The young LMC cluster NGC 1866 represents the ideal laboratory for testing stellar evolutionary models: in particular, it can be used to discriminate among classical stellar models, in which the extension of the convective regions is fixed by the classical Schwarzschild criterion, from models with overshooting, in which an ``extra-mixing'' is considered to take place beyond the classical limit of the convective zone. Addressing this subject in a recent work, Testa et al. (1999) reached the conclusion that the classical scheme for the treatment of convection represents a good and sufficient approximation for convective interiors. Using their own data, we repeat here the analysis. First we revise the procedure followed by Testa et al. (1999) to correct the data for completeness, second we calculate new stellar models with updated physical input for both evolutionary schemes, finally we present many simulations of the colour-magnitude diagrams and luminosity functions of the cluster using the ratio of the integrated luminosity function of main sequence stars to the number of giants as the normalization factor of the simulations. We also take into account several possible physical agents that could alter the color-magnitude diagram and the luminosity function: they are unresolved binary stars, dispersion in the age, stochastic effects in the initial mass function. Their effect is analyzed separately, with the conclusion that binary stars have the largest impact. The main result of this study is that the convective overshoot hypothesis (together with a suitable percentage of unresolved binaries) is really needed to fully match the whole pattern of data. The main drawback of the classical models is that they cannot reproduce the correct ratio of main sequence to post-main sequence stars.