It is well-known how to canonically quantize the Lagrangian
$$L = i \bar{\psi} \dot{\psi} - \omega \bar\psi \psi$$
I now wonder how one quantizes the Lagrangian with one real fermion
$$L = i \psi \dot\psi$$
Obviously there can be no mass term since it is anticommuting so $\psi \psi = 0$.
I find no obstacles when going through Dirac's procedure. First I get the conjugate momentum as
$$\pi = i \psi$$
which I view as a constraint since there is no time derivative in this relation,
$$\Phi = \pi - i \psi = 0$$
Next I construct the Dirac bracket (DB) from the Poisson brackets (PB)
$$\{\psi,\pi\}_{PB} = 1$$
$$\{\psi,\psi\}_{PB} = 0$$
$$\{\pi,\pi\}_{PB} = 0$$
by following the standard procedure. First I define
$$C = \{\Phi,\Phi\}_{PB} = - 2 i$$
and then
$$\{\psi,\psi\}_{DB} = \{\psi,\pi\}_{PB} C^{-1} \{\pi,\psi\}_{PB} = 1\cdot(-2i)^{-1}\cdot 1 = i/2$$
Quantizing amounts to replacing DB by anticommutator as
$$[\psi,\chi]_+ = i\hbar\{\psi,\chi\}_{DB}$$
In this case this gives
$$[\psi,\psi]_+ = i\hbar(i/2) = - \hbar/2$$
This amounts to
$$\psi \psi = - \hbar/4$$
contradicting the fact that $\psi \psi = 0$.
Is there no way out of this? Is it impossible to canonically quantize this theory? The path integral seems to exist and make sense.