From: Jeffrey A. Law <law@snake.cs.utah.edu>
Date: Tue, 2 May 1995 05:27:01 +0000 (-0600)
Subject: pa.c (hppa_expand_epilogue): Correctly handle restore of %rp for functions with a... 
X-Git-Tag: misc/cutover-egcs-0~4397
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=473fa8ccce8bb50c9ff8c591520cd8aadc7691d1;p=thirdparty%2Fgcc.git

pa.c (hppa_expand_epilogue): Correctly handle restore of %rp for functions with a stack size of exactly...

Tue May  2 01:25:29 1995  Jeffrey A. Law  <law@snake.cs.utah.edu>

	* pa.c (hppa_expand_epilogue): Correctly handle restore of %rp
	for functions with a stack size of exactly 8kbytes and no frame
	pointer.

From-SVN: r9561
---

diff --git a/gcc/config/pa/pa.c b/gcc/config/pa/pa.c
index ab43981c203e..f6a6960ca472 100644
--- a/gcc/config/pa/pa.c
+++ b/gcc/config/pa/pa.c
@@ -2439,8 +2439,15 @@ hppa_expand_epilogue ()
       set_reg_plus_d (STACK_POINTER_REGNUM,
 		      STACK_POINTER_REGNUM,
 		      - actual_fsize);
-      /* Uses value left over in %r1 by set_reg_plus_d.  */
-      load_reg (2, - (actual_fsize + 20 + ((- actual_fsize) & ~0x7ff)), 1);
+
+      /* This used to try and be clever by not depending on the value in
+	 %r30 and instead use the value held in %r1 (so that the 2nd insn
+	 which sets %r30 could be put in the delay slot of the return insn).
+	
+	 That won't work since if the stack is exactly 8k set_reg_plus_d
+	 doesn't set %r1, just %r30.  */
+      load_reg (2, - (actual_fsize + 20 + ((- actual_fsize) & ~0x7ff)),
+		STACK_POINTER_REGNUM);
     }
 
   /* Reset stack pointer (and possibly frame pointer).  The stack */