Re: [cdt-dev] [dsf-gdb] ExprMetaGetVarInfo

[Vladimir Prus <vladimir@xxxxxxxxxxxxxxxx>]

On 10/31/2014 07:44 PM, Marc Khouzam wrote:

> > > > So this design:
> > > > - introduces a new bag-of-data class, adding complexity
> > > > - prevents one MI component, MIExpressions, from accessing additional
> > > > data from MIVariableObject,
> > > >      also MI component
> > > >
> > > > Would it be sensible to eventually eliminate ExprMetaGetVarInfo (and
> > > > possibly related classes)?
> > >
> > > If they can be replaced by something better.  I'm not clear what you
> > > suggest doing to replace them.
> >
> > I suggest to use MIVariableObject directly, or make ExprMetaGetVarInfo
> > have no fields other than instance of MIVariableObject.
>
> Could be nicer, especially since MIVariableObject is a public class.
> However, when looking at calls to the constructor of ExprMetaGetVarInfo,
> not all parameters seem to be taken from an MIVariableObject; but
> maybe you'll be able work around that?

I've pushed a patch to Gerrit that does something like that:

	https://git.eclipse.org/r/41198

That's cleaned up version of what we use internally, and is somewhat conservative for now.
I could make ExprMetaGetVarInfo have no data fields, as suggested above, but I'm not sure
we can remove constructors - that seems like API.

--
Vladimir Prus
CodeSourcery / Mentor Embedded
http://www.mentor.com/embedded-software/