This is the mail archive of the 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]

PATCH to simplify make_decl_rtl

This patch resulted from trying to figure out how the target
callback encode_section_info should work - is it ok to just
"smash" the old SYMBOL_REF's XSTR in place, as may targets do?
The answer is clearly yes for the two places that the callback
is called on a fresh clearly-unshared DECL_RTL.  It is less
clear if it is safe or desirable on the remaining site where
it is called on an existing DECL_RTL.  But I though about the
changes that encode_section_info might do to a DECL_RTL, and
I doubted that all the callsbacks could robustly handle being
called twice on the same decl.  So I thought that it would
probably be simpler and more robust to get rid of the special
handling for DECL_RTL_SET_P in make_decl_rtl.

It seems to work.   Bootstrapped on Fedora 2.  It actually fixes
one C++ failure (builtins9.C), and 5 Java failures, though it
appears to add a new Java failure.  I think that's a fair trade.

I also appear to get a working linux-x-mingw compiler.

OK to check into mainline?
	--Per Bothner
2004-06-04  Per Bothner  <>

	* varasm.c (make_decl_rtl):  Remove special handling if DECL_RTL_SET_P.

Index: varasm.c
RCS file: /cvs/gcc/gcc/gcc/varasm.c,v
retrieving revision 1.428
diff -u -p -r1.428 varasm.c
--- varasm.c	30 May 2004 18:32:31 -0000	1.428
+++ varasm.c	4 Jun 2004 15:48:12 -0000
@@ -761,32 +761,6 @@ make_decl_rtl (tree decl, const char *as
 	   || TREE_CODE (decl) == LABEL_DECL)
     abort ();
-  /* For a duplicate declaration, we can be called twice on the
-     same DECL node.  Don't discard the RTL already made.  */
-  if (DECL_RTL_SET_P (decl))
-    {
-      /* If the old RTL had the wrong mode, fix the mode.  */
-      if (GET_MODE (DECL_RTL (decl)) != DECL_MODE (decl))
-	SET_DECL_RTL (decl, adjust_address_nv (DECL_RTL (decl),
-					       DECL_MODE (decl), 0));
-      /* ??? Another way to do this would be to maintain a hashed
-	 table of such critters.  Instead of adding stuff to a DECL
-	 to give certain attributes to it, we could use an external
-	 hash map from DECL to set of attributes.  */
-      /* Let the target reassign the RTL if it wants.
-	 This is necessary, for example, when one machine specific
-	 decl attribute overrides another.  */
-      targetm.encode_section_info (decl, DECL_RTL (decl), false);
-      /* Make this function static known to the mudflap runtime.  */
-      if (flag_mudflap && TREE_CODE (decl) == VAR_DECL)
-	mudflap_enqueue_decl (decl);
-      return;
-    }
   reg_number = decode_reg_name (asmspec);
   if (reg_number == -2)

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