[Debichem-devel] Elpa-2015.11 interface change

[Andreas Marek]

Am Sonntag, 22. November 2015, 14:22:41 schrieb Michael Banck:
Dear Michael,

I will have a look at this again, but to my understanding the changing in 
versioning has been consistent with
http://www.gnu.org/software/libtool/manual/html_node/Updating-version-info.html

But maybe we were wrong in some way. I will check this.

Best regards, Andreas


> Dear Elpa developers,
> 
> I had a look at the new Elpa-2015.11 release, and noticed (via
> discussion in the CP2K development team) that there have been interface
> changes since 2015.03, at least this one (from src/elpa2.F90):
> 
>   public :: solve_evp_real_2stage
> [...]
>  function solve_evp_real_2stage(na, nev, a, lda, ev, q, ldq, nblk,        &
> +                               matrixCols,                               &
>                                   mpi_comm_rows, mpi_comm_cols,           &
>                                   mpi_comm_all, THIS_REAL_ELPA_KERNEL_API,&
>                                   useQR) result(success)
> 
> The symbolic library changed its (internal library) version from 3.0.1
> to 3.0.2, while I think it should be revved to 4.0.0 due to (at least)
> the above change.  That would change the SONAME of the library from
> libelpa.so.3 to libelpa.so.4, which is not the case:
> 
> $ objdump -p elpa-2015.11.001/.libs/libelpa.so.3.0.2 | grep SONAME
>   SONAME               libelpa.so.3
> $ objdump -p elpa-2015.05.001/.libs/libelpa.so.3.0.1 | grep SONAME
>   SONAME               libelpa.so.3
> 
> Adding an additional function like solve_evp_real_2stage_2 with the
> additional parameter would not have been an issue for the SONAME (it
> would have increment the library version from 3.0.1 to 3.1.0), but
> changing or removing a function signature usually needs an API bump.
> 
> Was this discussed prior to release, and/or can there be anything done
> to rectify this?  Or is this not a concern for some reason?
> 
> 
> Best regards,
> 
> Michael