From mboxrd@z Thu Jan  1 00:00:00 1970
From: Mark Kettenis <kettenis@wins.uva.nl>
To: gdb-patches@sources.redhat.com
Cc: grossman@juniper.dot.net
Subject: [PATCH] Fix returning floating points values for x86
Date: Wed, 11 Jul 2001 01:49:00 -0000
Message-id: <200107110849.f6B8nGB19158@delius.kettenis.local>
X-SW-Source: 2001-07/msg00250.html

Fixes two FAILS in gdb.base/return2.exp.

Stu, I was about to commit this patch when I saw you FreeBSD patch.
Patch is almost identical (modulo comments), so it must be right :-).
I nevertheless took all the credit ;-).  I will come back to you on
some of the other stuff in your patch.

Checked in.

Mark


Index: ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>

	* i386-tdep.c (i386_extract_return_value): "Fix" comment.
	(i386_store_return_value): Frob FPU status and tag word to make
	sure the return value is the only value on the FPU stack.

Index: i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.31
diff -u -p -r1.31 i386-tdep.c
--- i386-tdep.c 2001/05/09 16:16:33 1.31
+++ i386-tdep.c 2001/07/11 08:38:27
@@ -803,7 +803,8 @@ i386_extract_return_value (struct type *
 	  return;
 	}
 
-      /* Floating-point return values can be found in %st(0).  */
+      /* Floating-point return values can be found in %st(0).
+         FIXME: Does %st(0) always correspond to FP0?  */
       if (len == TARGET_LONG_DOUBLE_BIT / TARGET_CHAR_BIT
 	  && TARGET_LONG_DOUBLE_FORMAT == &floatformat_i387_ext)
 	{
@@ -861,6 +862,8 @@ i386_store_return_value (struct type *ty
 
   if (TYPE_CODE (type) == TYPE_CODE_FLT)
     {
+      unsigned int fstat;
+
       if (NUM_FREGS == 0)
 	{
 	  warning ("Cannot set floating-point return value.");
@@ -889,6 +892,16 @@ i386_store_return_value (struct type *ty
 	  write_register_bytes (REGISTER_BYTE (FP0_REGNUM), buf,
 				FPU_REG_RAW_SIZE);
 	}
+
+      /* Set the top of the floating point register stack to 7.  That
+         makes sure that FP0 (which we set above) is indeed %st(0).
+         FIXME: Perhaps we should completely reset the status word?  */
+      fstat = read_register (FSTAT_REGNUM);
+      fstat |= (7 << 11);
+      write_register (FSTAT_REGNUM, fstat);
+
+      /* Mark %st(1) through %st(7) as empty.  */
+      write_register (FTAG_REGNUM, 0x3fff);
     }
   else
     {