[multiple changes]

2004-04-23  Emmanuel Briot  <briot@act-europe.fr>

	* adaint.c (__gnat_try_lock): No longer requires that the parent
	directory be writable, the directory itself is enough.
	(gnat_is_absolute_path): Change profile, so that the call from
	GNAT.OS_Lib can be made more efficient.

	* adaint.h (gnat_is_absolute_path): Change profile, so that the call
	from GNAT.OS_Lib can be made more efficient.

	* g-os_lib.adb (Is_Absolute_Path): More efficient implementation, avoid
	one copy of the file name. Found by code reading.

2004-04-23  Vincent Celier  <celier@gnat.com>

	* gnat_ugn.texi: Add documentation for gnatmake switch -eL
	Correct documentation on gnatmake switches transmitted to the compiler

	* ali.ads: Minor comment fix

2004-04-23  Javier Miranda  <miranda@gnat.com>

	* sem_ch6.adb: (Confirming Types): Code cleanup

	* decl.c (gnat_to_gnu_entity): Give support to anonymous access to
	subprogram types: E_Anonymous_Access_Subprogram_Type and
	E_Anonymous_Access_Protected_Subprogram_Type.

2004-04-23  Thomas Quinot  <quinot@act-europe.fr>

	* sem_dist.adb: Add a new paramter to the RAS_Access TSS indicating
	whether a pragma All_Calls_Remote applies to the subprogram on which
	'Access is taken.
	No functional change is introduced by this revision; the new parameter
	will be used to allow calls to local RCI subprograms to be optimized
	to not use the PCS in the case where no pragma All_Calls_Remote applies,
	as is already done in the PolyORB implementation of the DSA.

	* exp_dist.adb: Add a new paramter to the RAS_Access TSS indicating
	whether a pragma All_Calls_Remote applies to the subprogram on which
	'Access is taken.
	No functional change is introduced by this revision; the new parameter
	will be used to allow calls to local RCI subprograms to be optimized
	to not use the PCS in the case where no pragma All_Calls_Remote applies,
	as is already done in the PolyORB implementation of the DSA.

2004-04-23  Robert Dewar  <dewar@gnat.com>

	* Makefile.rtl: Add entry for s-addope.o in run time library list
	* Make-lang.in: Add entry for s-addope.o to GNAT1 objects
	* s-addope.ads, s-addope.adb: New files.

	* s-carsi8.adb, s-carun8.adb, s-casi16.adb, s-casi32.adb,
	s-casi64.adb, s-caun16.adb, s-caun32.adb, s-caun64.adb,
	s-finimp.adb, s-geveop.adb, s-stoele.adb: Modifications to allow
	System.Address to be non-private and signed.

	* sem_elim.adb: Minor reformatting (fairly extensive)
	Some minor code reorganization from code reading
	Add a couple of ??? comments

2004-04-23  Richard Kenner  <kenner@vlsi1.ultra.nyu.edu>

	* trans.c (tree_transform, build_unit_elab): Don't call getdecls.
        (tree_transform, case N_If_Statement): Remove non-determinism.

	* utils.c (begin_subprog_body): Just set DECL_CONTEXT in PARM_DECL.

2004-04-23  Sergey Rybin  <rybin@act-europe.fr>

	* gnat_rm.texi: Small fixes in the changes made in the 'pragma
	Eliminate' section.

	* snames.ads, snames.adb: Remove Name_Homonym_Number (Homonym_Number is
	no longer used as a parameter name for Eliminate pragma).

From-SVN: r81086

1 files changed, 5 insertions, 25 deletions

diff --git a/gcc/ada/utils.c b/gcc/ada/utils.c
index 3e3d6b5..8c25f48 100644
--- a/gcc/ada/utils.c
+++ b/gcc/ada/utils.c
@@ -1841,9 +1841,7 @@ static int function_nesting_depth;
 void
 begin_subprog_body (tree subprog_decl)
 {
-  tree param_decl_list;
   tree param_decl;
-  tree next_param;
 
   if (function_nesting_depth++ != 0)
     push_function_context ();
@@ -1859,32 +1857,14 @@ begin_subprog_body (tree subprog_decl)
      the C sense!  */
   TREE_STATIC (subprog_decl)   = 1;
 
-  /* Enter a new binding level.  */
+  /* Enter a new binding level and show that all the parameters belong to
+     this function.  */
   current_function_decl = subprog_decl;
   pushlevel (0);
 
-  /* Push all the PARM_DECL nodes onto the current scope (i.e. the scope of the
-     subprogram body) so that they can be recognized as local variables in the
-     subprogram.
-
-     The list of PARM_DECL nodes is stored in the right order in
-     DECL_ARGUMENTS.  Since ..._DECL nodes get stored in the reverse order in
-     which they are transmitted to `pushdecl' we need to reverse the list of
-     PARM_DECLs if we want it to be stored in the right order. The reason why
-     we want to make sure the PARM_DECLs are stored in the correct order is
-     that this list will be retrieved in a few lines with a call to `getdecl'
-     to store it back into the DECL_ARGUMENTS field.  */
-    param_decl_list = nreverse (DECL_ARGUMENTS (subprog_decl));
-
-    for (param_decl = param_decl_list; param_decl; param_decl = next_param)
-      {
-	next_param = TREE_CHAIN (param_decl);
-	TREE_CHAIN (param_decl) = NULL;
-	pushdecl (param_decl);
-      }
-
-  /* Store back the PARM_DECL nodes. They appear in the right order. */
-  DECL_ARGUMENTS (subprog_decl) = getdecls ();
+  for (param_decl = DECL_ARGUMENTS (subprog_decl); param_decl;
+       param_decl = TREE_CHAIN (param_decl))
+    DECL_CONTEXT (param_decl) = subprog_decl;
 
   init_function_start (subprog_decl);
   expand_function_start (subprog_decl, 0);