gdb/gdb-ia64-infcall-workaround.patch
Jan Kratochvil 5a6df07a01 - Fix gdb.ada/* regressions (Keith Seitz).
- Disable addon (finish) due to inline-cmds.exp: up from outer_inline2
    assert.
- Fix i386+x86_64 rwatch+awatch before run, regression against 6.8 (BZ
    541866).
- Fix crash on stale addrinfo->sectindex (more sensitive due to the PIE
    patch).
- Remove false gdb_assert on $sp underflow.
- Workaround ia64 inferior calls clearing SP.
- testsuite: Fix long timeout on arches with failing
    gdb.base/interrupt.exp.
- Fix ia64 part of the bt-clone-stop.exp fix.
2010-02-26 01:18:33 +00:00

118 lines
3.4 KiB
Diff

There was a regression by:
Re: [rfc, v3] Fix frame_id_inner comparison false positives
http://sourceware.org/ml/gdb-patches/2008-08/msg00578.html
http://sourceware.org/ml/gdb-cvs/2008-08/msg00182.html
916dde5d38b45a659514e47942ece70aec04cd78
the part:
* stack.c (return_command): Directly pop the selected frame.
Reproducible on:
hp-diablo-01.rhts.eng.bos.redhat.com
RHEL5.5-Server-20100216.nightly
kernel-2.6.18-187.el5.ia64
Not reproducible on:
hp-bl860c-03.rhts.eng.bos.redhat.com
RHEL-4.8
kernel-2.6.9-89.EL.ia64
To make ia64-tdep.c compilable +/- this stack.c check-in one must apply:
Re: [RFC] Remove addr, endaddr, offset from obj_section
http://sourceware.org/ml/gdb-patches/2008-09/msg00011.html
http://sourceware.org/ml/gdb-cvs/2008-09/msg00009.html
99072369ec3c8f94c9a596e5ce30bf1f1c4bf20e
It is reproducible by:
gdb.base/call-signal-resume.exp
although one must first remove the gdb_assert by:
gdb-infcall-sp-underflow.patch
Reproducer:
set confirm no
set breakpoint pending on
set height 0
set width 0
# testcase from: gdb-6.8-37.el5.src
file gdb.base/call-signals
break stop_one
run
call gen_signal ()
bt
frame 3
return
p/x $sp
break stop_two
info break
p/x $pc
p/x $sp
bt
set debug infrun 1
delete 1
continue
bt
p/x $pc
echo Bug is reproduced if sp is now 0:\n
p/x $sp
kill
quit
--- ./gdb/frame.c 2010-02-23 21:35:35.000000000 +0100
+++ ./gdb/frame.c 2010-02-23 21:46:34.000000000 +0100
@@ -517,7 +517,7 @@ frame_id_eq (struct frame_id l, struct f
a stack overflow strategy that cause the handler to be run on a
different stack. */
-static int
+int
frame_id_inner (struct gdbarch *gdbarch, struct frame_id l, struct frame_id r)
{
int inner;
--- ./gdb/frame.h 2009-09-13 18:28:28.000000000 +0200
+++ ./gdb/frame.h 2010-02-23 21:46:10.000000000 +0100
@@ -708,4 +708,7 @@ extern struct frame_info *create_new_fra
extern int frame_unwinder_is (struct frame_info *fi,
const struct frame_unwind *unwinder);
+extern int frame_id_inner (struct gdbarch *gdbarch, struct frame_id l,
+ struct frame_id r);
+
#endif /* !defined (FRAME_H) */
--- ./gdb/stack.c 2010-02-23 21:35:34.000000000 +0100
+++ ./gdb/stack.c 2010-02-23 21:44:39.000000000 +0100
@@ -1967,8 +1967,29 @@ If you continue, the return value that y
error (_("Not confirmed"));
}
- /* Discard the selected frame and all frames inner-to it. */
- frame_pop (get_selected_frame (NULL));
+ /* NOTE: cagney/2003-01-18: Is this silly? Rather than pop each
+ frame in turn, should this code just go straight to the relevant
+ frame and pop that? */
+
+ /* First discard all frames inner-to the selected frame (making the
+ selected frame current). */
+ {
+ struct frame_id selected_id = get_frame_id (get_selected_frame (NULL));
+ while (!frame_id_eq (selected_id, get_frame_id (get_current_frame ())))
+ {
+ struct frame_info *frame = get_current_frame ();
+ if (frame_id_inner (get_frame_arch (frame), selected_id,
+ get_frame_id (frame)))
+ /* Caught in the safety net, oops! We've gone way past the
+ selected frame. */
+ error (_("Problem while popping stack frames (corrupt stack?)"));
+ frame_pop (get_current_frame ());
+ }
+ }
+
+ /* Second discard the selected frame (which is now also the current
+ frame). */
+ frame_pop (get_current_frame ());
/* Store RETURN_VALUE in the just-returned register set. */
if (return_value != NULL)