[petsc-dev] petsc/petsc-dev someone changed DM to __PETSc_dm for the name of a DM attached to a Mat object but didn't change its use everywhere (commit 6e9d4e780b4d4f1b06b023485bbcf786a79fb6fe)

[Jed Brown]

Matt said he was afraid of your wrath because we discussed doing that a
long time ago and you were opposed. I'd rather have it in the Vec and Mat
data structures, if nothing else than because it removes the string lookups.

Note that the slots in Vec and Mat should be a PetscObject because their
destructor needs to be called. Also note that we don't automatically check
those slots for circular references so we need to think about the
destruction process to figure out if we still need to prevent that loop and
if so, how to check.

On Tue, Nov 20, 2012 at 10:47 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:

>
> On Nov 20, 2012, at 2:30 PM, Matt Knepley <
> notifications-noreply at bitbucket.org> wrote:
>
> > New comment on changeset:
> >
> >
> https://bitbucket.org/petsc/petsc-dev/changeset/6e9d4e780b4d4f1b06b023485bbcf786a79fb6fe#comment-70978
> >
> > Matt Knepley (knepley) said:
> >
> > What the heck are those calls doing in examples? The user should NEVER
> NEVER NEVER be doing that. There are accessors for that.
>
>    There is an accessor for doing
>
>   ierr =
> PetscObjectCompose((PetscObject)J,"__PETSc_dm",(PetscObject)PETSC_NULL);CHKERRQ(ierr);
>
> ?
>
>    BTW: if you are going to write accessors like MatSetDM() then why go
> through the PetscObjectCompose() nonsense? Why not just stick the dm into
> the mat data structure as a void * in your accessor?
>
>
>
>
> >
> > --
> > This is an comment notification from bitbucket.org.
> > You are receiving this because you are participating
> > in a comment thread.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20121120/d3806d85/attachment.html>