Re: Scotch upgrade [Was: Re: MUMPS and other imminent upgrades]
On Mon, 2017-06-12 at 10:31 +0100, Ghislain Vaillant wrote:
> On 12/06/17 10:11, Drew Parsons wrote:
> >
> > Thanks Ghislain, that'd be great if you can take on scotch.
>
> Please keep in mind that I cannot commit to a super speedy ETA with
> the
> holidays coming :-)
No great rush, it'll be a couple of years before the next stable
release...
> > > I'll give it another go. Do we agree that we only want to support
> > > the
> > > latest version (6.0.4) from Buster onwards, and drop the 5.x
> > > series
> > > currently in the archive?
> >
> > That's sensible I think. I don't think there's much to be gained
> > by
> > keeping the older version around. Scotch's clients (PETSc, FENiCS,
> > etc,
> > not sure about trilinos) are well maintained upstream so they
> > should
> > have no trouble building against the new scotch.
>
> Ok. Just to be clear on the plan. Christophe is in the process of
> sending his work on 6.0.4 to me. I'll make sure the update is good
> policy-wise, push to the packaging repository, and submit to
> experimental (or unstable?).
Yes, good plan. Best to put it in the experimental branch first I
think.
I wonder whether this should be listed in Debian Science policy as a
general policy?: new major updates (especially transitions with a
soname change) should always be pushed first to experimental.
Otherwise you risk the danger where one component of the library gets
into unstable and then testing, while another component is held back by
some failure. That might result in the library not working at all, for
a while.
Experimental also gives the opportunity to test the builds (and
runtime) of client programs before releasing to unstable.
Drew
Reply to: