there are many, many problems in this code style which lead to this kind of problem.
localised comments may help but naming things properly would help more (and negate a comment being required at all), actually adopting some better practices would be a better fix going forward.
looking at the called function's comment:
The counters should be initialized to 1, 5, and 7 before
* calling this for the first time.
so how about making a function with the same name, ending in '_first' (maybe rename the original _next) which initialises these variables internally before calling the version intended to be called multiple times...?
since the parameters are not well described by their names renaming them would help too... it is not a pointer to a 3 a 5 or a 7. it might not even be pointing at multiples... but current_multiple_of_three etc. would be better. there is some reason why these values are significant in this algorithm (which I do not know) and the best names would capture that.
the code is moderately difficult to read in the function itself - the method of swapping around a local copy of a pointer based on the conditions is not very easy to follow. although makes sense here to avoid excessive code... comments would help make it clearer what is going on.
localised comments may help but naming things properly would help more (and negate a comment being required at all), actually adopting some better practices would be a better fix going forward.
looking at the called function's comment:
The counters should be initialized to 1, 5, and 7 before * calling this for the first time.
so how about making a function with the same name, ending in '_first' (maybe rename the original _next) which initialises these variables internally before calling the version intended to be called multiple times...?
since the parameters are not well described by their names renaming them would help too... it is not a pointer to a 3 a 5 or a 7. it might not even be pointing at multiples... but current_multiple_of_three etc. would be better. there is some reason why these values are significant in this algorithm (which I do not know) and the best names would capture that.
the code is moderately difficult to read in the function itself - the method of swapping around a local copy of a pointer based on the conditions is not very easy to follow. although makes sense here to avoid excessive code... comments would help make it clearer what is going on.