[RFC][0/5] Python event handling

[RFC][0/5] Python event handling

[Oguz Kayral]

Hi,
This patch series adds inferior event handling support to GDB Python
scripting. The work is based on archer-tromey-python branch.

An example gdb session:

gsoc-gdb/install/bin/gdb ./simple
GNU gdb (GDB) 6.8.50.20090717-cvs
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
(gdb) require command check_segfault
(gdb) start
Temporary breakpoint 1 at 0x80483a5: file simple.c, line 7.
Starting program: /home/oguz/simple

Temporary breakpoint 1, main () at simple.c:7
7	    *p = simple();
(gdb) check_segfault
SIGSEGV handler registered.
(gdb) cont
Continuing.

Program received signal SIGSEGV, Segmentation fault.
0x080483af in main () at simple.c:7
7	    *p = simple();
event type : stop
stop reason: signal
stop signal: SIGSEGV
#0  0x080483af in main () at simple.c:7

Re: [RFC][0/5] Python event handling

[Tom Tromey]

>>>>> "Oguz" == Oguz Kayral <oguzkayral@gmail.com> writes:

Oguz> This patch series adds inferior event handling support to GDB Python
Oguz> scripting. The work is based on archer-tromey-python branch.

Hi Oguz.

I'm sorry about the long delay in my response here.  I spent most of
August on another part of the Python interface and didn't have much time
left over.

I have a few comments on the patches.  I will reply to the patches
needing comment.

One bigger thing, not specific to any particular patch, is that we like
to have documentation for all new Python APIs.

Tom

Re: [RFC][0/5] Python event handling

[Richard Ward]

Oguz Kayral wrote:
> Hi,
> This patch series adds inferior event handling support to GDB Python
> scripting.

Hi Oguz, I've been looking forward to this functionality, Good stuff!

I have a few comments though, if you don't mind.


Firstly two (fairly  unimportant) things about the function registration.

I think it would be nice if you could give upvalues to the registration 
function (similar to gobject's `connect'), though admittedly it is very 
easy to fake that in python using a class that defines __call__.

Also, I'm not sure whether this would ever be an issue but it seems 
advantageous to have the registration function return a unique object or 
number that could subsequently be used to unregister your callable, 
rather than using the callable object itself.


Next, in python-stopevent.c:
emit_stop_event (struct bpstats *bs, const char *stop_signal)
{
   /*...snip...*/
   for (i = 0; i < PyList_Size (callback_list); i++)
     {
       PyObject_CallObject (PyList_GET_ITEM (callback_list, i), args_tuple);
     }
}

A couple of issues with this. First, PyObject_Callback returns a new 
reference which is the return value of the function (or NULL), so it 
must be handled otherwise it will leak.

Secondly the user code may throw an exception, or the object may not be 
callable. The exception should be cleared, and it would be nice to also 
print the exception (PyErr_Print does both).

Most important here though is the iteration through the list. If the 
user disconnects a callback from inside a callback then the some 
callbacks could be skipped. It could be better to copy the list first, 
but that could still lead to some confusing behaviour, for example a 
callback being called after the user thinks it ought to have been 
disconnected.

Thanks,
Richard