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

Re: libsundials-dev



Le dimanche 28 mai 2017 à 11:23 +0100, Ghislain Vaillant a écrit :
> On Sat, 2017-05-27 at 23:52 -0700, Dima Kogan wrote:
> > Science people:
> > 
> > I'm looking at the sundials packaging again. This is an umbrella
> > library
> > that has several different sub-libraries for
> > 
> > - different ODE solvers
> > - linear algebra backends used by these solvers
> > 
> > The granularity question came up earlier, and some of you pointed
> > me
> > to
> > a section in the policy (section 8.1) that indicates that putting
> > at
> > least each solver into its own separate package is the right thing
> > to
> > do.
> > 
> > What about the -dev package? I THINK putting everything into a
> > single
> > libsundials-dev is OK, and shouldn't break anything. Is the only
> > downside the unnecessary dependencies a user would need to install
> > (for
> > instance a user that only wants serial computations would need to
> > install all the parallel libraries)?
> > 
> > How frowned-upon is this? A joint libsundials-dev would simplify
> > some
> > things, and make stuff less error-prone.
> > 
> > I can do a finer split, but want to make sure this is worth the
> > trouble
> > first.
> 
> Please correct me if I am wrong, but I thought that each library
> composing sundials had separate (so)versioning? If so, then you'd
> have
> no choice but ship them in separate -dev packages, right?

Since -dev packages do not contain the SOVERSION in their name, there
is no problem having only one -dev package (see suitesparse which is in
a similar situation, with one -dev package for several dynamic
libraries with different SOVERSIONs).

The only benefit of splitting the -dev package is to save some disk
space on developers' machines who don't need all the libraries. IMHO it
is not worth the trouble (in comparison with the extra maintenance work
and the overhead on Debian mirrors).

-- 
 .''`.    Sébastien Villemot
: :' :    Debian Developer
`. `'     http://sebastien.villemot.name
  `-      GPG Key: 4096R/381A7594


Reply to: