This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [rfc] dwarf2 unwinder and MIPS n32
- From: "Maciej W. Rozycki" <macro at mips dot com>
- To: Daniel Jacobowitz <drow at false dot org>
- Cc: gdb-patches at sourceware dot org, "Maciej W. Rozycki" <macro at linux-mips dot org>
- Date: Mon, 30 Apr 2007 13:40:54 +0100 (BST)
- Subject: Re: [rfc] dwarf2 unwinder and MIPS n32
- References: <20070428202914.GA8077@caradoc.them.org>
On Sat, 28 Apr 2007, Daniel Jacobowitz wrote:
> Maybe if the size of the register != the size of a void * we should
> store it as an unsigned integer. But that seems hackish to me.
Of course MIPS addresses are signed, so if you have a stack pointer in
KSEG0 (e.g. 0x8fff0000), you want to store it as a signed integer, don't
you?
Anyway, I have the following patch waiting in the queue for submission --
perhaps it is less hackish. ;-)
2007-04-30 Maciej W. Rozycki <macro@mips.com>
* dwarf2-frame.c (read_reg): Extract an address using the width
of the register to be used to hold it as the size.
(dwarf2_frame_prev_register): Store an address using the width
of the holding register as the size.
It can certainly jump the queue if you find it useful. ;-)
Maciej
13118.diff
Index: gdb/src/gdb/dwarf2-frame.c
===================================================================
--- gdb.orig/src/gdb/dwarf2-frame.c 2007-02-13 13:51:55.000000000 +0000
+++ gdb/src/gdb/dwarf2-frame.c 2007-02-13 13:51:56.000000000 +0000
@@ -229,6 +229,7 @@
static CORE_ADDR
read_reg (void *baton, int reg)
{
+ struct type reg_void_data_ptr = *builtin_type_void_data_ptr;
struct frame_info *next_frame = (struct frame_info *) baton;
struct gdbarch *gdbarch = get_frame_arch (next_frame);
int regnum;
@@ -236,6 +237,8 @@
regnum = DWARF2_REG_TO_REGNUM (reg);
+ TYPE_LENGTH(®_void_data_ptr) = register_size (gdbarch, regnum);
+
buf = alloca (register_size (gdbarch, regnum));
frame_unwind_register (next_frame, regnum, buf);
@@ -244,7 +247,7 @@
under the covers, and this makes more sense for non-pointer
registers. Maybe read_reg and the associated interfaces should
deal with "struct value" instead of CORE_ADDR. */
- return unpack_long (register_type (gdbarch, regnum), buf);
+ return extract_typed_address (buf, ®_void_data_ptr);
}
static void
@@ -1020,10 +1023,13 @@
enum lval_type *lvalp, CORE_ADDR *addrp,
int *realnump, gdb_byte *valuep)
{
+ struct type reg_void_data_ptr = *builtin_type_void_data_ptr;
struct gdbarch *gdbarch = get_frame_arch (next_frame);
struct dwarf2_frame_cache *cache =
dwarf2_frame_cache (next_frame, this_cache);
+ TYPE_LENGTH(®_void_data_ptr) = register_size (gdbarch, regnum);
+
switch (cache->reg[regnum].how)
{
case DWARF2_FRAME_REG_UNDEFINED:
@@ -1132,7 +1138,7 @@
if (valuep)
{
/* Store the value. */
- store_typed_address (valuep, builtin_type_void_data_ptr, cache->cfa);
+ store_typed_address (valuep, ®_void_data_ptr, cache->cfa);
}
break;