summary(brmfit) of any fits from brm() gives me errors, but I can still access the fit by just typing fit object. Is it normal? It happens to any toy examples.
I think there is something wrong with the C++ compiling, but any subsequent tasks can be carried out because posterior samples are still extracted from fit object.
I figured it out.
I found that rstan::summary() works, which means that I was using base::summary() on brmsfit object. So, I just remove all the objects and packages installed, and start over with only library(brms) and no library(rstan). Then, summary() works. So, it was caused by R confusing with three summary functions in the memory.