Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
RE: [cdt-dev] About default debuggers, etc

To say "if we only have TCF on host" indicates that you are misinterpreting what I'm saying. TCF is flexible. For any particular environment, you can have agents on boards, on probes, or on host machines, or some combination of these. Where you put them depends on how much you want them to do and the resource constraints of the devices. Agents on probes typically need a small footprint. You can complement what they do with an agent on the host.

John

At 11:19 AM 2/5/2010, Dominique Toupin wrote:

A lot of the TCF info (e.g. http://wiki.eclipse.org/DSDP/TM/TCF_FAQ) is about a TCF agent on target and about the communication between the host and target, the reference implementation was a target agent. If we only have TCF on host aren't we missing TCF benefits like auto-discovery, tunneling via other targets, etc. maybe Felix B. can give us some light.

> -----Original Message-----
> From: cdt-dev-bounces@xxxxxxxxxxx
> [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of John Cortell
> Sent: 5-Feb-10 11:38
> To: CDT General developers list.; CDT General developers list.
> Subject: RE: [cdt-dev] About default debuggers, etc
>
> TCF agents need not reside on the target; they can live on the host.
>
> John
>
> At 10:33 AM 2/5/2010, Dominique Toupin wrote:
>
> >It's not just a nice to have, if we add a TCF agent in
> addition to the
> >gdbserver we now need 2 processes instead of 1, doubling the debug
> >processes on the target is not good for many embedded systems (for
> >Windows and native debug this should be OK).
> >For many embedded systems the overhead, memory, etc. is important,
> >adding extra overhead for debug is not an option.
> >This is why we added TCF directly in the Linux tracing daemon
> >(equivalent of gdbserver but for tracing) not as a separate process.
> >
> >If we want TCF to pick up we need to have low overhead
> implementation,
> >with a few companies on board, gdbserver/GDB should talk TCF
> >efficiently.
> >
> > > -----Original Message-----
> > > From: cdt-dev-bounces@xxxxxxxxxxx
> > > [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of John Cortell
> > > Sent: 5-Feb-10 11:03
> > > To: CDT General developers list.; CDT General developers list.
> > > Subject: RE: [cdt-dev] About default debuggers, etc
> > >
> > > I agree. It would be nice if gdbserver talked TCF but I
> just don't
> > > see the incentive. GDB has been around much longer than
> TCF and is
> > > pervasive in both the desktop and embedded OS world. More
> likely is
> > > the use of TCF agents that act as a bridge to the gdb
> world, as per
> > > what Nokia contributed for Linux.
> > >
> > > John
> > >
> > >
> > > At 09:55 AM 2/5/2010, Dominique Toupin wrote:
> > >
> > > >We are also interested in TCF, actually we did a few
> > > improvements that
> > > >are now merged in the TCF repository (thanks to Michael
> > > >Sills-Lavoie) e.g.:
> > > >  - TCF binary transfer;
> > > >  - allow TCF services to be distributed outside of the
> TCF source
> > > >tree; Those TCF additions allowed us to created a Linux
> tracing TCF
> > > >agent, you can find a high level view at
> > > >http://git.dorsal.polymtl.ca/?p=lttng-dev-doc.git;a=blob_plai
> >n;f=arch_g
> > > >raph/lttng-arch.pdf
> > > >
> > > >Regarding debug it would be nice to have gdbserver/GDB talk TCF,
> > > >Ericsson alone might have a hard time convincing the GDB
> > > community but
> > > >if we are a few we could come up we a good plan, if other are
> > > >interested please let me know.
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: cdt-dev-bounces@xxxxxxxxxxx
> > > > > [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of John Cortell
> > > > > Sent: 4-Feb-10 20:57
> > > > > To: CDT General developers list.; CDT General developers list.
> > > > > Subject: Re: [cdt-dev] About default debuggers, etc
> > > > >
> > > > > At 07:49 PM 2/4/2010, Doug Schaefer wrote:
> > > > > >I'd really like the Nokia gang to state what they think the
> > > > > >strategy for EDC should be, or what they are intending
> > > it for since
> > > > > >they are investing pretty heavily in it. The one
> thing I think
> > > > > >is
> > > > > firm is that
> > > > > >it helps those who need to avoid GPL, and there is
> > > > > commercial need for
> > > > > >that from some vendors. As I also mentioned, I'm curious
> > > whether it
> > > > > >would make implementing a JNI debug story more easily as we
> > > > > have full
> > > > > >control over what it's doing.
> > > > >
> > > > > Doug, EDC provides a DSF-based solution for TCF agents.
> > > The hope is
> > > > > that TCF will end up being adopted widely as a "standard"
> > > > > run-control API, allowing for easier integration between
> > > silicon and
> > > > > tools vendors. Freescale is very interested in this
> technology.
> > > > >
> > > > > John
> > > > >
> > > > >
> > > > > _______________________________________________
> > > > > cdt-dev mailing list
> > > > > cdt-dev@xxxxxxxxxxx
> > > > > https://dev.eclipse.org/mailman/listinfo/cdt-dev
> > > > > _______________________________________________
> > > >cdt-dev mailing list
> > > >cdt-dev@xxxxxxxxxxx
> > > >https://dev.eclipse.org/mailman/listinfo/cdt-dev
> > >
> > >
> > > _______________________________________________
> > > cdt-dev mailing list
> > > cdt-dev@xxxxxxxxxxx
> > > https://dev.eclipse.org/mailman/listinfo/cdt-dev
> > > _______________________________________________
> >cdt-dev mailing list
> >cdt-dev@xxxxxxxxxxx
> >https://dev.eclipse.org/mailman/listinfo/cdt-dev
>
>
> _______________________________________________
> cdt-dev mailing list
> cdt-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/cdt-dev
> _______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cdt-dev




Back to the top