strange behavior temp/rescale, compute temp and ave/chunk with changes Nrepeat

Dear , I have a question .

I attached the in.melt script. When I change the Nrepeat parameter used in the fix: ave_1 and ave_2 , I get different values in ​​temp_all compute .

with Nrepeat = 1,I get the following(temp2.txt) :

Fix print output for fix print_2

11 1.2000000000000050626
12 1.1999999999999999556
13 1.200000000000001954
14 1.1999999999999995115
15 1.2000000000000012879
16 1.1999999999999972911
17 1.2000000000000001776
18 1.1999999999999952927
19 1.199999999999997069
20 1.2000000000000001776

which is the expected result. On the other hand , if I use Nrepeat = 10 ( in ave_1 and ave_2 ) I get the following(temp2.txt):

Fix print output for fix print_2

11 0.93761672873587798005
12 0.86984101699425353527
13 0.79960147604907094721
14 0.73083111779263265628
15 0.66790218602818351989
16 0.6147833639242605841
17 0.57422234871493982311
18 0.54728152284890385637
19 0.53334749378502577244
20 0.53052800888141260494

Furthermore latter behavior is corrected if I use uncompute chunk_1 .

I do not understand why ave / chunk and chunk / atom affects the behavior of temp / rescale , even when chunk_1 is not used in the second run . Is this a bug?

This was tested in the stable versions of lammps 10Aug15 , 7dec15 and 16feb16 .

Regards

in.melt (1.12 KB)