Date: Fri, 16 Apr 1999 00:48:13 +0900 From: Hidetoshi Shimokawa <simokawa@sat.t.u-tokyo.ac.jp> To: dfr@nlsystems.com Cc: gallatin@cs.duke.edu, freebsd-alpha@freebsd.org, obrien@NUXI.com Subject: Re: EGCS and Alpha builds Message-ID: <14102.2621.961081.71582J@ett.sat.t.u-tokyo.ac.jp> In-Reply-To: In your message of "Tue, 13 Apr 1999 20:18:09 %2B0100 (BST)" <Pine.BSF.4.05.9904132016520.51054-100000@herring.nlsystems.com> References: <14099.10351.647510.43847V@ett.sat.t.u-tokyo.ac.jp> <Pine.BSF.4.05.9904132016520.51054-100000@herring.nlsystems.com>
next in thread | previous in thread | raw e-mail | index | archive | help
At Tue, 13 Apr 1999 20:18:09 +0100 (BST),
Doug Rabson <dfr@nlsystems.com> wrote:
> The backtrace code in DDB is somewhat fragile. It has to parse the
> function prologue to figure out where the return address is. I expect egcs
> has changed the prologue. GDB does something similar; hopefully it will
> be fixed in gdb-4.18.
Thanks for the hint. The prologue has changed. Egcs uses 'subq/addq' to
manipulate stack pointer instead of 'lda'. I made an adhoc fix for
DDB/GDB. Don't you mind my commiting this DDB fix to the tree?
(warning supression included)
Index: db_trace.c
===================================================================
RCS file: /pub/FreeBSD-CVS/src/sys/alpha/alpha/db_trace.c,v
retrieving revision 1.2
diff -u -r1.2 db_trace.c
--- db_trace.c 1998/06/27 15:37:42 1.2
+++ db_trace.c 1999/04/15 06:05:07
@@ -25,12 +25,9 @@
static void
parse_proc(db_expr_t addr, struct alpha_proc* frame)
{
- db_sym_t sym;
+ c_db_sym_t sym;
db_expr_t func;
db_expr_t junk, pc, limit;
- int frame_size;
- int reg_mask;
- int got_frame;
frame->pcreg = -1;
frame->reg_mask = 0;
@@ -50,7 +47,13 @@
ins.bits = *(u_int32_t*) pc;
if (ins.memory_format.opcode == op_lda
&& ins.memory_format.ra == 30) {
+ /* gcc 2.7 */
frame->frame_size += -ins.memory_format.offset;
+ } else if (ins.operate_lit_format.opcode == op_arit
+ && ins.operate_lit_format.function == op_subq
+ && ins.operate_lit_format.rs == 30) {
+ /* egcs */
+ frame->frame_size += ins.operate_lit_format.literal;
} else if (ins.memory_format.opcode == op_stq
&& ins.memory_format.rb == 30
&& ins.memory_format.ra != 31) {
@@ -66,7 +69,6 @@
void
db_stack_trace_cmd(db_expr_t addr, boolean_t have_addr, db_expr_t count, char *modif)
{
- struct alpha_proc proc;
db_addr_t callpc;
db_addr_t frame;
@@ -82,10 +84,9 @@
}
while (count--) {
- u_int64_t *actframe;
- char * name;
+ const char * name;
db_expr_t offset;
- db_sym_t sym;
+ c_db_sym_t sym;
struct alpha_proc proc;
sym = db_search_symbol(callpc, DB_STGY_ANY, &offset);
Here is a gdb fix.
--- alpha-tdep.c.orig Thu Apr 8 21:24:45 1999
+++ alpha-tdep.c Fri Apr 16 00:44:29 1999
@@ -502,6 +502,10 @@
epilogue is reached. */
break;
}
+ else if ((word & 0xfff00fff) == 0x43c0053e) /* subq $sp,n,$sp */
+ {
+ frame_size += word & 0xff;
+ }
else if ((word & 0xfc1f0000) == 0xb41e0000 /* stq reg,n($sp) */
&& (word & 0xffff0000) != 0xb7fe0000) /* reg != $zero */
{
--
/\ Hidetoshi Shimokawa
\/ simokawa@sat.t.u-tokyo.ac.jp
PGP public key: finger -l simokawa@sat.t.u-tokyo.ac.jp
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?14102.2621.961081.71582J>