[petsc-dev] [tao-comments #191525] PetscNewLog signature change

[Satish Balay]

On Fri, 13 Dec 2013, Satish Balay wrote:

> On Fri, 13 Dec 2013, Satish Balay wrote:
> 
> > On Fri, 13 Dec 2013, Geoffrey Irving wrote:
> > 
> > > On Fri, Dec 13, 2013 at 11:49 AM, Geoffrey Irving <irving at naml.us> wrote:
> > > > On Thu, Dec 12, 2013 at 2:06 PM, Satish Balay <balay at mcs.anl.gov> wrote:
> > > >> On Thu, 12 Dec 2013, Lisandro Dalcin wrote:
> > > >>
> > > >>> On 12 December 2013 18:03, Todd Munson <tmunson at mcs.anl.gov> wrote:
> > > >>> >
> > > >>> > The idea is to distribute TAO with PETSc and treat it like any other set of PETSc solves.
> > > >>> > All the TaoLog (or TaoNewLog) macros should go away in my opinion and be replace with
> > > >>> > the correct PETSc equivalents.
> > > >>> >
> > > >>> > Jason?
> > > >>> >
> > > >>> > Todd.
> > > >>> >
> > > >>>
> > > >>> Not sure if you guys care about previous PETSc releases, but #define
> > > >>> TaoNew[Log] would ease supporting petsc-3.4 in the next TAO release.
> > > >>
> > > >> There won't be a next [standalone] tao release. It will be part of
> > > >> petsc-3.5.
> > > >>
> > > >> And I'm not sure if maintaining a separate tao_version.h makes sense.
> > > >>
> > > >> And there will be a minimal changing of adapting the makefiles to to
> > > >> this new mode [if one needs to maintain comatibility with
> > > >> both petsc-3.4+tao-2.2 and petsc-3.5/tao].
> > > >>
> > > >> You can check the current state of this migration at 'balay/tao-to-petsc' branch.
> > > >
> > > > That's awesome.  To confirm: I should just merge balay/tao-to-petsc
> > > > into the irving/hollow branch of petsc that I'm using to track the new
> > > > features I need for my simulator development?   I.e., is that branch
> > > > stable enough to track with repeated merges?
> > > 
> > > In order for me to use tao-to-petsc I need it to be compatible with
> > > master, so I fix it for the PetscNew changes.  The updated branch is
> > > 
> > >     irving/tao-to-petsc
> > > 
> > > Hopefully that can be merged into your tao-to-petsc branch.
> > 
> > I should probably rebase balay/tao-to-petsc over current master - and
> > then cherrypick this patch - and push it down the stack - into the
> > compile-fixes set of patches on this tree.
> 
> Ok - its rebased(to latest master)/cherry-picked(PetscLog change) and
> pushed now. If anyone has balay/tao-to-petsc checked out - you should
> do a 'git pull --rebase'

BTW: if you want 'irving/hollow' to track latest petsc+tao - you
should perhaps do the following: [i.e remove the unnecessary
master/merge commits in this branch - and rebase it off
balay/tao-to-petsc

git checkout balay/tao-to-petsc
git pull --rebase

git checkout irving/hollow-temp
git rebase --onto balay/tao-to-petsc master irving/hollow-temp
git log balay/tao-to-petsc..
git push -f

I see only 2 relavent commits in here - and [plenty of unneeded merges]

balay at asterix /home/balay/petsc (irving/hollow-temp)
$ git log --oneline balay/tao-to-petsc..
5e05b55 Mention cell number in another DMPlexComputeCellGeometry error
156ab2f snes ex12: Deduplicate SetupElement and friends

Satish