[petsc-dev] PETSc developers who use weird MPI

[Satish Balay]

On Sat, 2 Nov 2013, Jed Brown wrote:

> Satish Balay <balay at mcs.anl.gov> writes:
> > Should 'prbrune/sf-sfbasicops' and 'prbrune/mat-matcolor' be cleaned
> > up to remove the merge between the 'rebased' and 'non-rebased'
> > variants of the branches - before the changes got into master?
> 
> This looks possible.
> 
> *   5c44807 (prbrune/mat-matcolor) Merge branch 'prbrune/sf-sfbasicops' into prbrune/mat-matcolor
> |\  
> | *   8e99f7e (prbrune/sf-sfbasicops) Merge branch 'prbrune/sf-sfbasicops' of bitbucket.org:petsc/petsc into prbrune/sf-sfbasicops
> | |\  
> | | * f3fe83a mpiuni: change types to use unique IDs [for sf]
> | | * 13d1d6f mpiuni: add MPI_OP definitions required for sfbasic
> | * | a49f225 Added several additional MPI_Ops to PetscSFBasic.
> |  /  
> * | 3142415 Fix for multiple definitions in mat/color/interface/ custom fortran bindings
> 
> 
> $ diff -u  <(git show --format=fuller 53deab3) <(git show --format=fuller a49f225)                                                            
> --- /dev/fd/63  2013-11-02 13:40:30.964494674 -0600
> +++ /dev/fd/62  2013-11-02 13:40:30.964494674 -0600
> @@ -1,8 +1,8 @@
> -commit 53deab39b05c95f82ded0f9b5ed98e1eca77ec57
> +commit a49f225f21e5e022456a96fbd0a3a873045b2c5d
>  Author:     Peter Brune <brune at mcs.anl.gov>
>  AuthorDate: Wed Oct 9 14:54:51 2013 -0500
>  Commit:     Peter Brune <brune at mcs.anl.gov>
> -CommitDate: Wed Oct 9 14:54:51 2013 -0500
> +CommitDate: Thu Oct 17 13:38:04 2013 -0500
>  
>      Added several additional MPI_Ops to PetscSFBasic.
> 
> 
> $ git branch -r --contains a49f225
>   origin/prbrune/mat-matcolor
>   origin/prbrune/sf-sfbasicops
> 
> 
> Peter's rebased commit is a49f225, but it does exactly the same thing as
> the original (53deab3) and the only other branch it has been used by is
> 'prbrune/mat-matcolor', so it should be okay to:
> 
> $ git checkout prbrune/sf-sfbasicops
> $ git reset --hard f3fe83a        # gets rid of rebased version and merge
> 
> 
> $ git checkout prbrune/mat-matcolor
> $ git reset --hard HEAD^          # moves back to 3142415
> $ git merge prbrune/sf-sfbasicops
> 
> 
> Test that the above works, then force-push those two branches.
> 
> $ git push origin +prbrune/sf-sfbasicops +prbrune/mat-matcolor

barry/saws is now in master. This brought in prbrune/mat-matcolor
snapshot prior to the mpiuni fix into master.

How do we proceed here? [to get mpiuni fixes into master aswell]?
[i.e Peter do the above cleanup and merge to master - or merge to master as is?]

thanks,
Satish