I fixed this to be Matlab compatible on the development branch in this changeset (http://hg.savannah.gnu.org/hgweb/octave/rev/16b9ec39ff46).  Closing report.

Thanks Guillaume for testing with an ancient version of Matlab.  I think it is safe to bring Octave up to date and announce the change in the NEWS file.

One last test

linspace (1, [], 3)

Does this silently return an empty matrix or does it produce an error because the second input is invalid?

Oopsie that was the wrong window... (Octave)

Here's what ML r2014a really says:

>> x = [1:5]';
linspace (x, 10, 0)
ans =
Error using  .*
Matrix dimensions must agree.

Error in linspace ...   <ML code snipped>

Attached is an attempt at a cset for this.  It appears to work. 

One of the things I get from reading the Matlab documentation is that matrices and vectors are not accepted inputs.  I kind of agree that matrices don't make much sense.  Maybe we should delete the implementation for that?  On the other hand, I do kind of like the ability to use vectors.  But if the input is a row vector, shouldn't the result be expanded down rather than across?  It appears that the input is always made a column vector before expanding.

linspace ([1 2 3], 4, 3)
ans =

   1.0000   2.5000   4.0000
   2.0000   3.0000   4.0000
   3.0000   3.5000   4.0000

(file #34730)

Forgot to tack these on the end...

>> class(linspace(1, single(10), 10))
ans =
single
>> class(linspace(1, 10, single(10)))
ans =
double

So only floating point arrays are valid, double or single, and the types of the first two arguments work according to the normal casting rules, type of 3rd argument doesn't matter.

I don't know what behavior to match, but the code in Octave is quite clear.  As an example, liboctave/array/dMatrix.cc begins with

Matrix linspace (const ColumnVector& x1,
                 const ColumnVector& x2,
                 octave_idx_type n)

{
  if (n < 1) n = 1;

which guarantees that at least one element is returned regardless of the value of N.

I think there was already some special handling for linspace when N is small.  According to the documentation,

For compatibility with MATLAB, return the second argument (LIMIT)
if fewer than two values are requested.

What I think is supposed to happen is that for linspace (BASE, LIMIT, 1) it should return LIMIT rather than BASE.  But for linspace (BASE, LIMIT, 0) it does seem that an empty matrix rather than LIMIT should be returned.

Just to check, can the original reporter run the following examples in Matlab and report the results?

linspace (10, 20, 2)
linspace (10, 20, 1)
linspace (10, 20, 0)
linspace (10, 20, -1)

Octave is a little strange because even for -1 it returns 20 which is most likely to be incorrect.

Hello:

In Octave, linspace(10,20,0) returns

ans = 20

while in Matlab, the same order returns

ans = Empty matrix: 1-by-0

I think this is an incompatibility issue. I'm using Octave 4.0, so I don't know if the behavior is also present in old releases