Re: [RFC - Python] New ObjFile event

[Kevin Pouget <kevin dot pouget at gmail dot com>]

On Mon, May 23, 2011 at 10:43 AM, Kevin Pouget <kevin.pouget@gmail.com> wrote:
> Hello,
>
>
>> Kevin> + ? ? ? ?py-value.o \
>> Kevin> + ? ? ? ?py-newobjfileevent.o
>>
>> Alphabetical, please. ?Applies elsewhere in the Makefile patch too.
>
> fixed, sorry about that
>
>> Kevin> ?# found by configure; if GNU Make is not found, we fall back to a
>> Kevin> diff --git a/gdb/doc/ChangeLog b/gdb/doc/ChangeLog
>> Kevin> index 08d6c8c..3ea5068 100644
>> Kevin> --- a/gdb/doc/ChangeLog
>> Kevin> +++ b/gdb/doc/ChangeLog
>> Kevin> @@ -1,3 +1,9 @@
>> Kevin> +2011-03-29 ?Kevin Pouget <kevin.pouget@st.com>
>> Kevin> + ? ? ? ?* gdb.textinfo (Events In Python): Document `gdb.NewObjFileEvent'
>>
>> Missing blank line after the date line, and typo: "texinfo", not
>> "textinfo".
>
> I removed the Changelog entries from the patch as suggested by Pedro
> in another thread,
> typo fixed at the bottom of the mail
>
>> Kevin> +@item events.newobjfile
>> Kevin> +Emits @code{gdb.NewObjFileEvent} which indicates that a new object-file has
>> Kevin> +been loaded in the inferior.
>>
>> This is not an accurate description -- it may or may not be related to a
>> change in the inferior.
>>
>> E.g., if the user did "set auto-solib-add off", then no objfile will be
>> made when a shared library is loaded.
>>
>> Or, if the user does "add-symbol-file", then a new objfile will be made
>> even though nothing has changed in the inferior.
>>
>> Maybe this isn't the event you actually wanted? ?I mean -- it is fine to
>> have this, and I'd like it if you followed through :-), but it may not
>> do what you are expecting.
>
> you're right, I changed it to:
> ?+been loaded by @value{GDBN}.
>
> About the behaviour, I actually wanted to mimic the way GDB decides to
> enable [linux/thread-db] thread-debugging, ie checking the name of the
> shared library loaded by GDB. As you mentioned that's not perfect,
> "set auto-solib-add off" or "add-symbol-file" might change the
> behaviour, but I guess it's easy enough to cope with it
>
>> Kevin> + ?/* Will be NULL when clearing the symtab. */
>> Kevin> + ?if (objfile)
>> Kevin> + ? ?cleanup = ensure_python_env (get_objfile_arch (objfile), current_language);
>> Kevin> + ?else
>> Kevin> + ? ?cleanup = ensure_python_env (get_current_arch (), current_language);
>>
>> I think it is better to just return early if objfile==NULL.
>> Or, emit a different event in this case.
>
> I found some more documentation about objfile == NULL in the Internals:
>> Called with objfile equal to NULL to indicate previously loaded symbol table data has now been invalidated.
>
> so as you suggested, I just return immediately and leave symbol table
> invalidation for another patch
>
>> Kevin> +2011-03-29 ?Kevin Pouget <kevin.pouget@st.com>
>> Kevin> + ? ? ? ?* testsuite/gdb.python/py-events-shlib.c: New file.
>>
>> Missing newline.
>
> cf above, ChangeLog removed from the diff patch and fixed at the bottom
>
>> Kevin> +# Start with a fresh gdb.
>> Kevin> +
>> Kevin> +gdb_exit
>> Kevin> +gdb_start
>>
>> I think clean_restart is preferred here.
>> But maybe skip_python_tests has to be inserted in the middle of the
>> sequence?
>
> fixed
>
>> Kevin> +gdb_test "run" ".*event type: new_objfile.*
>> Kevin> +.*new objfile name.*"
>>
>> This style of test is hard to read. ?I'd prefer a \n in there.
>
> Fixed (I didn't want to test the '\n' specifically but just make it
> more readable:).
> However, JFYI the test file already contains such formatting:
>
> # Test continue event and breakpoint stop event
> gdb_test "continue" ".*event type: continue.*
> .*event type: stop.*
> .*stop reason: breakpoint.*
> .*breakpoint number: 2.*
> all threads stopped"
>
> #test exited event.
> gdb_test "continue" ".*event type: continue.*
> .*event type: exit.*
> .*exit code: 12.*"
>
>
>
> thanks for your review,
>
> Kevin


Hello,

could you please tell me what you think about this patch, I think I
address all the points you mentioned in your last mail

I added a NEWS entry, and as far as I've seen, the documentation was
not reviewed.

Cordially,

Kevin

2011-03-29  Kevin Pouget  <kevin.pouget@st.com>

	Allow Python notification of new object-file loadings.
	* Makefile.in (SUBDIR_PYTHON_SRCS): Add py-newobjfilevent.c.
	(SUBDIR_PYTHON_OBS): Add py-newobjfileevent.o.
	Add build rule for this file.
	* python/py-event.h (emit_new_objfile_event): New prototype.
	(newobjfile): New Python event emitter.
	* python/py-evts.c (gdbpy_initialize_py_events): Add newobjfile to
	Python event registry.
	* python/py-inferior.c: Include objfiles.h
	(python_new_objfile): New function.
	(gdbpy_initialize_inferior): Add python_new_objfile to the new objfile
	observers.
	* python/py-newobjfileevent.c: New file.
	* python-internal.h (gdbpy_initialize_new_objfile_event): New
	prototype.
	(gdbpy_current_objfile): New global variable.
	* python/python.c (gdbpy_current_objfile): Make global.
	(_initialize_python): Add gdbpy_initialize_new_objfile_event call.

2011-03-29  Kevin Pouget  <kevin.pouget@st.com>

	Allow Python notification of new object-file loadings.
	* gdb.texinfo (Events In Python): Document `gdb.NewObjFileEvent'
	events emitter.
	(Objfiles In Python): Indicate that `gdb.current_objfile' is also set
	during new object-file callbacks.

Attachment: 0001-New-ObjFile-event
Description: Binary data