[MITgcm-devel] new TAF version

[Jean-Michel Campin]

Hi,

This morning (~8.am), staf switched to a new TAF version: 2.9.0
(was 2.8.16 before)
This does introduce little changes (at machine truncation level) as
shown by testreport run on ref. platform (i.e., baudelaire with gfortran)
where all except 1 experiment pass with 13 or more identical digits for AD-grad.

The only one that now fails (@ 6 digits) is the main lab_sea exp.
but it has always been very sensitive.

Since I am not ready to update the lab_sea output_adm (+ tlm ?) now,
and since Martin made a little changes in pkg/seaice today, I will 
force lab_sea to stay with version 2.8.16 for tonight tests ;
And sometime this WE will move back to (new) default TAF version for lab_sea 
(+ update the output_adm.txt there).

Cheers,
Jean-Michel

Experiments that show some changes after switching to TAF 2.9.0
(were all 16 with previous TAF version):
Y Y Y Y 16>13<16 pass  1D_ocean_ice_column  (e=0, w=2)
Y Y Y Y 16>14<16 pass  global_ocean.90x40x15  (e=0, w=13)
Y Y Y Y 16>14<16 pass  global_ocean.90x40x15.bottomdrag
Y Y Y Y 16>15<16 pass  halfpipe_streamice  (e=0, w=8)
Y Y Y Y 16> 6<16 FAIL  lab_sea  (e=0, w=7)
Y Y Y Y 16>14<16 pass  lab_sea.noseaicedyn
Y Y Y Y 16>15<16 pass  obcs_ctrl  (e=0, w=0)