This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Add bp_location to Python interface
- From: Kevin Pouget <kevin dot pouget at gmail dot com>
- To: gdb-patches at sourceware dot org
- Cc: pmuldoon at redhat dot com
- Date: Mon, 9 Jan 2012 12:46:30 +0100
- Subject: Re: [PATCH] Add bp_location to Python interface
- References: <CAPftXUL7Gag8pOzh6NF2t=nowBgSYa_Je3CkMQHgxexrNz+OHg@mail.gmail.com> <m3liqn9lym.fsf@redhat.com> <CAPftXULeNBNLT=9EixD3sEdtxcLkRuUFY6ny=2k9NSUQqHf+Bg@mail.gmail.com> <m3k4656dkx.fsf@redhat.com> <CAPftXU+NjNead1ZHEc8_P84py3esmOnAKVa2Y1KpxKX_augKsw@mail.gmail.com>
On Tue, Dec 13, 2011 at 4:28 PM, Kevin Pouget <kevin.pouget@gmail.com> wrote:
>
> On Fri, Dec 9, 2011 at 3:10 PM, Phil Muldoon <pmuldoon@redhat.com> wrote:
> > Kevin Pouget <kevin.pouget@gmail.com> writes:
> >
> >> On Thu, Dec 8, 2011 at 3:28 PM, Phil Muldoon <pmuldoon@redhat.com> wrote:
> >>
> >
> >>>> +A breakpoint location is represented with a @code{gdb.BpLocation} object,
> >>>> +which offers the following attributes (all read only) and methods.
> >>>> +Please note that breakpoint locations are very transient entities in
> >>>> +@value{GDBN}, so one should avoid keeping references to them.
> >>
> >>>
> >>> While it is good to note that, I'm not sure ?what we are explaining here
> >>> other than when the breakpoint is deleted, all location objects that are
> >>> associated with that object are invalid too. ?Or, are you noting we
> >>> should allow the user to interact after the fact? If that is the case,
> >>> what is "is_valid" for? ?Why note the transient nature of the object?
> >>
> >>> I'm not sure what we are explaining here other than when the breakpoint is deleted
> >>
> >> that's what I wanted to emphasize here, bp_locations are cleaned/reset
> >> more often than 'high-level' breakpoints are removed. My knowledge
> >> about that is quite limited, but for instance this (cleaned up) stack:
> >>
> >> #0 ?gdbpy_bplocation_free py-bploc.c:60
> >> #1 ?free_bp_location (loc=) breakpoint.c:5685
> >> #2 ?decref_bp_location (blp=) breakpoint.c:5707
> >> #3 ?update_global_location_list (should_insert=1) ?breakpoint.c:10575
> >> #4 ?update_breakpoint_locations (b=, sals=..., ? ? sals_end=...)
> >> breakpoint.c:11787
> >> #5 ?breakpoint_re_set_default (b=) breakpoint.c:11937
> >> #6 ?breakpoint_re_set_one (bint=) breakpoint.c:11968
> >> #8 ?breakpoint_re_set () breakpoint.c:11992
> >> #9 ?solib_add (pattern=0x0, from_tty=0, target=, ?readsyms=1) solib.c:926
> >> #10 bpstat_what (bs_head=) breakpoint.c:4487
> >> #11 handle_inferior_event (ecs=) infrun.c:4394
> >> #12 wait_for_inferior ()
> >>
> >> shows that bplocations might be removed when a shared library is loaded
> >
> > Good point, I did not think of this scenario. ?Looking at several
> > sections of GDB code, update_global_location_list can just be called
> > whenever.
> >
> >>> If the breakpoint is deleted and the user still has reference to a
> >>> location object, I think we should just run a validation routine and
> >>> refuse to do anything but raise an exception at that point (like
> >>> is_valid, but triggered for all function/attribute calls).
> >>
> >> hum, I think that's what I already do in the code :)
> >>
> >> I've updated the gdb.BpLocation.is_valid documentation, maybe it's clearer?
> >
> > Yeah, I saw it later. ?Forgot to delete that comment!
> >
> >>> @defun BpLocation.is_valid ()
> >>> Returns @code{True} if the @code{gdb.BpLocation} object is valid,
> >>> @code{False} if not. ?A @code{gdb.BpLocation} object may be invalidated by
> >>> GDB at any moment for internal reasons. All other @code{gdb.BpLocation} methods
> >>> and attributes will throw an exception if the object is invalid.
> >>> @end defun
> >>
> >>>> +
> >>>> +@defvar BpLocation.owner
> >>>> +This attribute holds a reference to the @code{gdb.Breakpoint} object which
> >>>> +owns this location.
> >>>> +@end defvar
> >>
> >>> Note which attributes are read only/writable. ?This and others.
> >> I specify a few lines above that all the attributes are read only, but
> >> I've noted it on each attribute as well.
> >
> > Right, but there is no guarantee that the user will read the entirety
> > of that section's documentation. ?So we always put attribute access rights
> > in the attribute section.
>
> ok, I thought that as soon as it was clearly mentioned somewhere, that
> was enough. But it doesn't cost anything to me to write it for each
> attribute, so no problem!
>
> >>> Also many different breakpoints can be at one location. ?I'm not sure if it
> >>> is worth pointing out here that "this breakpoint" only owns the
> >>> location insofar as the scope of that breakpoint (there could be other
> >>> breakpoints set to that location).
> >>
> >> from an implementation point of view, the is a BP ?<1--------n>
> >> BpLocation relation, even if two locations have the same address.
> >>
> >> Also, I think that the end-users won't have problems to understand
> >> these BpLocations, as they're already exposed with "info breakpoints":
> >
> > I was purely talking from a scripting point of view, but ok. ?My view
> > is, if the user has to run experiments to collect data on the actual
> > behavior, then our documentation needs to be a little better.
>
> yes, no doubt about that!
>
> > You could just put in that "from an implementation point of view" paragraph in
> > the documentation somewhere?
>
> I've added it to 'BpLocation.owner' description:
>
> @defvar BpLocation.owner
> This attribute holds a reference to the @code{gdb.Breakpoint} object which
> owns this location. ?This attribute is not writable. ?From an implementation
> point of view, there is a @code{1 ... n} relation between a breakpoint and
> its locations, even if two breakpoints are set on at same address.
>
>
>
> >>>> +struct bploc_object
> >>>> +{
> >>>> + ?PyObject_HEAD
> >>>> +
> >>>> + ?/* The location corresponding to this py object. ?NULL is the location
> >>>> + ? ? has been deleted. ?*/
> >>>> + ?struct bp_location *loc;
> >>>
> >>> Typo in the comment "NULL is the location has been deleted.". ?Also nit
> >>> pick shouldn't it be "The location corresponding to a gdb.Breakpoint
> >>> object"?
> >>
> >> typo fixed, but not the nit: it's the same idea as breakpoint,
> >> if the 'backend' breakpoint is deleted, 'struct breakpoint_object . bp' is NULL,
> >> if the 'backend' location is deleted, 'struct bploc_object . loc' is
> >> NULL
> >
> > My objection was to "py object", I would just find it clearer to name
> > the object. ?Also, the explanation above would go a long way for future
> > hackers to understand the relationship. ?Maybe add that to the code too?
>
> You're right, it looks obvious to me, but it would take time to figure
> out by itself.
> Here are a few more bits of comments:
>
> /* The location corresponding to this gdb.BpLocation object. ?It's the same
> ? ? idea as gdb.Breakpoint, if the 'backend' location is deleted, LOC is
> ? ? set to NULL. ?No reference to the location is owned here (in terms of
> ? ? ref. counting) in order not to change the internal behavior. ?*/
> struct bp_location *loc;
>
> >>
> >>>> +
> >>>> + ?/* 1 if the owner BP has been deleted, 0 otherwise. ?*/
> >>>> + ?int invalid_owner;
> >>>> +
> >>>> + ?/* Cache for the gdb.Value object corresponding to loc->address. ?*/
> >>>> + ?PyObject *py_address;
> >>>> +};
> >>>
> >>> I'm not sure if breakpoint locations can change. ?I do not think so, but
> >>> why do we need to cache loc->address?
> >>
> >> I assumed that it can't change, but didn't actually investigate all
> >> the implementation.
> >> My feeling is "caching is easy, creating a Py object has a cost, so
> >> let's cache!", but I've got no idea about the actual cost, so I'll
> >> change it if someone insists [and convinces me] :)
> >
> > I don't know enough about breakpoint locations to be able to advise you,
> > but what does update_global_location_list do? ?Just shuffle add/delete
> > locations to the list? ?I am not against caching. ?Just not sure why we
> > need it. ?I tend towards the conservative with things like these, if
> > only purely because we keep missing reference counts in the existing
> > code.
>
> I somehow managed to convinced myself to remove this address caching,
> 1/I'm not sure about the actual py object creation, certainly quite low,
> 2/BpLocation objects have a short lifespan, so the cached addresses
> would be discarded quite frequently
> so all in all, I can't see anymore any relevant reason to cache it :)
>
> >>>> +
> >>>> +/* Require that LOCATION be a valid bp_location; throw a Python
> >>>> + ? exception if it is invalid. ?*/
> >>>> +#define BPLOCPY_REQUIRE_VALID(Location) ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? \
> >>>> + ? ?do { ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?\
> >>>> + ? ? ?if ((Location)->loc == NULL) ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?\
> >>>> + ? ? ? ?return PyErr_Format (PyExc_RuntimeError, ? ? ? ? ? ? ? ? ? ? ? ?\
> >>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? _("BpLocation invalid.")); ? ? ? ? ? ? ? ? \
> >>>> + ? ?} while (0)
> >>>
> >>> I prefer error messages a little more descriptive. ?That is just a
> >>> personal thing of mine, but "BpLocation invalid" would seem cryptic when
> >>> thrown in the context of a script.
> >>
> >> I'm not sure how I could improve it, at this point (checking validity
> >> at the beginning of all the attributes), we just know that the backend
> >> location has been freed, but we don't know why.
> >
> > "BpLocation object, breakpoint location is NULL". Or something like that.
>
> I'm still not sure, I think that when I read "BpLocation invalid", I'd
> take a look at the 'is_valid' method description, and understand
> that "A @code{gdb.BpLocation} object may be invalidated by GDB at any
> moment for internal reasons."
> 'breakpoint location is NULL' sounds like an internal error to me ...
>
> Maybe "BpLocation invalid: internal location object freed up/removed by GDB." ?
>
> >>>> +PyObject *
> >>>> +bplocation_to_bplocation_object (struct bp_location *loc)
> >>>> +{
> >>>> + ?bploc_object *bploc_obj;
> >>>> +
> >>>> + ?gdb_assert (loc != NULL);
> >>>
> >>> Is this a fatal error that we need to shutdown GDB for? ?gdb_assert
> >>> seems pretty strong from an API point of view.
> >>
> >> hum, I'm not sure here,
> >> I wrote this line to actually shutdown GDB instead of segfaulting on
> >> the next line,
> >> for instance 'py-pspace.c::pspace_to_pspace_object' and
> >> 'py-objfile.c::objfile_to_objfile_object" will just segfault if
> >> there're called with a NULL parameter.
> >>
> >> throwing an Python/GDB exception wouldn't make much sense to me
> >> either, as there end-user won't be able to deal with it
> >
> > Neither would killing GDB. ?I almost never use gdb_assert because of the
> > possible negative user implications. ?Can you discern that GDB has
> > entered an unstable state because a function has received a NULL
> > argument for a bp_location? ?If you cannot answer "yes" with a high
> > degree of certainty, it is my opinion we should raise an exception and
> > return NULL. ?We can make sure in the Python API that, if this is the
> > case, we will return the exception to the script boundary for the script
> > to deal with. ?From an API point-of-view I think we should defer to
> > script authors the opportunity to deal with and decide what to do in
> > this scenario. ?If GDB is truly unstable, then I agree, assert is
> > correct. ?I think the other two cases are bugs.
>
> I get your point and quite agree with you, the only thing is that
> we're talking about 'gdb_assert' and not 'assert' itself.
>
> > Neither would killing GDB.
> and here's my point, 'gdb_assert' does't kill GDB, but raises a
> warning message asking the user if he wants to quit or continue.
> My feeling about *assert is that it's a development/beta-test
> features, which might/should be turned into noop for production.
>
> > From an API point-of-view I think we should defer to script authors the opportunity to deal with
> I don't think so, because (from my developer point of view) I see it
> as very improbable that this situation would occur, or even impossible
> if I were a good programmer!
> I don't think that API should cover internal 'buggy' situations,
> otherwise they would be endless!
>
> > The Maintainers may differ on this, wait for their opinion.
> yes, I'm curious about different point of views
>
> >>>> +bplocpy_get_address (PyObject *self, void *closure)
> >>>> +{
> >>>> + ?bploc_object *self_bploc = (bploc_object *) self;
> >>>> +
> >>>> + ?BPLOCPY_REQUIRE_VALID (self_bploc);
> >>>> +
> >>>> + ?if (!self_bploc->py_address)
> >>>> + ? ?{
> >>>> + ? ? ?/* Get the address Value object as a void *. ?*/
> >>>> + ? ? ?volatile struct gdb_exception except;
> >>>> + ? ? ?struct type *void_ptr_type;
> >>>> + ? ? ?struct value *val = NULL ; /* Initialize to appease gcc warning. ?*/
> >>>> +
> >>>> + ? ? ?TRY_CATCH (except, RETURN_MASK_ALL)
> >>>> + ? ? ? ?{
> >>>> + ? ? ? ? ?void_ptr_type = lookup_pointer_type (
> >>>> + ? ? ? ? ? ? ?builtin_type (python_gdbarch)->builtin_void);
> >>>> + ? ? ? ? ?val = value_from_pointer (void_ptr_type, self_bploc->loc->address);
> >>>> + ? ? ? ?}
> >>>> + ? ? ?GDB_PY_HANDLE_EXCEPTION (except);
> >>>> +
> >>>> + ? ? ?self_bploc->py_address = value_to_value_object (val);
> >>>> + ? ? ?Py_XINCREF (self_bploc->py_address);
> >>>> + ? ?}
> >>>> + ?Py_XINCREF (self_bploc->py_address);
> >>>
> >>> I don't really mind it, but I prefer explicit return NULL when dealing
> >>> with cases of exceptions. ?I find the other logic hard to read. ?This is
> >>> not a request for a change. Is there a case where py_address will be
> >>> NULL? ?Yes, there is, value_to_value_object can return NULL. ?If it
> >>> returns NULL, then there is an exception set. ?I much prefer to exit
> >>> then and there, other the conditional XINCREF step, and returning at the
> >>> function exit. ?Still, this is just a stylistic thing, and probably
> >>> personal thing. ?The second XINCREF can just be a plain old INCREF as we
> >>> already tested for NULL.
> >>
> >> makes sense to me, my "single return point" idea doesn't stand against
> >> error handling
> >
> > Note this is purely a personal style issue, there was nothing wrong with
> > the correctness of your code.
>
> sure, I try to make it clear when I don't agree with the suggestion;
> this one was perfectly fine,
> I think that's actually the way I wrote it at the first time :)
>
> >>> This brings me to the address cache argument. ?Is it worthwhile managing
> >>> the cache increment counts instead of just returning the address each
> >>> time? ?I ask as I am not sure if you have done any metrics that indicate
> >>> this is a slow operation.
> >>
> >> no, I didn't run any measurement so far; I'll wait for a maintainer
> >> point of view about that
> >
> > Ok.
> >
> >>> We need to make sure that the this is not a watchpoint.
> >>
> >> why not?
> >> I don't know much about watchpoints, but as per my quick
> >> investigations, they share the same high-level structure (struct
> >> breakpoint), and they seem to use bp_location the same way ?
> >
> > I'm not sure as watchpoints are based on expressions. ?Maybe it is ok,
> > but note that watchpoints are removed and reinserted on scope of the
> > expression. ?It might be fine. ?A few tests would prove the issue.
>
> I'll try to build some test to ensure that it's fine
>
>
> thanks for your comments and reviews,
ping
patch refreshed against current HEAD
Thanks,
Kevin
--
2012-01-09 Kevin Pouget <kevin.pouget@st.com>
Add bp_location to Python interface
* Makefile.in (SUBDIR_PYTHON_OBS): Add py-bploc.o
(SUBDIR_PYTHON_SRCS): Add python/py-bploc.c
Add build rule for this file.
* breakpoint.h (struct bploc_object): Forward declaration.
(struct bp_location): Add py_bploc_obj.
* breakpoint.c (free_bp_location): Call gdbpy_bplocation_free.
* python/py-bploc.c: New file.
* python/py-breakpoint.c (bppy_locations): New function.
(breakpoint_object_methods): New method binding: locations().
* python/python-internal.h (bploc_object): New typedef.
(bplocation_to_bplocation_object): New prototype.
(gdbpy_initialize_bplocation): Likewise.
* python/python.c (gdbpy_bplocation_free): New empty stub.
(_initialize_python): Call gdbpy_initialize_bplocation.
* python/python.h (gdbpy_bplocation_free): New prototype.
doc/
Add bp_location to Python interface
* gdb.texinfo (Breakpoints In Python): Document
gdb.Breakpoint.locations and gdb.BpLocation.
testsuite/
Add bp_location to Python interface
* gdb.python/py-breakpoint.exp: Test gdb.BpLocation.
From e61ed33ace88dd336eea6f5e3b7f750243087822 Mon Sep 17 00:00:00 2001
From: Kevin Pouget <kevin.pouget@st.com>
Date: Wed, 18 May 2011 10:02:23 -0400
Subject: [PATCH] Add bp_location to Python interface
---
gdb/Makefile.in | 6 +
gdb/NEWS | 6 +
gdb/breakpoint.c | 3 +
gdb/breakpoint.h | 6 +
gdb/doc/gdb.texinfo | 44 ++++
gdb/python/py-bploc.c | 302 ++++++++++++++++++++++++++++
gdb/python/py-breakpoint.c | 39 ++++
gdb/python/python-internal.h | 6 +
gdb/python/python.c | 7 +
gdb/python/python.h | 2 +
gdb/testsuite/gdb.python/py-breakpoint.exp | 52 +++++
11 files changed, 473 insertions(+), 0 deletions(-)
create mode 100644 gdb/python/py-bploc.c
diff --git a/gdb/Makefile.in b/gdb/Makefile.in
index cd93768..e4cdd5a 100644
--- a/gdb/Makefile.in
+++ b/gdb/Makefile.in
@@ -273,6 +273,7 @@ SUBDIR_PYTHON_OBS = \
py-auto-load.o \
py-block.o \
py-bpevent.o \
+ py-bploc.o \
py-breakpoint.o \
py-cmd.o \
py-continueevent.o \
@@ -305,6 +306,7 @@ SUBDIR_PYTHON_SRCS = \
python/py-auto-load.c \
python/py-block.c \
python/py-bpevent.c \
+ python/py-bploc.c \
python/py-breakpoint.c \
python/py-cmd.c \
python/py-continueevent.c \
@@ -2033,6 +2035,10 @@ py-bpevent.o: $(srcdir)/python/py-bpevent.c
$(COMPILE) $(PYTHON_CFLAGS) $(srcdir)/python/py-bpevent.c
$(POSTCOMPILE)
+py-bploc.o: $(srcdir)/python/py-bploc.c
+ $(COMPILE) $(PYTHON_CFLAGS) $(srcdir)/python/py-bploc.c
+ $(POSTCOMPILE)
+
py-breakpoint.o: $(srcdir)/python/py-breakpoint.c
$(COMPILE) $(PYTHON_CFLAGS) $(srcdir)/python/py-breakpoint.c
$(POSTCOMPILE)
diff --git a/gdb/NEWS b/gdb/NEWS
index a9a7859..c529f96 100644
--- a/gdb/NEWS
+++ b/gdb/NEWS
@@ -9,6 +9,12 @@
* The binary "gdbtui" can no longer be built or installed.
Use "gdb -tui" instead.
+* Python scripting
+
+ ** A new method "gdb.Breakpoint.locations" has been added, as well as
+ the class gdb.BpLocation to provide further details about breakpoint
+ locations.
+
*** Changes in GDB 7.4
* GDB now handles ambiguous linespecs more consistently; the existing
diff --git a/gdb/breakpoint.c b/gdb/breakpoint.c
index 6bcedc4..623f2df 100644
--- a/gdb/breakpoint.c
+++ b/gdb/breakpoint.c
@@ -5679,6 +5679,9 @@ static void
free_bp_location (struct bp_location *loc)
{
loc->ops->dtor (loc);
+
+ gdbpy_bplocation_free (loc);
+
xfree (loc);
}
diff --git a/gdb/breakpoint.h b/gdb/breakpoint.h
index c1d3be9..8d6fa19 100644
--- a/gdb/breakpoint.h
+++ b/gdb/breakpoint.h
@@ -26,6 +26,7 @@
struct value;
struct block;
struct breakpoint_object;
+struct bploc_object;
struct get_number_or_range_state;
struct thread_info;
struct bpstats;
@@ -403,6 +404,11 @@ struct bp_location
/* Source file name of this address. */
char *source_file;
+
+ /* Python object associated with this location. May be NULL if the location
+ is not yet exported to Python. */
+
+ struct bploc_object *py_bploc_obj;
};
/* This structure is a collection of function pointers that, if available,
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index 2f4aa4f..f8375d6 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -24358,6 +24358,50 @@ commands, separated by newlines. If there are no commands, this
attribute is @code{None}. This attribute is not writable.
@end defvar
+@findex gdb.locations
+@defun gdb.locations ()
+Return a tuple containing a sequence of @code{gdb.BpLocation} objects
+(see below) associated with this breakpoint. A breakpoint with no location
+is a pending breakpoint (@xref{Set Breaks, , pending breakpoints}).
+@end defun
+
+A breakpoint location is represented with a @code{gdb.BpLocation} object,
+which offers the following attributes (all read only) and methods.
+Please note that breakpoint locations are very transient entities in
+@value{GDBN}, so one should avoid keeping references to them.
+
+@defvar BpLocation.owner
+This attribute holds a reference to the @code{gdb.Breakpoint} object which
+owns this location. This attribute is not writable. From an implementation
+point of view, there is a @code{1 ... n} relation between a breakpoint and
+its locations, even if two breakpoints are set on at same address.
+
+@end defvar
+
+@defvar BpLocation.enabled
+This attribute indicates whether this location is currently enabled or not.
+This attribute is not writable.
+@end defvar
+
+@defvar BpLocation.inferior
+This attribute holds a reference to the @code{gdb.Inferior} inferior object
+in which this breakpoint location has been inserted. The value will be
+@code{None} if there is no inferior associated with this location. This
+attribute is not writable.
+@end defvar
+
+@defvar BpLocation.address
+This attribute holds a @code{gdb.Value} object corresponding to the address
+at which the breakpoint has been inserted. This attribute is not writable.
+@end defvar
+
+@defun BpLocation.is_valid ()
+Returns @code{True} if the @code{gdb.BpLocation} object is valid,
+@code{False} if not. A @code{gdb.BpLocation} object may be invalidated by
+GDB at any moment for internal reasons. All other @code{gdb.BpLocation}
+methods and attributes will throw an exception if the object is invalid.
+@end defun
+
@node Finish Breakpoints in Python
@subsubsection Finish Breakpoints
diff --git a/gdb/python/py-bploc.c b/gdb/python/py-bploc.c
new file mode 100644
index 0000000..5ba3948
--- /dev/null
+++ b/gdb/python/py-bploc.c
@@ -0,0 +1,302 @@
+/* Python interface to breakpoint locations.
+
+ Copyright (C) 2011 Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ This program is free software; you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+
+#include "defs.h"
+#include "inferior.h"
+#include "python-internal.h"
+#include "observer.h"
+#include "gdbarch.h"
+
+struct bploc_object
+{
+ PyObject_HEAD
+
+ /* The location corresponding to this gdb.BpLocation object. It's the same
+ idea as gdb.Breakpoint, if the 'backend' location is deleted, LOC is
+ set to NULL. No reference to the location is owned here (in terms of
+ ref. counting) in order not to change the internal behavior. */
+ struct bp_location *loc;
+
+ /* 1 if the owner BP has been deleted, 0 otherwise. */
+ int invalid_owner;
+};
+
+/* Require that LOCATION be a valid bp_location; throw a Python
+ exception if it is invalid. */
+#define BPLOCPY_REQUIRE_VALID(Location) \
+ do { \
+ if ((Location)->loc == NULL) \
+ return PyErr_Format (PyExc_RuntimeError, \
+ _("BpLocation invalid.")); \
+ } while (0)
+
+static PyTypeObject bploc_object_type;
+
+/* Call by free_bp_location when loc is about to be freed. */
+
+void
+gdbpy_bplocation_free (struct bp_location *loc)
+{
+ if (loc->py_bploc_obj)
+ {
+ loc->py_bploc_obj->loc = NULL;
+ Py_DECREF (loc->py_bploc_obj);
+ }
+}
+
+/* Dissociate the bp_location from the Python object. */
+
+static void
+bplocpy_dealloc (PyObject *self)
+{
+ bploc_object *self_bploc = (bploc_object *) self;
+
+ if (self_bploc->loc != NULL)
+ self_bploc->loc->py_bploc_obj = NULL;
+
+ self->ob_type->tp_free (self);
+}
+
+/* Create or acquire a ref to the bp_location object (gdb.BpLocation)
+ that encapsulates the struct bp_location from GDB. */
+
+PyObject *
+bplocation_to_bplocation_object (struct bp_location *loc)
+{
+ bploc_object *bploc_obj;
+
+ gdb_assert (loc != NULL);
+ if (loc->py_bploc_obj)
+ {
+ Py_INCREF (loc->py_bploc_obj);
+ return (PyObject *) loc->py_bploc_obj;
+ }
+
+ bploc_obj = PyObject_New (bploc_object, &bploc_object_type);
+ if (!bploc_obj)
+ return NULL;
+
+ bploc_obj->loc = loc;
+ bploc_obj->invalid_owner = 0;
+ Py_INCREF (bploc_obj);
+ loc->py_bploc_obj = bploc_obj;
+
+ return (PyObject *) bploc_obj;
+}
+
+/* Python function to get the BP owning this location, if any. */
+
+static PyObject *
+bplocpy_get_owner (PyObject *self, void *closure)
+{
+ bploc_object *self_bploc = (bploc_object *) self;
+
+ BPLOCPY_REQUIRE_VALID (self_bploc);
+
+ if (self_bploc->invalid_owner)
+ Py_RETURN_NONE;
+
+ if (self_bploc->loc->owner
+ && self_bploc->loc->owner->py_bp_object)
+ {
+ Py_INCREF (self_bploc->loc->owner->py_bp_object);
+ return (PyObject *) self_bploc->loc->owner->py_bp_object;
+ }
+
+ Py_RETURN_NONE;
+}
+
+/* Python function to test whether or not this breakpoint location is
+ enabled. */
+
+static PyObject *
+bplocpy_get_enabled (PyObject *self, void *closure)
+{
+ bploc_object *self_bploc = (bploc_object *) self;
+
+ BPLOCPY_REQUIRE_VALID (self_bploc);
+
+ if (self_bploc->loc->enabled)
+ Py_RETURN_TRUE;
+
+ Py_RETURN_FALSE;
+}
+
+/* Python function to get the address of this breakpoint location. The
+ gdb.Value object will be cached if this is the first access. Returns
+ NULL in case of failure, with a Python exception set. */
+
+static PyObject *
+bplocpy_get_address (PyObject *self, void *closure)
+{
+ bploc_object *self_bploc = (bploc_object *) self;
+ volatile struct gdb_exception except;
+ struct type *void_ptr_type;
+ struct value *val = NULL; /* Initialize to appease gcc warning. */
+ PyObject *py_address;
+
+ BPLOCPY_REQUIRE_VALID (self_bploc);
+
+ /* Get the address Value object as a void * value. */
+ TRY_CATCH (except, RETURN_MASK_ALL)
+ {
+ void_ptr_type = lookup_pointer_type (
+ builtin_type (python_gdbarch)->builtin_void);
+ val = value_from_pointer (void_ptr_type, self_bploc->loc->address);
+ }
+ GDB_PY_HANDLE_EXCEPTION (except);
+
+ py_address = value_to_value_object (val);
+ if (!py_address)
+ return NULL;
+
+ return py_address;
+}
+
+/* Python function to get the inferior hosting this breakpoint location.
+ Return Py_None if there is no inferior associated with the program space of
+ this location, or NULL in case of failure, with a python exception set. */
+
+static PyObject *
+bplocpy_get_inferior (PyObject *self, void *closure)
+{
+ bploc_object *self_bploc = (bploc_object *) self;
+ struct inferior *inf;
+ PyObject *infobj;
+
+ BPLOCPY_REQUIRE_VALID (self_bploc);
+
+ inf = find_inferior_for_program_space (self_bploc->loc->pspace);
+ if (!inf)
+ Py_RETURN_NONE;
+
+ infobj = inferior_to_inferior_object (inf);
+ Py_XINCREF (infobj);
+
+ return infobj;
+}
+
+/* Python function which checks the validity of a bp location object. */
+
+static PyObject *
+bplocpy_is_valid (PyObject *self, PyObject *args)
+{
+ bploc_object *self_bploc = (bploc_object *) self;
+
+ if (self_bploc->loc)
+ Py_RETURN_TRUE;
+ Py_RETURN_FALSE;
+}
+
+/* Callback triggered when a breakpoint is deleted. This will invalidate
+ the corresponding bp_location Python object owners. */
+
+static void
+bplocpy_breakpoint_deleted (struct breakpoint *b) {
+ struct bp_location *loc;
+
+ for (loc = b->loc; loc; loc = loc->next)
+ {
+ if (loc->py_bploc_obj)
+ loc->py_bploc_obj->invalid_owner = 1;
+ }
+}
+
+/* Initialize the Python bp_location code. */
+
+void
+gdbpy_initialize_bplocation (void)
+{
+ if (PyType_Ready (&bploc_object_type) < 0)
+ return;
+
+ Py_INCREF (&bploc_object_type);
+ if (PyModule_AddObject (gdb_module, "BpLocation",
+ (PyObject *) &bploc_object_type) < 0)
+ return;
+
+ observer_attach_breakpoint_deleted (bplocpy_breakpoint_deleted);
+}
+
+static PyGetSetDef bploc_object_getset[] =
+{
+ { "owner", bplocpy_get_owner, NULL,
+ "Each breakpoint location must belong to exactly one higher-level \
+breakpoint. This pointer is NULL iff this bp_location is no \
+longer attached to a breakpoint (read-only).",
+ NULL },
+ { "enabled", bplocpy_get_enabled, NULL,
+ "Is this particular location enabled.", NULL },
+ { "address", bplocpy_get_address, NULL,
+ "The address at which the breakpoint has been set.", NULL },
+ { "inferior", bplocpy_get_inferior, NULL,
+ "The inferior in which this breakpoint location has been set.", NULL },
+ { NULL } /* Sentinel. */
+};
+
+
+static PyMethodDef bploc_object_methods[] =
+{
+ { "is_valid", bplocpy_is_valid, METH_NOARGS,
+ "Return true if this breakpoint location is valid, false if not." },
+ { NULL } /* Sentinel. */
+};
+
+static PyTypeObject bploc_object_type =
+{
+ PyObject_HEAD_INIT (NULL)
+ 0, /* ob_size */
+ "gdb.BpLocation", /* tp_name */
+ sizeof (bploc_object), /* tp_basicsize */
+ 0, /* tp_itemsize */
+ bplocpy_dealloc, /* tp_dealloc */
+ 0, /* tp_print */
+ 0, /* tp_getattr */
+ 0, /* tp_setattr */
+ 0, /* tp_compare */
+ 0, /* tp_repr */
+ 0, /* tp_as_number */
+ 0, /* tp_as_sequence */
+ 0, /* tp_as_mapping */
+ 0, /* tp_hash */
+ 0, /* tp_call */
+ 0, /* tp_str */
+ 0, /* tp_getattro */
+ 0, /* tp_setattro */
+ 0, /* tp_as_buffer */
+ Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE, /* tp_flags */
+ "GDB breakpoint location object", /* tp_doc */
+ 0, /* tp_traverse */
+ 0, /* tp_clear */
+ 0, /* tp_richcompare */
+ 0, /* tp_weaklistoffset */
+ 0, /* tp_iter */
+ 0, /* tp_iternext */
+ bploc_object_methods, /* tp_methods */
+ 0, /* tp_members */
+ bploc_object_getset, /* tp_getset */
+ 0, /* tp_base */
+ 0, /* tp_dict */
+ 0, /* tp_descr_get */
+ 0, /* tp_descr_set */
+ 0, /* tp_dictoffset */
+ 0, /* tp_init */
+ 0 /* tp_alloc */
+};
diff --git a/gdb/python/py-breakpoint.c b/gdb/python/py-breakpoint.c
index 195ed2b..25ef62f 100644
--- a/gdb/python/py-breakpoint.c
+++ b/gdb/python/py-breakpoint.c
@@ -584,6 +584,43 @@ bppy_get_ignore_count (PyObject *self, void *closure)
return PyInt_FromLong (self_bp->bp->ignore_count);
}
+
+/* Python function which returns the BpLocation objects associated
+ with this breakpoint. */
+
+static PyObject *
+bppy_locations (PyObject *self, PyObject *args)
+{
+ breakpoint_object *self_bp = (breakpoint_object *) self;
+ PyObject *list, *tuple;
+ struct bp_location *loc;
+ int err;
+
+ BPPY_REQUIRE_VALID (self_bp);
+
+ list = PyList_New (0);
+ if (!list)
+ return NULL;
+
+ err = 0;
+ for (loc = self_bp->bp->loc; loc; loc = loc->next)
+ {
+ PyObject *loc_obj = bplocation_to_bplocation_object (loc);
+ err = PyList_Append (list, loc_obj);
+ if (err == -1)
+ {
+ Py_DECREF (list);
+ return NULL;
+ }
+ Py_DECREF (loc_obj);
+ }
+
+ tuple = PyList_AsTuple (list);
+ Py_DECREF (list);
+
+ return tuple;
+}
+
/* Python function to create a new breakpoint. */
static int
bppy_init (PyObject *self, PyObject *args, PyObject *kwargs)
@@ -975,6 +1012,8 @@ static PyMethodDef breakpoint_object_methods[] =
"Return true if this breakpoint is valid, false if not." },
{ "delete", bppy_delete_breakpoint, METH_NOARGS,
"Delete the underlying GDB breakpoint." },
+ { "locations", bppy_locations, METH_NOARGS,
+ "Get a list of gdb.BpLocation objects associated with this breakpoint." },
{ NULL } /* Sentinel. */
};
diff --git a/gdb/python/python-internal.h b/gdb/python/python-internal.h
index 328e5d8..14608c2 100644
--- a/gdb/python/python-internal.h
+++ b/gdb/python/python-internal.h
@@ -165,6 +165,9 @@ typedef struct breakpoint_object
extern breakpoint_object *bppy_pending_object;
+/* Defined in py-bploc.c */
+typedef struct bploc_object bploc_object;
+
typedef struct
{
PyObject_HEAD
@@ -217,6 +220,8 @@ PyObject *pspy_get_printers (PyObject *, void *);
PyObject *objfile_to_objfile_object (struct objfile *);
PyObject *objfpy_get_printers (PyObject *, void *);
+PyObject *bplocation_to_bplocation_object (struct bp_location *loc);
+
thread_object *create_thread_object (struct thread_info *tp);
thread_object *find_thread_object (ptid_t ptid);
PyObject *find_inferior_object (int pid);
@@ -245,6 +250,7 @@ void gdbpy_initialize_pspace (void);
void gdbpy_initialize_objfile (void);
void gdbpy_initialize_breakpoints (void);
void gdbpy_initialize_finishbreakpoints (void);
+void gdbpy_initialize_bplocation (void);
void gdbpy_initialize_lazy_string (void);
void gdbpy_initialize_parameters (void);
void gdbpy_initialize_thread (void);
diff --git a/gdb/python/python.c b/gdb/python/python.c
index 5212d4e..fb88b5e 100644
--- a/gdb/python/python.c
+++ b/gdb/python/python.c
@@ -1101,6 +1101,12 @@ gdbpy_breakpoint_has_py_cond (struct breakpoint_object *bp_obj)
"scripting is not supported."));
}
+void
+gdbpy_bplocation_free (struct breakpoint_object *bp_obj)
+{
+ return;
+}
+
#endif /* HAVE_PYTHON */
@@ -1244,6 +1250,7 @@ message == an error message without a stack will be printed."),
gdbpy_initialize_objfile ();
gdbpy_initialize_breakpoints ();
gdbpy_initialize_finishbreakpoints ();
+ gdbpy_initialize_bplocation ();
gdbpy_initialize_lazy_string ();
gdbpy_initialize_thread ();
gdbpy_initialize_inferior ();
diff --git a/gdb/python/python.h b/gdb/python/python.h
index 9e461f7..516a94c 100644
--- a/gdb/python/python.h
+++ b/gdb/python/python.h
@@ -47,4 +47,6 @@ int gdbpy_should_stop (struct breakpoint_object *bp_obj);
int gdbpy_breakpoint_has_py_cond (struct breakpoint_object *bp_obj);
+void gdbpy_bplocation_free (struct bp_location *loc);
+
#endif /* GDB_PYTHON_H */
diff --git a/gdb/testsuite/gdb.python/py-breakpoint.exp b/gdb/testsuite/gdb.python/py-breakpoint.exp
index 8ed3742..6a1a43b 100644
--- a/gdb/testsuite/gdb.python/py-breakpoint.exp
+++ b/gdb/testsuite/gdb.python/py-breakpoint.exp
@@ -304,3 +304,55 @@ gdb_py_test_silent_cmd "python wp1 = wp_eval (\"result\", type=gdb.BP_WATCHPOIN
gdb_test "continue" ".*\[Ww\]atchpoint.*result.*Old value =.*New value = 788.*" "Test watchpoint write"
gdb_test "python print never_eval_bp1.count" "0" \
"Check that this unrelated breakpoints eval function was never called."
+
+# gdb.BpLocation
+
+# Start with a fresh gdb.
+clean_restart ${testfile}
+
+if ![runto_main] then {
+ fail "Cannot run to main."
+ return 0
+}
+delete_breakpoints
+gdb_test_no_output "set detach-on-fork off" "don't detach on fork"
+gdb_test "call fork()" "New process .*" "create a second inferior"
+
+gdb_breakpoint "main"
+gdb_test "py print len(gdb.breakpoints())" "1" "ensure that threre is only one BP"
+gdb_test_no_output {py bp0 = gdb.breakpoints()[0]} "save breakpoint 0"
+gdb_test "py print len(bp0.locations())" "2" "ensure that threre are 2 locations"
+
+gdb_test_no_output {py loc0 = bp0.locations()[0]} "save location 0"
+gdb_test_no_output {py loc1 = bp0.locations()[1]} "save location 1"
+
+gdb_test "py print loc0.owner == loc1.owner == bp0" "True" "verify ownership"
+gdb_test "py print loc0.address == loc1.address " "True" "verify addresses are identical"
+# how to check address location ? != address(main)
+
+gdb_test {py print loc0.inferior == gdb.inferiors()[0]} "True" "verify inferior for loc 0" #inf 2
+gdb_test {py print loc1.inferior == gdb.inferiors()[1]} "True" "verify inferior for loc 1" #inf 1
+
+gdb_test "py print loc0.enabled == loc1.enabled == True" "True" "verify that locations are enabled"
+
+gdb_test "py print loc0.inferior.num" "2" "ensure that loc0 is on inferior 2"
+
+gdb_test "detach inferior 2" "Detaching from program:.*" "detach inferior 2"
+gdb_test "inferior 1" "Switching to inferior .*" "switch to inferior 1"
+gdb_test_no_output "remove-inferiors 2" "remove inferior 2"
+gdb_test "py print loc0.inferior" "None" "removed inferior set to None"
+
+delete_breakpoints
+gdb_test "py print bp0.is_valid()" "False" "verify that BP has been invalidated"
+gdb_test "py bp0.locations()" ".*RuntimeError: Breakpoint .* is invalid.*"\
+ "verify that locations can't accessed on an invalid breakpoint"
+
+gdb_test "py print loc0.is_valid()" "False" "verify that location is invalid"
+gdb_test "py print loc0.owner" ".*RuntimeError: BpLocation invalid.*"\
+ "verify that owner can't be accessed"
+gdb_test "py print loc0.enabled" ".*RuntimeError: BpLocation invalid.*"\
+ "verify that location can't be accessed"
+gdb_test "py print loc0.address" ".*RuntimeError: BpLocation invalid.*"\
+ "verify that location can't be accessed"
+gdb_test "py print loc0.inferior" ".*RuntimeError: BpLocation invalid.*"\
+ "verify that inferior can't be accessed"
--
1.7.6.5