I am reading a paper in which the authors derive the asymptotics for the following series: $$ \sum_{k=0;\,k+=2}^{N(1+m)-2}\left(1-\frac{k}{2}\log{\frac{k+2}{k}}\right)\sim \frac{1}{2}\log{\left(\pi N(m+1)\right)+\mathcal{O}(1/N)}\,, $$ where $m$ is a real number and $-1\le m\le 1$. For completeness I report here what they write about this formula: "where the last line can be obtained by recognizing that the sum can be written as a convergent part plus a divergent sum which is the harmonic number, and then using the asymptotic expression of the latter".
I am looking for an argument to get the result they provide, do you have any suggestion?
Suggested method: Using the quadratic approximation $\log(1+x) = x - \frac{x^2}2 + O(x^3)$, rewrite $$ \log\frac{k+2}k = \log\biggl( 1+\frac2k \biggr) = \frac2k - \frac2{k^2} + O\biggl( \frac1{k^3} \biggr) $$ and examine the sum that results.