[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [cdt-debug-dev] MI-debugger for embedded applications
|
...
> >
> > It is a very long discussion with Oyvind spawning across lots of PR.
> > But basically, the rationale is; the users have finish the debug session
> > and there is no reason to synchronize with the "abort" command and if it
> > fails (and it does many times, for x reasons, gdb can not kill the
> > inferior) we still go through the termination sequence and exit GDB.
>
> Without waiting, the GDB executes the "abort" command and gives an answer. The
> next "postCommand()" reads this answer with a invalid sequence number. Then
> the terminate method doesn't run successful.
> If the "abort" command was launched with "postCommand()" and a suitable
> timeout, the termination sequence should continue even if the "abort" command
> fails.
>
I do not follow, each command as a specific sequence number, whether you
wait() for the answer or not it should not influence the next command, meaning
the next command is waiting for __its__ sequence number and discarding all the others.
To be clearer, all commands are push in a queue, when a responce comes back the
command with the correct sequence number is remove from the queue and notify.
How about to make a PR and attach the MI trace, so we can see what's going on.
> >
> > > Another problem is the terminate() method in MISession.
> > > After the call of "postCommand0(exit, 2000)", eclipse hangs totally.
> > > I don't now why, but the problem dosn't occur with a "Thread.sleep(2000)"
> > > before the "postCommand()"
> >
> > Are you using Windows ?
>
> I'm using NOVELL/SUSE Linux 9.1. It's faster then Windows with cygwin.
>
8-)
Ok.
>
> >
> > > Which way do you prefer for future development of the CDT-debugger? A
> > > dedicated embedded debugger plugin or the enhancement of the MI debugger
> > > interface?
> >
> > Both ...
> > Make a private plugin and for fixes that should be in the general common
> > MI base code; a PR with a patch and the rationale.
>
> O.k. I'll do so.
>
Thanks, that would be great.
