TikZ: Track the bounding rectangle of created nodes (Storing values?)

Question

In the main part of the tikzpicture I'd like to simply create blocks. So I need somewhere in theirs' style definition to perform tracking of their common bounding rectangle. I've tried to access current bounding box, but it is obviously set to node's rectangle, thus, unfortunately, duplicating the path picture bounding box.

I'm going to use this box when creating another type of nodes (say, foo, which is to be implemented shortly). So, again, it should be accessible from inside foo's style definition. The goal is to make this calculations invisible from the user (drawer) perspective, though he will be restricted to declare nodes in two consecutive sections.

Is there a way to do this task in a somewhat more comfortable way than writing it with low-level commands, such as ifdim, etc.? What is preferred way of storing values, when drawing with TikZ? I guess, this is not by using \newcommand, which defines macros available at global scope. Up to now I've managed to store values using \pgfkeys but I stuck at getting the main picture's bounding box.

\documentclass[landscape]{article}

\usepackage{tikz}
\usetikzlibrary{scopes,chains,shapes.multipart,backgrounds,positioning,fit,calc}

\newcommand{\setup}{
    \tikzset{
        node distance=5mm and 3cm,
        block/.style={rectangle, %draw, thick,
            path picture={
                \draw[red, thick] (current bounding box.south west) rectangle (current bounding box.north east);
            }
        },
        %foo/.style={circle},
        every edge/.style={->, rounded corners, thick, draw,
            to path={
                let \p1=(\tikztostart.east),
                    \p2=($ (\tikztostart.east) + (1cm,0) $),
                    \p3=(\tikztotarget.west) in
                (\p1) -- (\p2) \ifdim\y2=\y3  -- \else |- \fi (\p3)
                \tikztonodes
            }
        }
    }
}

\begin{document}

\begin{tikzpicture}
    \setup

    % First section. This blocks will build up the bounding rectangle, which is to be used in the next section.
    \node[block=1] (A) {System A};
    \node[block=2] (B) [above right=of A] {System B};
    \node[block=3] (C) [right=of A] {System C};
    \node[block=4] (D) [below right=of A] {System D};

    % Second section. Positioning of this nodes depends on the resulting bounding rectangle of the previous section.
    %\node[foo] (o1) {circle 1};
    %\node[foo] (o2) {circle 2};

    \path
    (A) edge (B)
        edge (C)
        edge (D);
\end{tikzpicture}

\end{document}

Answer 1

The fit library of TikZ can be used to create a surrounding rectangle or other shaped node of multiple nodes or coordinates.

% Preamble:
\usetikzlibrary{fit}
% ..
\node[inner sep=0pt,outer sep=0pt,fit=(a) (b) (c) (d) (e)] (bb) {};
% use `(bb.north west)` etc. to access the boundaries

What is prefered way of storing values, when drawing with TikZ?

You can use coordinate (<name>) inside a \path to save the current position. Alternative there is the \coordinate (<name>) at (<coordinate>); macro.

Answer 2

Well, I've played more with nodes and styles... The conclusion that inside path picture the current bounding box duplicates path picture bounding box was completely false. It stands what it should for. So, this code works (thanks to global definition of nodes):

block/.style={rectangle,
    path picture={
        \node[rectangle, fit=(current bounding box), inner sep=0pt, outer sep=0pt] (my bounding box) {};
    }
},

Unfortunately, I didn't manage to figure out, why in the example of the initial post each node gets surrounded by red border while that red border should increase in size with the creation of each new node.