CVX3: bug report in kron

1

I’d like to report a bug in CVX 3.0 (version downloaded from website mid-Feb. 2017).

The following crashes:

cvx_begin
variable X(2,2) symmetric;
variable y(2);
kron(X, eye(2));
kron(y, ones(2, 1));
cvx_end

with the following output:

Error using reshape (line 16)
Size vector must have at least two elements.

Error in find (line 35)

Error in kron (line 31)

Error in bug (line 4)
kron(X, eye(2));
 
29  error( e2 );

As a workaround, use the following kron implementation, written using an outer product:

function res = workaroundkron(x, y)
    x1 = size(x, 1);
    x2 = size(x, 2);
    y1 = size(y, 1);
    y2 = size(y, 2);
    res = y(:) * x(:)';
    res = reshape(res, [y1 y2 x1 x2]);
    res = permute(res, [1 3 2 4]);
    res = reshape(res, [y1*x1 y2*x2]);
end

Note: I have done minimal testing of this workaround, use at your own risk.

2

Output of cvx_version (private info < removed > ):

---------------------------------------------------------------------------
CVX: Software for Disciplined Convex Programming       (c)2014 CVX Research
Version 3.0beta, Build 1175 (1326ef2)              Mon Nov 23 14:29:34 2015
---------------------------------------------------------------------------
Installation info:
    Path: <removed>
    MATLAB version: 8.6 (R2015b)
    OS: Linux amd64 version 4.4.0-62-generic
    Java version: 1.8.0_121
Verfying CVX directory contents:
    WARNING: The following files/directories are missing:
        <removed>/examples/ + 504 files, 30 subdirectories
        <removed>/mosek/glx/ + 2 files, 0 subdirectories
    These omissions may prevent CVX from operating properly.
Loading preferences:
Preferences already loaded:
    Local: <removed>
License host:
    Username: <removed>
    Host ID: <removed>
Installed license:
    No license installed.
No valid licenses found.
    <removed>
    for the username and first hostid listed above.
---------------------------------------------------------------------------
3

May I ask why would you use version 3.0?

Any status on the development of CVX 3.0?
I wish to see it goes forward.

4

I wanted to test more solvers, and have control over the primal/dual form (cvx_dualize).