GUIDE gui compatibility between 2015b and 2009b

I have been working on a gui tool using GUIDE in 2009b and it has been functional in that environment. After making a trivial geometry edit in 2015b, I was worried to find that while my gui worked flawlessly in 2015b, it no longer could be run or edited in 2009b, resulting in a pile of errors:

??? Error using ==> hgload at 53
invalid Figure file format

Error in ==> openfig at 72
    [fig, savedvisible] = hgload(filename, struct('Visible','off'));

Error in ==> gui_mainfcn>local_openfig at 286
    gui_hFigure = openfig(name, singleton, visible);

Error in ==> gui_mainfcn at 159
        gui_hFigure = local_openfig(gui_State.gui_Name, gui_SingletonOpt, gui_Visible);

Error in ==> testgui at 42
    gui_mainfcn(gui_State, varargin{:});

After doing a casual browse of the file contents and creating a test figure file, I notice that 2015b appends a hgM_070000 struct when any edits are made. Deleting this allows the gui to work in 2009b again.

I haven't yet found anything in documentation or bug reports about this. Is there a known incompatibility with older versions of Matlab, or is there something I can do to ensure compatibility beyond manually editing the file contents?

EDIT: After dredging the release notes, I see a long history of deprecated or altered functionality WRT gui development. With that in mind, I suppose it's probably expected that some things in particular are incompatible. The question still stands, though; this change to the file appears to break compatibility regardless of the gui content.

I'm wondering if it's even reasonable to bother using GUIDE as a crutch or if I'm better off just figuring out how to build everything programmatically. Alternatively, I suppose I could always just build the gui in 2009b and never ever touch it again for fear of making it error-bomb other users. If there's an intermediate choice, I'd be interested.

Answer 1

[Walter Roberson]

GUIDE has never been designed for backwards compatibility in what it generates.

In particular since R2014b it is likely to generate HG2 objects that cannot be read by older releases.

[DGM]

I kind of had a feeling that was it. I'll probably try a programmatic approach, if not with this project, then the next. Thanks for the insight.