[Debichem-devel] chemps2 version 1.7.1 (new release)

Michael Banck mbanck at debian.org
Mon Jul 25 16:30:02 UTC 2016


Hi,

sorry for the late reply...

On Wed, Jun 29, 2016 at 10:39:10AM +0200, Sebastian Wouters wrote:
> (1) There's a new upstream v1.7.1, which I've added to the alioth git repo
> http://anonscm.debian.org/cgit/debichem/packages/chemps2.git/commit/?id=7dc05ec3efb740a19d60805522486ceecc341359
> 
> (2) Regarding:
> 
> > That's unrelated to the 1.7 upload (thanks!), but I've noticed that
> > chemps2 Build-Depends on libatlas-base-dev, what is the rationale for
> > that?
> 
> > That means that psi4 is now being built with ATLAS as well and I had to
> > defend myself when being asked about Blas:
> > https://github.com/psi4/psi4/issues/391
> 
> > Unless chemps2 really requires ATLAS, I think it should build against
> > refblas, and then the users should pick their preferred BLAS/LAPACK
> > implementation via update-alternatives, what do you think?
> 
> With refblas, do you mean "libblas-dev" and "liblapack-dev"? I switched
> libatlas-base-dev to libblas-dev and liblapack-dev. If that is not what you
> intended, please feel free to adapt (and please explain to me what refblas
> is).

Yeah, exactly, thanks.

> (3) In the debian/libchemps2-2.symbols file, there are a couple of
> additions and three replacements:
> 
> - _ZN7CheMPS28DavidsonC1Eiiiddb at Base 1.7
> - _ZN7CheMPS28DavidsonC2Eiiiddb at Base 1.7
> + _ZN7CheMPS28DavidsonC1Eiiiddbc at Base 1.7.1
> + _ZN7CheMPS28DavidsonC2Eiiiddbc at Base 1.7.1
> - _ZNK7CheMPS26CASPT25solveEd at Base 1.7
> + _ZNK7CheMPS26CASPT25solveEdb at Base 1.7.1
> 
> Please note that these correspond to optional arguments (which have default
> values):
> 
> https://github.com/SebWouters/CheMPS2/blob/5991b9104276483d07979aff64affadbd95f533e/CheMPS2/include/chemps2/Davidson.h#L46
> https://github.com/SebWouters/CheMPS2/blob/5991b9104276483d07979aff64affadbd95f533e/CheMPS2/include/chemps2/CASPT2.h#L85
> 
> and therefore do not break anything.

Ok.


Michael



More information about the Debichem-devel mailing list