In Green, Schwarz, Witten Volume 1, section 3.2, BRST quantization is introduced in a general way. A Lie algebra $G$ is defined with elements $$ [K_i,K_j] = f_{ij}{}^k K_k \tag{3.2.1}$$ where $f_{ij}{}^k$ is the structure constant. Antighosts $b_i$ and ghosts $c^i$ transform in the adjoint and dual adjoint representations respectively. They obey $$ \{c^i,b_j\} = \delta^i_j\tag{3.2.2} $$ The nilpotent BRST operator is $$ Q = c^i K_i - \frac{1}{2}f_{ij}{}^kc^i c^j b_k \tag{3.2.4}$$ The indices here are all summed.
This is then applied to the Virasoro algebra without central charge $$ [L_m,L_n] = (m-n)L_{m+n} $$ where $$ L_m = L^{(\alpha)}_m + L^{(c)}_m - a\delta_m \tag{3.1.58}$$ The ghost contribution is $$ L^{(c)}_m = \sum_{n=-\infty}^\infty(m-n)b_{m+n}c_{-n} \tag{3.1.49} $$ According to the book, the BRST operator is \begin{align*} Q &= \sum_{-\infty}^\infty L^{(\alpha)}_{-m}c_m - \frac{1}{2}\sum_{-\infty}^\infty (m-n):c_{-m} c_{-n} b_{m+n} : -ac_0 \tag{3.2.11}\\ &= \sum_{-\infty}^\infty :(L^{(\alpha)}_{-m} + \frac{1}{2}L^{(c)}_{-m} - a\delta_m)c_m: \tag{3.2.12} \end{align*}
It seems that the antighosts $b_i$ are now $b_m$, and the ghosts $c^i$ are now $c_{-m}$, so that $$\{c_m,b_n\}=\delta_{m+n}\tag{3.1.44}.$$ The non-zero structure constants are $f_{mn}^{(m+n)} = (m-n)$.
My question is: how is the normal ordered ghost term derived? I think the ghost contribution to the Virasoro operator should be normal ordered, since the ordering ambiguity is adsorbed into $a$. So, following equation $(3.2.4)$, \begin{align} Q &= \sum_{-\infty}^\infty c_{-m}L_m - \frac{1}{2}\sum_{-\infty}^{\infty} (m-n)c_{-m}c_{-n}b_{m+n} \\ &= \sum_{-\infty}^\infty L^{(\alpha)}_{-m} c_{m} + \sum_{-\infty}^\infty c_{-m}L^{(c)}_m - ac_0 - \frac{1}{2}\sum_{-\infty}^{\infty} (m-n)c_{-m}c_{-n}b_{m+n} \end{align}
Inserting $L^{(c)}_m$, it seems that \begin{align} \sum_{-\infty}^\infty (m-n) c_{-m}:b_{m+n}c_{-n}: - \frac{1}{2}\sum_{-\infty}^{\infty}(m-n)c_{-m}c_{-n}b_{m+n} &= -\frac{1}{2}\sum_{-\infty}^\infty (m-n) : c_{-m}c_{-n}b_{m+n}: \end{align}
I am unsure as to how to relate these. It seems that trying to combine the two terms will give a $-3/2$ since the operators are anticommuting, as well as a divergence sum. Perhaps equation $(3.1.49)$ is off by some factor? Above that equation it is written that $L_m = \frac{1}{\pi}\int^\pi_{-\pi} d\sigma e^{im\sigma} T_{++}$, and computing this I get $i$ times equation $(3.1.49)$.
Any help is appreciated, thanks.