Re: Crash reports from GNOME bindings

[Gustavo Carneiro <gjc inescporto pt>]

On Mon, 2006-06-19 at 20:11 -0600, Brent Smith wrote:
> Fernando Herrera wrote:
> > On 6/18/06, Gustavo Carneiro <gjc inescporto pt> wrote:
> >>   This sounds like a very good idea.   But could you give more details?
> >> What does the --include option accept?  A string, file name, ...?  I
> >> rather pass information through a pipe, really, anything else is bound
> >> to reach either a cmdline length limit, or force you to create a
> >> temporary file (if done wrong we'll be seeing those security fixes due
> >> to bad tmpfile handling in a few months).
> > 
> > --include points to a filename including the trace. You have also a
> > --kill <pid> command (not working yet) to get your application killed
> > by bug-buddy after the bug report.
> > 
> > I guess that getting a trace in python on mono is not as expensive as
> > the gdb thing, so there would not be a big delay after the crash and
> > the bug-buddy interface coming up. But if we have a big delay we could
> > use instead a named pipe to feed the trace over it, so the bindings
> > can call bug-buddy inmidiately and then getting/feeding the trace
> > while bug-buddy shows the progress bar.
> > 
> 
> What if bug-buddy accepted input from stdin with "--include -"?  Then 
> the caller could use g_spawn_async_with_pipes().

  Sounds good.

> 
> Any security implications there?

  None that I can see.


-- 
Gustavo Carneiro <gjc inescporto pt>
INESC Porto