Icepack code coverage

[eclare108213]

The new codecov test by @anders-dc (#193) does a great job of pointing our the holes in our test coverage. Many of the "misses" are write statements for various failure modes, which we hope never to write (unless we want to include some failures in our testing). By far, the biggest coverage gaps are for zsalinity and skeletal-layer bgc -- I'll put the bgc ones in a separate issue. Here's the rundown of everything else that's missing. Maybe some of these are covered in the full base-suite, and this test was only travis? All of these could be covered in 4 more tests (one per kcatbound value).

namelist:

• l_mpond_fresh = T
• days_per_year=360
• dumpfreq = 'd'
• calc_Tsfc = F
• default_season = 'winter', 'summer'
• update_ocn_f = T
• restore_ocn = T AND trestore=0
• atm_data_type = 'CFS'
• highfreq = T
• kcatbound = -1, 0, 2, 3
• krdg_partic = 0
• krdg_redist = 0
• tr_pond_lvl = T AND frzpond = 'cesm'
• tr_pond_topo = T AND ktherm .ne. 2
• heat_capacity = F
• conduct = 'bubbly' AND ktherm = 1

configuration:

• nslyr > 1
• ncat = 1 (kcatbound = -1)

potential bugs:

1. icepack_atmo.F90, line 652 should be .or. instead of .and.
2. why are lines 830-835 in icedrv_forcing.F90 not used?

[eclare108213]

Most of these options are included in -s alt02, alt03, alt04 in PR #196, and bug 1 is fixed. Still missing:

• calc_Tsfc=F (I don't think this is possible in our configurations -- this is used by UK MetOffice in their coupled system)
• kcatbound=3 (I suggest we remove this option, which I couldn't get to work well enough to publish -- it'll probably generate more questions than it's worth)
• there's some issue with tfrz_option = linear_salt (doesn't seem to matter whether there are quotes around it in namelist, which is suspicious)
• the icedrv_forcing.F90 question remains open

[apcraig]

There is an updated code coverage report based on runs on conrad with the full test suite,

https://codecov.io/gh/CICE-Consortium/Icepack/tree/586247ab794415d85ec5ca91232bb26270351cab

Lets have a look at that and decide if new tests are needed to improve coverage. Also, I am going to create a new PR shortly to finally push the codecov capability to the Icepack consortium master.

[apcraig]

The code coverage tool has been updated in #313 and #314. It can be run by a few people and the results are available at

https://codecov.io/gh/CICE-Consortium/Icepack

and the latest valid results on a file-by-file list is

https://codecov.io/gh/CICE-Consortium/Icepack/list/fcc265787936d49f460e6ab8a549133e116685e0/

We should review those results and see where we can update testing to further improve coverage.

[apcraig]

Just fyi, we move to lcov and the new code coverage output is here, https://apcraig.github.io/