Re: [Bug-readline] SIGSEGV in _rl_dispatch_callback()

[Chet Ramey]

On 3/7/14, 7:07 PM, Anatol Pomozov wrote:
> Hi
> 
> On Fri, Mar 7, 2014 at 3:46 PM, Anatol Pomozov <address@hidden> wrote:
>> Hi
>>
>> On Thu, Mar 6, 2014 at 10:49 PM, Felix Yan <address@hidden> wrote:
>>> Hi,
>>>
>>> I've generated a coredump for this problem, with both python 2.7.6 (patched 
>>> to
>>> compile with readline 6.3) and readline 6.3 enabled debug symbols.
>>>
>>> Link to download: https://paste.xinu.at/QM0Ws/
>>>
>>> I've also tried to debug it somewhat, but I'm not good at it. Just saw 
>>> python
>>> called:
>>>
>>> p = readline_until_enter_or_signal(prompt, &signal);
>>>
>>> and the two args here are fine. Then readline called rl_callback_read_char()
>>> with no arg.
> 
> I compiled python 3.3.5rc and still see this issue. So something weird
> happens between python module and readline that leads to the crash.
> 
> Adding python developers who changed python/Modules/readline.c recently to CC.

>From what I can figure externally, somehow the key sequence context gets
set to NULL but the multikey state (which means that readline is reading a
multi-character key sequence, like the arrow keys) remains enabled.  I'm
not sure how this can happen unless a signal arrives at the wrong time.

A couple of questions: is the session running in vi or emacs mode? Does
this only happen when someone uses the arrow keys to move up and down?

Can someone get a debgging session running against this and set a
watchpoint on _rl_kscxt? When that gets set to 0, the multikey state
should always be turned off.

Chet
-- 
``The lyf so short, the craft so long to lerne.'' - Chaucer
                 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, ITS, CWRU    address@hidden    http://cnswww.cns.cwru.edu/~chet/