The following info are included in the zip file:

(1) Few adj matrix examples in this m file and one in the attachment.  Adj matrices may not be PSD. 
(2) Utilities,dfs, bfs, sparse_to_csr, (lightly modified from the original) to generate predecessor for treeplot. 
    see https://github.com/dgleich/gaimc
(3) A modified treeplot that outputs vertex xy.
(4) Attachment K.mat
(5) No existing octave function available to display the network defined by the adj matrix. 
    spy or networkvisualizer are used instead.  Matlab's graph algorithm is the better and typically generates
    best display with minimum edge crossing.

(file #56676)

I cleaned up the code for treeplot (https://hg.savannah.gnu.org/hgweb/octave/rev/dcc639b91ba9).  But, this is really only a start on the issue.

treelayout.m seems to be broken.  I'm going to discuss with Maintainers whether we can just replace it with the code from treeplot.m.

[x,y] = treelayout (p)
x =

    0.2000    0.3000    0.4000    0.5000    0.6000    0.7000    0.8000


y =

    0.2500    0.5000    0.2500    0.7500    0.2500    0.5000    0.2500

Also, can you try this to see if Matlab requires the input to be a row vector?

p = [2 4 2 0 6 4 6]';
treeplot (p)

In Octave, treeplot does not exercise treelayout for the vertex  x/y coordinates.  There is no direct access to the vertex x/y for labeling etc. after the treeplot call.

In matlab, treeplot calls treelayout.  To access vertex coordinates, a duplicated treelayout call will do. 

From my limited experiences, for display, Octave treeplot xy coordinates are better than Octave treelayout.  Matlab's xy from treelayout is better than Octave's xy from treelayout and treeplot. 

This report is to request your consideration for
(a) a consolidation of treeplot and treelayout vertex x/y computation. 
(b) The consolidated xy similar to matlab treelayout. Thanks.