Using solve on a equation of specific format woulld return 0x0 cell unexpectively

[xerubium]

Octave Version: 10.1.0
Symbolic Package version: 3.2.1

Issue:
Solving some equations in this format: 1/kx^n-C==0 by using solve, where k is an positive integer except 1, n is a positive real number except 1 and C is an integers, may return {}(0x0) as if it is not solvable, while it is expected to return the nth root of (Ck).

For example:

>> syms x
>> solve(sym(1)/2*x^2-1==0)
ans = {}(0x0)

The equation remains returning 0x0 unexpectedly as far as the coefficient of x^n can be simplified to the form of 1/k, so in this example, rewriting it to sym(2)/4 or 1/sym(2) or sym(1)/sym(2) won't make a difference. Oppositely, if multiply the whole equation such that the coefficient of x^n can no longer be simplified to the form of 1/k, using solve will return the correct results:

>> solve(3/2*(sym(1)/2*x^2-1)==0)
warning: passing floating-point values to sym is dangerous, see "help sym"
warning: called from
    double_to_sym_heuristic at line 50 column 7
    sym at line 384 column 11
    mtimes at line 54 column 3

ans = (sym 2x1 matrix)

  [   ___]
  [-\/ 2 ]
  [      ]
  [  ___ ]
  [\/ 2  ]

And weirdly, removing the ==0 also return the correct results:

>> solve(sym(1)/2*x^2-1)
ans = (sym 2x1 matrix)

  [   ___]
  [-\/ 2 ]
  [      ]
  [  ___ ]
  [\/ 2  ]

Or rewriting the command to solve(f1,0):

>> solve(sym(1)/2*x^2-1,0)
ans = (sym 2x1 matrix)

  [   ___]
  [-\/ 2 ]
  [      ]
  [  ___ ]
  [\/ 2  ]

But some solve(f1,f2) will weirdly return 0x0 also:

>> solve(sym(1)/2*x^2-1,2*x)
ans = {}(0x0)

which is actually solvable:

>> solve(sym(1)/2*x^2-2*x-1==0)
ans = (sym 2x1 matrix)

  [      ___]
  [2 - \/ 6 ]
  [         ]
  [      ___]
  [2 + \/ 6 ]