Re: [cdt-dev] Disassembly View

[Doug Schaefer <cdtdoug@xxxxxxxxx>]

On Tue, Sep 14, 2010 at 4:56 PM, John Cortell <rat042@xxxxxxxxxxxxx> wrote:
> At 03:25 PM 9/14/2010, Doug Schaefer wrote:
>>
>> On Tue, Sep 14, 2010 at 3:31 PM, John Cortell <rat042@xxxxxxxxxxxxx>
>> wrote:
>> > That's what I figured you were hinting at, but TCF is an
>> > Eclipse-agnostic,
>> > lower-level framework that needs something to glue it to Eclipse. DSDP
>> > is
>> > that glue for several of us. What you're using or planning on using, if
>> > not
>> > DSDP, is the mystery.
>>
>> org.eclipse.tm.tcf.debug. TCF is a services framework, don't
>> particularly want another one between it and the user, especially with
>> some of the complex workflows we want to support. Felix and Eugene can
>> talk more about that at the summit. We didn't get to finish the
>> conversation we started at EclipseCon.
>
> Creating a custom glue layer between TCF and Platform Debug is certainly an
> option, but I would think most companies are not eager to reinvent that
> wheel. After all, that isn't a trivial layer. DSF (sorry, not sure why I
> said DSDP before) certainly isn't a simplification, but it does provide a
> lot of common function. Layers are not inherently bad. Otherwise, why not
> just tie gdb directly to Platform Debug. I think you can see where I'm
> going. Anyway, I'm very curious to hear more on this next week.

It's definitely a different philosophy. The idea is that with TCF
standard services definitions, everything between it and Platform
debug is common. All agents are supposed to follow the protocol and
behave the same way. CDI and DSF exist to support multiple agent
protocols and behaviors and that gives rise to their complexity. If
everyone talked MI and talked exactly the same way, I think we would
have ended up with something simpler (or CDI would have done just
fine).