The standard method to determine the transition temperature (T g ) of glass transition is the jump in the specific heat ∆C p . Despite this importance, standard theory for this jump is lacking.The difficulties encompass from lack of proper treatments of specific heat of liquids, hysteresis, to the timescale issue. The first part of this paper provides a non-empirical method to calculate specific heat in the glass transition, with resolving these difficulties. The method consists of molecular dynamics (MD) simulations based on density-functional theory (DFT) and thermodynamics methods. The total-energy approach based on DFT, in which the total energy is the most reliable energy for any state of matter, enables us to calculate specific heat, irrespective of solids or liquids.A serious problem for glass-transition states is involvement of complicated energy dissipation processes. This problem is resolved by employing adiabatic MD simulations, by which the relationship between the internal energy and equilibrium temperature is calculated. The problems of hysteresis and the timescale issue are alleviated by restricting the scope of calculations to equilibrium states only. The second part of this paper describes an application of the theory to the specific-heat jump of glycerol in order to show the validity of the methods. In spite of severe limitations due to the small size of supercells, a reasonable value for the specific-heat jump is obtained. By decomposing ∆C p into contributions of the structural energy, phonon, and thermal expansion parts, we have a sound interpretation for the specific-heat jump: the major contribution to ∆C p comes from the change in the structural energy. From this, a neat energy diagram about the glass transition is obtained: this greatly help our understanding on the glass transition. An outcome of this study is verification of the empirical relationship between the fragility and specific-heat jump, each reflecting the change in the energy barrier and the change in the internal energy, respectively. These two energies are scaled by the ratio k = T g /∆T g , where ∆T g is the width of the transition, through which the two quantities are interrelated. This is useful for organizing various relationships that are found empirically.