mcc generates an executable - exe runs but can't actually plot a map located in a matlab toolbox - how do I compile to get the plotting working?

I've tried the following:

mcc -m QuickLook

It generates my executable and when i run it, the GUI opens just fine. When I go to plot a map onto an axes on my figure, it is trying to access the following file:

C:\Program Files\MATLAB\R2010bSP1\toolbox\map\mapdemos\landareas.shp

It cannot find the file and the executable crashes and the figure is left incomplete.

If I just run the GUI in Matlab, it works. If I run the executable in MATLAB (or outside) it behaves the same way and crashes. How do I compile so that this file is included?

My path includes all necessary recursive paths already. I also tried using -I and including a path to the mapdemos folder but that didn't resolve the issue either. Help!

[Matt]

Anyone else have any ideas?

Answer 1

[Walter Roberson]

Did you try using -a to add the file to your build? You might also have to slightly modify the calling sequence of whatever uses the file, in order that it will know where to look for the file. See ctfroot()

[Matt]

I did not. Will give it a try. Thx.

[Matt]

Ok, tried that. Didn't help at all. ctfroot returns

C:\Program Files\MATLAB\R2010bSP1

So it's finding what I believe to be the correct path. I just don't think it's actually loading it!

I added the landareas.shp file directly to my current directory and the mcc then DID load it. (So why not just add all the missing files from the Matlab toolboxes? Because there are too many that I need and because this isn't the "correct" way to do it.)

So for some reason it doesn't appear to be loading the ctfroot recursively. What am I missing here?

[Walter Roberson]

I would not expect ctfroot to return that in compiled mode. In command line mode, maybe. ctfroot would be dynamically updated to be the directory that the CTF file was expanded into, and so would generally be the directory that anything you had used -a for would be placed into.

See also toolboxdir

The demo files would not be expected to be needed by compiled programs, and depfun is not able to deduce that you need files that are named in a string. This is the same basic reason that the %#function pragma exists. depfun is not, for example, able to notice that you need Xyz.m if you use feval('Xyz')

[Matt]

Eeek. I understood most of this...

When I typed ctfroot in Matlab command window, that output was what I saw. I understand it returns something different once compiled but wasn't sure how to see this.

Regarding depfun - is there a way to make MATLAB include demo files? Or was your solution to simply use -a option? I don't know why that didn't do anything for me.

[Walter Roberson]

It is theoretically impossible to find all possible filename references, as strings can be computed.

[Matt]

Well my new question then would be why the -a option didn't appear to do a darn thing for me. I added the file to my current working directory and it DID add it, but it did not add it with the -a option. Any guesses as to why not?

[Walter Roberson]

Sorry, I do not have the MATLAB Compiler to experiment with.

Related information:

http://blogs.mathworks.com/loren/2008/08/11/path-management-in-deployed-applications/

http://www.mathworks.com/matlabcentral/newsreader/view_thread/252484