It should be solved simultaneously. WLS belongs to GLS. Or GLS concept contains WLS which is a way to solve heteroskedasticity in cross-sectinal data. The Serial correlation can be thought of  a heteroskedasticity problem in time domain.

Both problems violate the homoskedasticity assumption about error term in OLS.

其实 自相关和异方差是同一个问题，应该用FGLS，可行的广义最小二乘法

其实 自相关和异方差是同一个问题，应该用FGLS，可行的广义最小二乘法

哦，学习了。