Fix float/double return on x86-64

Fix float/double return on x86-64

[Andreas Jaeger]

The appended patch fixes these three failures on x86-64:
FAIL: gdb.base/return.exp: correct value returned double test (known problem with sparc solaris)
FAIL: gdb.base/return2.exp: float value returned successfully
FAIL: gdb.base/return2.exp: double value returned successfully

The code does not handle complex floating point values, I'm not sure
that it's correct there but I decided to fix this issue first.

Ok to commit for both branches?

Andreas

2003-06-25  Andreas Jaeger  <aj@suse.de>

	* x86-64-tdep.c (x86_64_store_return_value): Return double and
	float values in SSE register.


============================================================
Index: src/gdb/x86-64-tdep.c
--- src/gdb/x86-64-tdep.c	15 Jun 2003 11:01:46 -0000	1.79
+++ src/gdb/x86-64-tdep.c	25 Jun 2003 19:10:46 -0000
@@ -712,8 +727,9 @@ x86_64_store_return_value (struct type *
 {
   int len = TYPE_LENGTH (type);
 
-  if (TYPE_CODE_FLT == TYPE_CODE (type))
-    {
+  /* First handle long doubles.  */
+  if (TYPE_CODE_FLT == TYPE_CODE (type) && len == 16)
+    { 
       ULONGEST fstat;
       char buf[FPU_REG_RAW_SIZE];
 
@@ -741,6 +757,12 @@ x86_64_store_return_value (struct type *
          for the tag word is 0x3fff.  */
       regcache_raw_write_unsigned (regcache, FTAG_REGNUM, 0x3fff);
     }
+  else if (TYPE_CODE_FLT == TYPE_CODE (type))
+    {
+      /* Handle double and float variables.  */
+      regcache_cooked_write (regcache,  X86_64_XMM0_REGNUM, valbuf);
+    }
+  /* XXX: What about complex floating point types?  */
   else
     {
       int low_size = REGISTER_RAW_SIZE (0);

-- 
 Andreas Jaeger, SuSE Linux AG, aj@suse.de, http://www.suse.de/~aj
  GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

Re: Fix float/double return on x86-64

[Mark Kettenis]

   Date: Fri, 11 Jul 2003 12:30:37 +0200
   From: Michal Ludvig <mludvig@suse.cz>

   Andreas Jaeger told me that:
   > 
   > The appended patch fixes these three failures on x86-64:
   > FAIL: gdb.base/return.exp: correct value returned double test (known problem with sparc solaris)
   > FAIL: gdb.base/return2.exp: float value returned successfully
   > FAIL: gdb.base/return2.exp: double value returned successfully
   > 
   > The code does not handle complex floating point values, I'm not sure
   > that it's correct there but I decided to fix this issue first.
   > 
   > Ok to commit for both branches?

   Mark, could you comment on this one, please? Could it go in?

Sorry, yes, I thought I already approved that one.  Turns out it
slipped through the cracks after all.  Please go ahead and check it in
on both branches.

Mark