Yes, I was wrong. I missed this check in "nmin_filter": ~~~ if (data->curlen <= data->n) return; ~~~ Which means that your patch will never execute the "nmin_filter" (that is where the bug lays) until it is called in...helfper (Helder Pereira)
I suppose I found the problem. Taking the first example: ~~~ [20, 32, 32, 21, 30, 25, 29, 13, 14].min(2) ~~~ This will call the function "nmin_run" in the file "enum.c", which sets "bufmax" to 4 times the number of minimums (n) w...helfper (Helder Pereira)