[petsc-users] MatMPIBAIJSetPreallocationCSR i and j indices

[Barry Smith]

On Sep 29, 2013, at 9:31 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> Jed Brown <jedbrown at mcs.anl.gov> writes:
> 
>> Barry Smith <bsmith at mcs.anl.gov> writes:
>> 
>>>   I am fine with this. In fact I would go ahead and make the change
>>>   in the branch I already made exactly for this work. Sadly, git
>>>   allowed me to fuck up the branch so I cannot easily work on that
>>>   branch since it empty, if only git could have warned me I was
>>>   fucking up…..
>> 
>> You can still create a new branch and we can merge later so that it
>> looks like your current commit had been in the branch all along,
>> provided we merge it to 'maint' before merging something else there.
> 
> On second thought, I think this should be reverted in 'maint' because it
> would break existing code that calls the *SetPreallocationCSR functions.
> Interface "improvements" belong in 'master' where they can be documented
> in release notes and people know that some things might change.

   It actually isn't an improvement, it is a bug fix; the current code was broken (though I am not sure if it was always broken). But I am fine with removing my "fix".
> 
> I'm preparing a branch that respects MAT_ROW_ORIENTED and keeps
> MatSeqBAIJSetPreallocationCSR consistent.
   
  That's fine.