      if (DECL_EXTERNAL (decl) && DECL_BUILT_IN (decl))
-	return;
-
       /* What we would really like to do here is to filter out all mere
 	 file-scope declarations of file-scope functions which are never
 	 referenced later within this translation unit (and keep all of ones
===================================================================
Index: tree.h
--- tree.h	5 Feb 2004 22:01:36 -0000	1.458.2.3
+++ tree.h	8 Feb 2004 01:09:39 -0000
@@ -1627,13 +1627,6 @@ struct tree_type GTY(())
 #define DECL_POINTER_ALIAS_SET_KNOWN_P(NODE) \
   (DECL_POINTER_ALIAS_SET (NODE) != - 1)
 
-/* In a FUNCTION_DECL for which DECL_BUILT_IN does not hold, this is
-   the approximate number of statements in this function.  There is
-   no need for this number to be exact; it is only used in various
-   heuristics regarding optimization.  */
-#define DECL_ESTIMATED_INSNS(NODE) \
-  (FUNCTION_DECL_CHECK (NODE)->decl.u1.i)
-
 /* Nonzero for a decl which is at file scope.  */
 #define DECL_FILE_SCOPE_P(EXP) 					\
   (! DECL_CONTEXT (EXP)						\
===================================================================
Index: tree-inline.c
--- tree-inline.c	23 Jan 2004 23:36:03 -0000	1.90.4.1
+++ tree-inline.c	8 Feb 2004 01:09:39 -0000
@@ -95,9 +95,6 @@ typedef struct inline_data
   int in_target_cleanup_p;
   /* A list of the functions current function has inlined.  */
   varray_type inlined_fns;
-  /* The approximate number of instructions we have inlined in the
-     current call stack.  */
-  int inlined_insns;
   /* We use the same mechanism to build clones that we do to perform
      inlining.  However, there are a few places where we need to
      distinguish between those two situations.  This flag is true if
@@ -1569,11 +1566,6 @@ expand_call_inline (tree *tp, int *walk_
      the equivalent inlined version either.  */
   TREE_USED (*tp) = 1;
 
-  /* Our function now has more statements than it did before.  */
-  DECL_ESTIMATED_INSNS (VARRAY_TREE (id->fns, 0)) += DECL_ESTIMATED_INSNS (fn);
-  /* For accounting, subtract one for the saved call/ret.  */
-  id->inlined_insns += DECL_ESTIMATED_INSNS (fn) - 1;
-
   /* Update callgraph if needed.  */
   if (id->decl)
     {
@@ -1590,11 +1582,6 @@ expand_call_inline (tree *tp, int *walk_
   }
   VARRAY_POP (id->fns);
 
-  /* If we've returned to the top level, clear out the record of how
-     much inlining has been done.  */
-  if (VARRAY_ACTIVE_SIZE (id->fns) == id->first_inlined_fn)
-    id->inlined_insns = 0;
-
   /* Don't walk into subtrees.  We've already handled them above.  */
   *walk_subtrees = 0;
 
@@ -1634,9 +1621,6 @@ optimize_inline_calls (tree fn)
   /* Don't allow recursion into FN.  */
   VARRAY_TREE_INIT (id.fns, 32, "fns");
   VARRAY_PUSH_TREE (id.fns, fn);
-  if (!DECL_ESTIMATED_INSNS (fn))
-    DECL_ESTIMATED_INSNS (fn) 
-      = (*lang_hooks.tree_inlining.estimate_num_insns) (fn);
   /* Or any functions that aren't finished yet.  */
   prev_fn = NULL_TREE;
   if (current_function_decl)
===================================================================
Index: cgraphunit.c
--- cgraphunit.c	1 Feb 2004 13:01:09 -0000	1.46.2.4
+++ cgraphunit.c	8 Feb 2004 01:09:38 -0000
@@ -327,10 +327,9 @@ cgraph_analyze_function (struct cgraph_n
   cgraph_create_edges (decl, DECL_SAVED_TREE (decl));
 
   node->local.inlinable = tree_inlinable_function_p (decl);
-  if (!DECL_ESTIMATED_INSNS (decl))
-    DECL_ESTIMATED_INSNS (decl)
+  if (!node->local.self_insns)
+    node->local.self_insns
       = (*lang_hooks.tree_inlining.estimate_num_insns) (decl);
-  node->local.self_insns = DECL_ESTIMATED_INSNS (decl);
   if (node->local.inlinable)
     node->local.disregard_inline_limits
       = (*lang_hooks.tree_inlining.disregard_inline_limits) (decl);
===================================================================
Index: cp/optimize.c
--- cp/optimize.c	25 Jan 2004 02:52:32 -0000	1.103.4.1
+++ cp/optimize.c	8 Feb 2004 01:09:39 -0000
@@ -232,10 +232,6 @@ maybe_clone_body (tree fn)
       /* Clone the body.  */
       clone_body (clone, fn, decl_map);
 
-      /* There are as many statements in the clone as in the
-	 original.  */
-      DECL_ESTIMATED_INSNS (clone) = DECL_ESTIMATED_INSNS (fn);
-
       /* Clean up.  */
       splay_tree_delete (decl_map);
 
===================================================================
Index: cp/decl.c
--- cp/decl.c	7 Feb 2004 09:43:01 -0000	1.1174.2.7
+++ cp/decl.c	8 Feb 2004 01:09:39 -0000
@@ -1853,8 +1853,6 @@ duplicate_decls (tree newdecl, tree oldd
 		 regardless of declaration matches.  */
 	      SET_DECL_RTL (newdecl, DECL_RTL (olddecl));
 	    }
-	  else
-	    DECL_ESTIMATED_INSNS (newdecl) = DECL_ESTIMATED_INSNS (olddecl);
 
 	  DECL_RESULT (newdecl) = DECL_RESULT (olddecl);
 	  /* Don't clear out the arguments if we're redefining a function.  */
@@ -10257,9 +10255,6 @@ start_function (tree declspecs, tree dec
 
   /* Start the statement-tree, start the tree now.  */
   begin_stmt_tree (&DECL_SAVED_TREE (decl1));
-
-  /* Don't double-count statements in templates.  */
-  DECL_ESTIMATED_INSNS (decl1) = 0;
 
   /* Let the user know we're compiling this function.  */
   announce_function (decl1);
===================================================================
Index: testsuite/gcc.dg/visibility-8.c
--- testsuite/gcc.dg/visibility-8.c	1 Jan 1970 00:00:00 -0000
+++ testsuite/gcc.dg/visibility-8.c	8 Feb 2004 01:09:41 -0000
@@ -0,0 +1,16 @@
+/* Test hidden visibility on built-in functions (for libc).  PR 13856.  */
+/* { dg-do compile } */
+/* { dg-options "-std=gnu99" } */
+/* { dg-require-visibility "" } */
+/* { dg-final { scan-assembler "\\.hidden.*__GI_fputs_unlocked" } } */
+
+int fputs_unlocked (const char *restrict, int *restrict)
+   __asm__ ("__GI_fputs_unlocked")
+   __attribute__ ((visibility ("hidden")));
+
+int
+fputs_unlocked (str, fp)
+     const char *str;
+     int *fp;
+{
+}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-02-08 01:59 -------
Subject: Bug 13856

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	zack@gcc.gnu.org	2004-02-08 01:59:12

Modified files:
	gcc            : ChangeLog c-common.c c-common.h c-decl.c 
	                 c-opts.c c-parse.in cgraphunit.c dwarf2out.c 
	                 tree-inline.c tree.h 
	gcc/cp         : ChangeLog decl.c name-lookup.c optimize.c 
	                 parser.c 
	gcc/testsuite  : ChangeLog 
	gcc/testsuite/gcc.c-torture/execute: string-opt-15.c 
	gcc/testsuite/gcc.dg: fwritable-strings-1.c 
Added files:
	gcc/testsuite/gcc.dg: visibility-8.c 

Log message:
	* c-common.c (shadow_warning): Delete.
	* c-common.h (free_parser_stacks, shadow_warning, sw_kind): Delete.
	* c-decl.c (warn_if_shadowing): Issue shadow warnings directly.
	* c-opts.c (c_common_parse_file): Don't call free_parser_stacks.
	* c-parse.in (free_parser_stacks): Delete.
	cp:
	* name-lookup.c (pushdecl): Issue shadow warnings directly.
	* parser.c (free_parser_stacks): Delete.
	testsuite:
	* gcc.c-torture/execute/string-opt-15.c: Define memcmp with void *
	arguments.
	* gcc.dg/fwritable-strings-1.c: Expect the deprecation notice.
	------
	Bug 13856
	* c-decl.c (diagnose_mismatched_decls): Only give special
	treatment when olddecl is DECL_BUILT_IN, if C_DECL_INVISIBLE
	is also true.
	(merge_decls): Don't clear DECL_BUILT_IN_CLASS and
	DECL_FUNCTION_CODE when defining a built-in function.
	Don't update DECL_ESTIMATED_INSNS.
	* dwarf2out.c (dwarf2out_decl): Don't ignore built-in
	FUNCTION_DECLs.
	* tree.h: Delete DECL_ESTIMATED_INSNS.
	* tree-inline.c (struct inline_data): Delete inlined_insns field.
	(expand_call_inline, optimize_inline_calls): Don't update
	DECL_ESTIMATED_INSNS nor inlined_insns.
	* cgraphunit.c (cgraph_analyze_function): Don't update
	DECL_ESTIMATED_INSNS.
	cp:
	* optimize.c (maybe_clone_body): Don't update DECL_ESTIMATED_INSNS.
	* decl.c (duplicate_decls, start_function): Likewise.
	testsuite:
	* gcc.dg/visibility-8.c: New testcase.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.2704&r2=2.2705
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/c-common.c.diff?cvsroot=gcc&r1=1.480&r2=1.481
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/c-common.h.diff?cvsroot=gcc&r1=1.219&r2=1.220
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/c-decl.c.diff?cvsroot=gcc&r1=1.475&r2=1.476
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/c-opts.c.diff?cvsroot=gcc&r1=1.100&r2=1.101
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/c-parse.in.diff?cvsroot=gcc&r1=1.197&r2=1.198
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cgraphunit.c.diff?cvsroot=gcc&r1=1.50&r2=1.51
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/dwarf2out.c.diff?cvsroot=gcc&r1=1.491&r2=1.492
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-inline.c.diff?cvsroot=gcc&r1=1.91&r2=1.92
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree.h.diff?cvsroot=gcc&r1=1.463&r2=1.464
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.3946&r2=1.3947
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/decl.c.diff?cvsroot=gcc&r1=1.1182&r2=1.1183
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/name-lookup.c.diff?cvsroot=gcc&r1=1.36&r2=1.37
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/optimize.c.diff?cvsroot=gcc&r1=1.104&r2=1.105
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/parser.c.diff?cvsroot=gcc&r1=1.170&r2=1.171
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.3477&r2=1.3478
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.c-torture/execute/string-opt-15.c.diff?cvsroot=gcc&r1=1.2&r2=1.3
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/visibility-8.c.diff?cvsroot=gcc&r1=1.1&r2=1.2
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/fwritable-strings-1.c.diff?cvsroot=gcc&r1=1.1&r2=1.2



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[zack at gcc dot gnu dot org]

------- Additional Comments From zack at gcc dot gnu dot org  2004-02-08 02:01 -------
fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[aj at suse dot de]

------- Additional Comments From aj at suse dot de  2004-02-08 06:41 -------
Subject: Re: c-decl.c rewrite: fix bug 13856

"Zack Weinberg" <zack@codesourcery.com> writes:

> [...]
>         Bug 13856

You should use "PR c/13856", so that bugzilla picks it up
automatically,

Andreas


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[zack at codesourcery dot com]

------- Additional Comments From zack at codesourcery dot com  2004-02-08 07:10 -------
Subject: Re: c-decl.c rewrite: fix bug 13856

Andreas Jaeger <aj@suse.de> writes:

> "Zack Weinberg" <zack@codesourcery.com> writes:
>
>> [...]
>>         Bug 13856
>
> You should use "PR c/13856", so that bugzilla picks it up
> automatically,

This notation works too.

zw


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856

[Bug c/13856] [3.4/3.5 Regression] hidden support broken with builtin functions

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2004-02-08 07:25 -------
Subject: Re: c-decl.c rewrite: fix bug 13856

"Zack Weinberg" <zack@codesourcery.com> writes:

| Andreas Jaeger <aj@suse.de> writes:
| 
| > "Zack Weinberg" <zack@codesourcery.com> writes:
| >
| >> [...]
| >>         Bug 13856
| >
| > You should use "PR c/13856", so that bugzilla picks it up
| > automatically,
| 
| This notation works too.

For humans looking at ChangeLog, it makes it easier to mentally
caterogize the bug (i.e. middle-end, C, optimization, target) without
having to go through bugzilla.

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13856