In Ahlfors' Complex Analysis text, the Bernoulli numbers, $B_k$, are defined as the coefficients in a Laurent development: $$(e^z-1)^{-1}=\frac{1}{z}-\frac{1}{2}+ \sum_1^\infty (-1)^{k-1} \frac{B_k}{(2k)!} z^{2k-1}. $$ (I'm aware that this definition is different from the modern one on Wikipedia.)
Over the course of proving Stirling's formula, the author states that it can be shown that for all $\nu \geq1$ $$ (-1)^{\nu-1} \frac{1}{\pi} \int_0^\infty \eta^{2\nu-2} \log \left( \frac{1}{1-e^{-2 \pi \eta}} \right) \mathrm{d} \eta=(-1)^{\nu-1} \frac{1}{(2\nu-1)2\nu} B_\nu$$ "by means of residues".
I have tried showing that by reaching the function in the definition of the Bernoulli numbers, using integration by parts. I found that $$ \frac{1}{\pi} \int_0^\infty \eta^{2\nu-2} \log \left( \frac{1}{1-e^{-2 \pi \eta}} \right) \mathrm{d} \eta=\frac{2}{2\nu-1} \int_0^\infty \eta^{2\nu-1} (e^{2 \pi \eta}-1)^{-1} \mathrm{d} \eta $$ Sadly, I can't find any point $\eta$ with a residue containing $B_\nu$... Maybe if the powers of $\eta$ were in the denominator I could have done something.
Can anyone help me prove this identity?