[RFA] read_reg() patch

[RFA] read_reg() patch

[Martin M. Hunt]

[-- Attachment #1: Type: text/plain, Size: 372 bytes --]

2004-07-20  Kevin Buettner and Martin Hunt  <hunt@redhat.com>
  
        * dwarf2-frame.c (read_reg): Add a call to
store_unsigned_integer.
	This extracts the least significant register_size() bits and
	extends it to the size of a pointer. Without this, big-endian
	targets where pointer size != register size breaks.
   
-- 
Martin M. Hunt <hunt@redhat.com>
Red Hat Inc.

[-- Attachment #2: read_reg.patch --]
[-- Type: text/x-patch, Size: 641 bytes --]

Index: dwarf2-frame.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2-frame.c,v
retrieving revision 1.36
diff -w -u -r1.36 dwarf2-frame.c
--- dwarf2-frame.c	15 Jun 2004 01:04:19 -0000	1.36
+++ dwarf2-frame.c	20 Jul 2004 18:54:54 -0000
@@ -214,6 +214,8 @@
 
   buf = (char *) alloca (register_size (gdbarch, regnum));
   frame_unwind_register (next_frame, regnum, buf);
+  store_unsigned_integer (buf, TYPE_LENGTH (builtin_type_void_data_ptr),
+			  extract_unsigned_integer (buf, register_size (gdbarch, regnum)));
   return extract_typed_address (buf, builtin_type_void_data_ptr);
 }
 

Re: [RFA] read_reg() patch

[Mark Kettenis]

   From: "Martin M. Hunt" <hunt@redhat.com>
   Date: Tue, 20 Jul 2004 12:16:51 -0700

   2004-07-20  Kevin Buettner and Martin Hunt  <hunt@redhat.com>

	   * dwarf2-frame.c (read_reg): Add a call to
   store_unsigned_integer.
	   This extracts the least significant register_size() bits and
	   extends it to the size of a pointer. Without this, big-endian
	   targets where pointer size != register size breaks.

The formatting of this ChangeLog entry isn't right.  Please look at
the existing ChangeLogs to see how you should write them.  Also move
the explanation of what the code does to the code.  The ChangeLog
should just say what's changed, not why.

I also think that the code itself is pretty unreadable.  There are too
many nested function calls.  I think it could be improved by using a
(temporary) local variable to store the result of
extract_unsigned_integer().

Mark

Re: [RFA] read_reg() patch

[Martin M. Hunt]

[-- Attachment #1: Type: text/plain, Size: 357 bytes --]


> I also think that the code itself is pretty unreadable.  There are too
> many nested function calls.  I think it could be improved by using a
> (temporary) local variable to store the result of
> extract_unsigned_integer().

How's this?

2004-07-26  Martin Hunt  <hunt@redhat.com>
	* dwarf2-frame.c (read_reg): Add a call to
   	store_unsigned_integer.


[-- Attachment #2: read_reg.patch --]
[-- Type: text/x-patch, Size: 1039 bytes --]

Index: dwarf2-frame.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2-frame.c,v
retrieving revision 1.38
diff -u -p -r1.38 dwarf2-frame.c
--- dwarf2-frame.c	23 Jul 2004 22:05:20 -0000	1.38
+++ dwarf2-frame.c	26 Jul 2004 20:54:39 -0000
@@ -209,11 +209,20 @@ read_reg (void *baton, int reg)
   struct gdbarch *gdbarch = get_frame_arch (next_frame);
   int regnum;
   char *buf;
+  CORE_ADDR tmp;
 
   regnum = DWARF2_REG_TO_REGNUM (reg);
 
   buf = (char *) alloca (register_size (gdbarch, regnum));
   frame_unwind_register (next_frame, regnum, buf);
+
+  /* This extracts the least significant register_size() bits and
+     extends it to the size of a pointer. Without this, big-endian
+     targets where pointer size != register size breaks. */
+  tmp = extract_unsigned_integer (buf, register_size (gdbarch, regnum));
+  store_unsigned_integer (buf, TYPE_LENGTH (builtin_type_void_data_ptr), 
+			  tmp);
+		
   return extract_typed_address (buf, builtin_type_void_data_ptr);
 }
 

Re: [RFA] read_reg() patch

[Mark Kettenis]

   From: "Martin M. Hunt" <hunt@redhat.com>
   Date: Mon, 26 Jul 2004 13:57:34 -0700

   > I also think that the code itself is pretty unreadable.  There are too
   > many nested function calls.  I think it could be improved by using a
   > (temporary) local variable to store the result of
   > extract_unsigned_integer().

   How's this?

   2004-07-26  Martin Hunt  <hunt@redhat.com>
	   * dwarf2-frame.c (read_reg): Add a call to
	   store_unsigned_integer.

Great, although I'd chose a somewhat less generic name for the
variable than `tmp'; although the `val' I've got in mind is pretty
generic too :-(.  However, the type shouldn't be CORE_ADDR, but
ULONGEST.  Consider a patch with that change pre-approved.

Mark