[ERROR] Divergence at iteration: object 'msg' not found

[pachadotdev]

I'm fitting a GEPPML and I found this error

fit_costs <- fepois(
    tij_bar ~ log(dist) + contig + comlang_off + colony | exporter + importer,
    data = d_slice,
    glm.iter = 100
)

NOTE: 293 observations removed because of NA values (LHS: 293).
Warning: feglm: Step halving due to non-finite deviance (7 times).
Error in fepois(tij_bar ~ log(dist) + contig + comlang_off + colony |  : 
  Error in paste0("Divergence at iteration ", iter, ": ", msg, ". Weighted-OLS returns NA coefficients. Last evaluated coefficients with finite deviance are returned for information purposes.") : 
  object 'msg' not found

The cause of the problem is that "msg" exists for problems related to mu/eta, but not to NA coefficients (https://github.com/lrberge/fixest/blob/master/R/ESTIMATION_FUNS.R#L2644)

[pachadotdev]

in #316 I modified the code, to get the correct halving warning

NOTE: 293 observations removed because of NA values (LHS: 293).
The variable 'colony' has been removed because of collinearity (see $collin.var).
Warning: feglm: Step halving due to non-finite deviance (4 times).
Warning messages:
1: In feglm.fit(env = env) :
  Demeaning algorithm: Absence of convergence after reaching the maximum number of iterations (10000).
2: Divergence at iteration 24: non-finite deviance. Step halving: no valid correction found. Last evaluated coefficients with finite deviance are returned for information purposes.

[lrberge]

Thanks again for reporting. Now fixed.