variable style and fix heat

Dear lammps-user,
I am using fix heat command to heat a gold nanoparticle. According to the documentation, eflux parameter can be equal style variable and atom style variable. If eflux is equal style, the eflux determines the change in aggregate energy of the entire group of atoms per unit time. If eflux is atom style, the eflux is is the energy flux for a single atom**. However, in my simulation, variable style seem has no effect on the system temperature.** below is my input file:

units metal
dimension 3
boundary p p p
atom_style full

lattice fcc 4.09
region box block -20 20 -20 20 -20 20
create_box 1 box

region particle sphere 0 0 0 50 units box
create_atoms 1 region particle

pair_style eam

pair_coeff * * Au_u3.eam

neighbor 2.0 bin
neigh_modify every 5 delay 0 check yes

#output
thermo 1000
thermo_style custom step temp pe ke vol press

timestep 0.001
fix 1 all npt temp 300 300 1 iso 0 0 10
run 50000
unfix 1

timestep 0.001
variable laser equal 100 # equal style or atom style with same value give the same output.

thermo 100

fix 21 all heat 1 v_laser
fix 22 all nph iso 0 0 2.0
run 7000
unfix 21
run 100000
unfix 22

If laser is equal style with value of 100, the thermal output is :

56700 381.33876 -117068.31 1511.6841 4399984.6 -249.94925
56800 367.71696 -117006.07 1457.6852 4385565.8 19.692795
56900 364.76474 -116983.59 1445.9822 4377429.9 311.95045
57000 412.65113 -117163.47 1635.811 4389475.3 -264.04444
Loop time of 73.5191 on 60 procs for 7000 steps with 30669 atoms

If laser is atom style with value of 100, the thermal output is :

56700 381.34114 -117068.3 1511.6935 4400003.1 -249.9643
56800 367.71815 -117006.06 1457.69 4385582.2 19.690581
56900 364.76703 -116983.59 1445.9913 4377444.9 311.93687
57000 412.6559 -117163.48 1635.8299 4389489.1 -264.06692

Why the variable style do not affect the thermo output? Is my simulation setup incorrect?
My lammps version is lammps-22Aug18.
Any help will be appreciated.

I can confirm this behavior with the latest development version. this means that either the code or the documentation needs to be updated.

Steve, any suggestion? i already found and fixed a memory access bug when trying to access the global scalar computed by the fix when using an atom style variable for eflux (included in PR #2010).

axel.