Re: -var-info-path-expression

[Vladimir Prus <ghost@cs.msu.su>]

On Thursday 04 January 2007 01:39, Daniel Jacobowitz wrote:
> On Mon, Dec 25, 2006 at 12:00:42PM +0300, Vladimir Prus wrote:
> > 
> > This patch implements new MI command, -var-info-path-expression,
> > that, given a variable object, returns full expression that corresponds
> > to it.  Both KDevelop and Eclipse have now code that guesses such
> > full expression, and that code is rather hacky, and not exactly correct.
> > 
> > Moreover, as soon as MI is taught to get the true type 
> > polymorphic C++ objects and display fields of the real type, such full expression 
> > cannot be computed in frontend at all.
> > 
> > The essence of this patch -- which expressions are returned in which cases -- are
> > ported without change from the Apple branch. The code structure is much
> > different though -- this patch capitalizes on MI refactoring patches I've posted
> > recently.
> > 
> > This is lightly tested by hand, I'll write automated tests later. There's no docs
> > either -- again, will be written after discussion.
> > 
> > There is a couple of issues with this patch:
> > 
> > 	- I don't much like 'var-info-path-expression' name, but
> > 	naming of MI commands is not very important.
> 
> True - I think it's fine.
> 
> > 	- I'm not sure why we can't report full expression in the
> > 	output of -var-list-children. The code I have does not seem
> > 	very computationally expensive.
> 
> Also true.  If this would be more useful, I'd be happy to do it that
> way - would you still need -var-info-path-expression?

No, I think -var-info-path-expression will not be needed then. Except if we
do it that way, we'll have "all of test are broken because they don't expect
an extra field" situation. Grr. I'm not sure what to do here.

- Volodya