Git history cleanup

[phil-blain]

• [x] rebase --autosquash 9/07/20
• [x] rebase on top of upstream/master (will get conflicts, eg. drivers....)
• [x] remove "WIP: TOREMOVE:" commits
• [x] check commit messages for typos ex. "fgmres : crrectly initialize workspace_[xy] and arnoldibasis[xy]" 9/07/20
• [x] check all commit messages 9/07/20
• [x] check what can/should be squashed, and squash them ("WIP" commits)

[phil-blain]

"dynamics: move basal stres residual velocity to ice_dyn_shared"

[phil-blain]

"ice_dyn_vp.F90; remove references to EVP"

[phil-blain]

"ice_dyn_vp: write solver diagnostics only for master_task" -> inside and -> inside an

[phil-blain]

did the autosquash. Interesting Git stuff that I discovered:

• I get a weird conflict with "fixup!" commit (fixup commit applied in the wrong subroutine). Asked for explanations on the Git mailing list: https://lore.kernel.org/git/CABPp-BEu=7xSLk5AE8sQX+S-vvYXxgn+avQ8+9ttdSKDjikb9w@mail.gmail.com/T/#t note: need to understand why the merge-base is as it is...
• should git rebase --keep-base imply --reapply-cherry-picks ??
• git log --merge does not work on non-merge merge conflicts (needs MERGE_HEAD)
• hunk headers are not shown by git diff in case of merge conflicts (git diff --cc) ?

EDIT 13/07/20

$ git -P  diff HEAD:cicecore/cicedynB/dynamics/ice_dyn_vp.F90 picard-doc:cicecore/cicedynB/dynamics/ice_dyn_vp.F90
diff --git a/cicecore/cicedynB/dynamics/ice_dyn_vp.F90 b/cicecore/cicedynB/dynamics/ice_dyn_vp.F90
index cadc331..81197a9 100644
--- a/cicecore/cicedynB/dynamics/ice_dyn_vp.F90
+++ b/cicecore/cicedynB/dynamics/ice_dyn_vp.F90
@@ -669,12 +669,12 @@ subroutine anderson_solver (icellt,   icellu,  &
          byfix    , & ! part of by that is constant during Picard
          umassdti     ! mass of U-cell/dte (kg/m^2 s)

-      real (kind=dbl_kind), dimension(nx_block,ny_block,max_blocks,4), intent(out) :: &
-         zetaD        ! zetaD = 2zeta (viscous coeff)
-
       type (ice_halo), intent(in) :: &
          halo_info_mask !  ghost cell update info for masked halo

+      real (kind=dbl_kind), dimension(nx_block,ny_block,max_blocks,4), intent(out) :: &
+         zetaD        ! zetaD = 2zeta (viscous coeff)
+
       real (kind=dbl_kind), dimension (nx_block,ny_block,max_blocks), intent(inout) :: &
          fpresx   , & ! x fixed point residual vector, fx = uvel - uprev_k
          fpresy   , & ! y fixed point residual vector, fy = vvel - vprev_k
@@ -727,7 +727,6 @@ subroutine anderson_solver (icellt,   icellu,  &
          res_diff   , & ! difference between current and previous residuals
          fpfunc     , & ! current value of fixed point function
          fpfunc_old , & ! previous value of fixed point function
-         Fvec       , & ! (Fx,Fy) (nonlinear residual) as vector
          tmp            ! temporary vector for BLAS calls

       real (kind=dbl_kind), dimension(ntot,im_andacc) :: &

So I did not succeed in getting the same blob add the end... Not that important since it's just a moved argument, and the deletion of an unused variable. Still, better to make sure everything is fine.

[phil-blain]

corrected the messages mentioned above. still need to check all commit messages.

[phil-blain]

I did git rebase --keep-base --reapply-cherry-picks my-first-commit~1 -x 'git commit --amend', with 'my-first-commit' being the first commit I authored. I reread all my commit messages (97) and edited them to make sure they follow best practices:

• use a prefix for the code area
• don't capitalize anything in the commit title
• describe the state before the commit in the present tense
• describe the state after the commit in the imperative mood, as if giving an order to the code base
• describe why the changes are made, not necessarily why (since this can be inferred from the code)

[phil-blain]

I squashed a few other commits together, but left a bunch as is. It's simpler this way.

[phil-blain]

I rebased on top of upstream/master. Got a few conflicts, but nothing difficult. mostly in ice_init because this file is not merge-friendly (since several variables are on the same line in the namelist section).

[phil-blain]

I did another rebase, editing the commit that introduced VP namelist parameters, to play more nicely with the changes in https://github.com/CICE-Consortium/CICE/pull/468

[phil-blain]

I removed the last remaining "WIP: TOREMOVE" commit (most were dropped during the rebase onto upstream/master).

[phil-blain]

I checked that git -P diff HEAD:cicecore/cicedynB/dynamics/ice_dyn_vp.F90 picard-doc:cicecore/cicedynB/dynamics/ice_dyn_vp.F90 gives the same result as above.