The Python "No solution" test in the sample test cases actually has infinitely many solutions.

    @test.it('No solution')
    def b():
        tester(solve("x+2y=1", "2x=2-4y"), None)

It seems that, in all the tests, all the solutions are in fact integers (and one can exploit this to work without approximation errors).
The description should mention this; otherwise, I guess different tests should be used.