Auto: Return cor & converge errors earlier than at the end

[SimonDedman]

glm.fit: algorithm did not converge (bin only?)
if (!conv)
warning("glm.fit: algorithm did not converge", call. = FALSE)

glm.fit: fitted probabilities numerically 0 or 1 occurred (bin only)
if (any(mu > 1 - eps) || any(mu < eps))
warning("glm.fit: fitted probabilities numerically 0 or 1 occurred",
call. = FALSE)

In cor(y_i, u_i) : the standard deviation is zero (gaus only?)
from cor, base function

http://blog.obeautifulcode.com/R/A-Warning-About-Warning/
question asked here

Can just use warnings() or print(warnings()) ? After gbm.steps so L393 & 440. Tried