gneric package breaks the restriction of No_Elaboration_Code ?

gneric package breaks the restriction of No_Elaboration_Code ?

[ytomino]

Hello.
I found unnecessary elaboration code is generated
that is probably caused by generic instantiation.
Please look the sources in below:

---- t_g.ads ----
pragma Restrictions (No_Elaboration_Code);
pragma No_Run_Time;
generic
procedure t_g;
pragma Pure (t_g);
---- t_g.adb ----
pragma Restrictions (No_Elaboration_Code);
pragma No_Run_Time;
procedure t_g is
begin
   null;
end t_g;
---- t_i.ads ----
pragma Restrictions (No_Elaboration_Code);
pragma No_Run_Time;
with t_g;
package t_i is
   pragma Pure (t_i);
   procedure nested is new t_g; -- *1
end t_i;
---- t_m.adb ----
pragma Restrictions (No_Elaboration_Code);
pragma No_Run_Time;
with t_g;
package t_i is
   pragma Pure (t_i);
   procedure nested is new t_g;
end t_i;
-----------------

And comple these, look b~t_m.adb.

% gnatmake -g t_m
gcc -c -g t_m.adb
gcc -c -g t_i.ads
gcc -c -g t_g.adb
gnatbind -x t_m.ali
gnatlink t_m.ali -g

---- b~t_m.adb ----
...
   procedure adainit is
      E2 : Boolean; pragma Import (Ada, E2, "t_i_E");

   begin
      null;

      t_i'elab_spec; -- *2
      E2 := True;
   end adainit;
...
-------------------

Also, I used nm, found __elabs procedure.

% nm t_i.o
00000261 D _t_i_E
00000008 T _t_i___elabs
00000000 T _t_i__nested

Next, remove generic instantiation (*1), and re-compile these,
then, elaboration code disappeared from adainit in b~t_m.adb.

---- b~t_m.adb ----
...
   procedure adainit is

   begin
      null;

   end adainit;
...
-------------------

__elabs procedure disappeared too.

% nm t_i.o
00000025 D _t_i_E

I think this is gcc's bug. (I tried with gcc-4.5.1/4.5.2)
Or, possibly this is correct on the spec ?

Re: gneric package breaks the restriction of No_Elaboration_Code ?

[Simon Wright]

ytomino <aghia05@gmail.com> writes:

> I think this is gcc's bug. (I tried with gcc-4.5.1/4.5.2)

Same with 4.6.0 experimental at 20110225.

Re: gneric package breaks the restriction of No_Elaboration_Code ?

[ytomino]

On Feb 28, 9:42 pm, Simon Wright <si...@pushface.org> wrote:
> ytomino <aghi...@gmail.com> writes:
> > I think this is gcc's bug. (I tried with gcc-4.5.1/4.5.2)
>
> Same with 4.6.0 experimental at 20110225.

Thanks for confirmation.

I reported this to GCC Bugzilla.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48013

and I made dirty-patch. It tentatively seems correct.
I don't know the details of gcc, cannot judge whether this solution is
correct or ill...
(Most of adainit was decreased with this patch, therefore disused .o
were removed on linking. I'm happy a little... Is it good or bad???)

Index: gcc/ada/gcc-interface/trans.c
===================================================================
--- gcc/ada/gcc-interface/trans.c	(revision 170538)
+++ gcc/ada/gcc-interface/trans.c	(working copy)
@@ -223,6 +223,34 @@
    of configurations.  */
 static const char *extract_encoding (const char *) ATTRIBUTE_UNUSED;
 static const char *decode_name (const char *) ATTRIBUTE_UNUSED;
+
+static bool is_empty (tree x);
+static bool is_empty (tree x)
+{
+  if (x){
+    switch (TREE_CODE (x)){
+    case STATEMENT_LIST:
+      do {
+        struct tree_statement_list_node *i = STATEMENT_LIST_HEAD(x);
+        while (i){
+          if (! is_empty (i->stmt)){
+            return false;
+          }
+          i = i->next;
+        }
+      }while (0);
+      break;
+    case STMT_STMT:
+      if (! is_empty (STMT_STMT_STMT (x))){ /* elm->exp.operands[0]
*/
+        return false;
+      }
+      break;
+    default:
+      return false;
+    }
+  }
+  return true;
+}

 /* This is the main program of the back-end.  It sets up all the
table
    structures and then generates code.  */
@@ -644,7 +672,7 @@
       gnu_stmts = gnu_body;
       if (TREE_CODE (gnu_stmts) == BIND_EXPR)
 	gnu_stmts = BIND_EXPR_BODY (gnu_stmts);
-      if (!gnu_stmts || !STATEMENT_LIST_HEAD (gnu_stmts))
+      if (is_empty (gnu_stmts))
 	Set_Has_No_Elaboration_Code (info->gnat_node, 1);
       else
 	{

Re: gneric package breaks the restriction of No_Elaboration_Code ?

[ytomino]

Sorry, I missed copy & paste t_m.adb.

pragma Restrictions (No_Elaboration_Code);
pragma No_Run_Time;
with t_i;
procedure t_m is
begin
   null;
end t_m;