[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Skip VTK7 and go directly VTK8? Was: Re: vtk6 and vtk7



However, for the name xxxx-vtk we have packages in oldstable
> that refers to the vtk5 python and tcl bindings.

Ah right, I faintly remember that too. Are those ever going to go away?

Technically yes, but if someone wants to use some software that
> requires python2-vtk6 and another one that requires python3-vtkX, the
> conflict would mean that they could not install both software packages
> at the same time, so we should try to avoid this situation.

Indeed. It may be a pick-your-poison situation: Either don't have any Python 3 interfaces, or get conflicting packages. (The egomaniac that I am I'd vote for the latter.) Anyhow, we can try and consolidate those two. Not sure if possible.

Gert, are you already done with your iteration on the package? If yes, and the coinstallability is the only remaining issue, we could start working at this in the vtk8 repo.

Cheers,
Nico

On Tue, Jul 11, 2017 at 6:03 PM Gert Wollny <gw.fossdev@gmail.com> wrote:
Am Dienstag, den 11.07.2017, 14:51 +0000 schrieb Nico Schlömer:
> Thanks Gert for the explanation.
>
> Do I understand correctly that we keep vtk6 alive for backwards
> compatibility of older packages?
Well, that was my proposal (I also didn't expect that VTK8 would come
so soon, after just one point release and one bugfix release of VTK7).

> If the idea for the new vtk{7,8,...} releases is to just keep them
> coming, we could pick up an earlier idea and call the new package
> simply `vtk`. This would make skipping 7 less weird, too.
I think the idea was to make virtual packages, and yes, I fully support
that idea. However, for the name xxxx-vtk we have packages in oldstable
that refers to the vtk5 python and tcl bindings.

>
> > Then, there are quite a few packages that depend on python-vtk6,
> and currently there are some files in python-vtk* that are not co-
> > installable for two different versions of vtk,
>
> Can't we just make them conflict?
Technically yes, but if someone wants to use some software that
requires python2-vtk6 and another one that requires python3-vtkX, the
conflict would mean that they could not install both software packages
at the same time, so we should try to avoid this situation.

> > For that reason I wanted to get rid of the vtk6 python bindings,
> and go forward with VTK7, and in order to not break the reverse 
> > dependencies, python2 was the choice, since it is currently the
> default version of python in Debian.
>
> I see. My idea was to keep python2 for vtk6 around, and python3
> bindings for the newer releases such that developers can choose.
That sounds reasonable.

Best,
Gert

>
> Cheers,
> Nico
>
> On Tue, Jul 11, 2017 at 4:29 PM Gert Wollny <gw.fossdev@gmail.com>
> wrote:
> > Am Dienstag, den 11.07.2017, 13:28 +0000 schrieb Nico Schlömer:
> > > > I'd say we skip VTK7 altogether, and go directly for VTK8.
> > >
> > > Don't know if this is a good idea. I know that at least FEniCS
> > > only supports VTK6 and 7, and I think not many libraries have
> > > upgraded their code for 8. (After all, it's a breaking change.)
> >
> > The major problem is with the VTK-OpenGL and OpenGL2 rendering
> > interfaces that can not be compiled in a the same time. The latter
> > requires OpenGL 3.2 and mesa doesn't support compatibility
> > profiles, - but I know that at least ginkgocadx and itksnap both
> > require an OpenGL compatibility profile, this means we will have to
> > keep two versions of VTK alive - at least until everything is
> > ported - but some packages might be dead upstream (ginkgocadx for
> > instance is kind of dead, and the changes required to make it
> > compatible with an OpenGL core profile are big, but it is of rather
> > high interest in debian-med). 
> >
> > Given that currently all packages that require VTK are at vtk6,
> > IMHO it is the best to keep vtk6 for the time being, and jump right
> > to the new version, which is VTK8 for packages that make use of
> > VTK-OpenGL2. And since I don't think that we want three versions of
> > VTK in buster, my suggestion would be to skip vtk7. In the end,
> > buster is still two years away, and if we plan now to get as many
> > packages as possible to VTK8, then the extra work for a transition
> > to VTK7 would be wasted.
> >
> > > I just noticed that in your new code, my Python 3 efforts aren't
> > > included. This is what motivated me to work on VTK7, and it'd be
> > > a pity if it didn't get in.
> >
> > I understand that, but I started again from vtk6 porting all the
> > patches and first wanted it to build like it was with VTK6. 
> > Then, there are quite a few packages that depend on python-vtk6,
> > and currently there are some files in python-vtk* that are not co-
> > installable for two different versions of vtk, even if one uses
> > python2 and the other one python3 (Though I have to admit that I
> > don't know whether your changes took care of this). For that reason
> > I wanted to get rid of the vtk6 python bindings, and go forward
> > with VTK7, and in order to not break the reverse dependencies,
> > python2 was the choice, since it is currently the default version
> > of python in Debian.
> >
> > Now, if we can get for instance python-vtk6 and python3-vtk8 co-
> > installable, I'd say we go for that.
> >
> > On the other hand, if you feel like we really should have a
> > VTK7+Python3 package, then just change the package names and
> > dependencies accordingly, and I will upload the vtk7 package.
> >
> > best, 
> > Geert
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > > Cheers,
> > > Nico
> > >
> > > On Tue, Jul 11, 2017 at 3:08 PM Gert Wollny <gw.fossdev@gmail.com
> > > > wrote:
> > > > > > I'm sorry, but the last discussion about this package was
> > > > > about two month ago
> > > > > > and nothing had happened past the version that was
> > > > > commented
> > > > >
> > > > > That's right. The discussion was moved from the mailing list
> > > > > to a private exchange between Ghislain and myself, the reason
> > > > > being (I thought) not to bother anyone with the technical
> > > > > details. I guess that was a mistake.
> > > >
> > > > The package is supposed to be group maintained, for that reason
> > > > I think it is always a good idea to keep discussions on the
> > > > list, and debian-science is not exactly high traffic.
> > > >
> > > > > Before overriding the the content I would have expected an
> > > > > email, though.
> > > >
> > > > Yes this was wrong from my side, but I did a pull before and
> > > > didn't see any change compared to May, 
> > > > that's why I thought you had abandoned the effort.
> > > >
> > > > > Anyways. Let's just take it from here and see if we can get
> > > > > VTK7 out. After all, VTK8 has already been released as well.
> > > >
> > > > VTK7 would have to remain in experimental for now, because it
> > > > requires gl2ps > 1.4 currently only available from exp, and no
> > > > package depends on anything vtk7, so
> > > > I'd say we skip VTK7 altogether, and go directly for VTK8. I
> > > > would guess that most of the patches for vtk7 also apply to
> > > > vtk8, maybe the new gl2ps patch is no longer needed, because it
> > > > might have been applied upstream.
> > > >
> > > > I've created an empty vtk8 repo as 
> > > > https://anonscm.debian.org/git/debian-science/packages/vtk8.git
> > > >
> > > > If you feel like it you could also rename the ITP #810254 for
> > > > vtk8 and take ownership.
> > > >
> > > > Sorry again,
> > > > Gert
> > > >
> > > >

Reply to: