On 02/06/2021 22.22, Sebastian Ramacher wrote:
So, what if we use the free versions between 103 and 200 to transition
to a Debian-specific version? Attached is a patch that does that. It's
not an optimal solution as we would have a release with a
Debian-specific SOVERSIONs. For good reason, we usually try to avoid
those. So, it would be good for someone more familiar with the users of
hdf5 to check if that would be an issue in this specific case.
Can we be even more explicit about that and use a SOVERSION like "106deb" ?
That might be a bit tricky with libtool, though.
I think you should only remove Conflicts/Replaces on libhdf5*-103 and
leave all other there. (Have they been in buster already?)
You definitively want to have a
Breaks: libhdf5-foo-10x(-1)
in each libhdf5-foo-106 package as we don't want to mix the current
bullseye packages with your proposed packages
Wouldn't it be sufficient to rename only the packages where the
soversion is the same in buster and bullseye?
I'll try to make a table:
library buster bullseye
proposed soversion
libhdf5_serial.so.103 libhdf5-103 libhdf5-103-1
103deb or 106
libhdf5_serial_fortran.so.100 libhdf5-103 n/a
-
libhdf5_serial_fortran.so.102 n/a
libhdf5-fortran-102 keep
libhdf5_serial_hl.so.100 libhdf5-103 libhdf5-hl-100
100deb or 106
libhdf5_serialhl_fortran.so.100 libhdf5-103
libhdf5-hl-fortran-100 100deb or 106
libhdf5_cpp.so.103 libhdf5-cpp-103
libhdf5-cpp-103-1 103deb or 106
libhdf5_hl_cpp.so.100 libhdf5-cpp-103
libhdf5-hl-cpp-100 100deb or 106
libhdf5_mpich.so.103 libhdf5-mpich-103
libhdf5-mpich-103-1 103deb or 106
libhdf5_mpich_fortran.so.100 libhdf5-mpich-103 n/a
-
libhdf5_mpich_fortran.so.102 n/a
libhdf5-mpich-fortran-102 keep
libhdf5_mpich_hl.so.100 libhdf5-mpich-103
libhdf5-mpich-hl-100 100deb or 106
libhdf5_mpichhl_fortran.so.100 libhdf5-mpich-103
libhdf5-mpich-hl-cpp-100 100deb or 106
libhdf5_mpich_cpp.so.103 n/a
libhdf5-mpich-cpp-103-1 keep or 103
libhdf5_mpich_hl_cpp.so.100 n/a
libhdf5-mpich-hl-cpp-100 keep
libhdf5_openmpi.so.103 libhdf5-openmpi-103
libhdf5-openmpi-103-1 103deb or 106
libhdf5_openmpi_fortran.so.100 libhdf5-openmpi-103 n/a
-
libhdf5_openmpi_fortran.so.102 n/a
libhdf5-openmpi-fortran-102 keep
libhdf5_openmpi_hl.so.100 libhdf5-openmpi-103
libhdf5-openmpi-hl-100 100deb
libhdf5_openmpihl_fortran.so.100 libhdf5-openmpi-103
libhdf5-openmpi-hl-fortran-100 100deb
libhdf5_openmpi_cpp.so.103 n/a
libhdf5-openmpi-cpp-103-1 keep or 103
libhdf5_openmpi_hl_cpp.so.100 n/a
libhdf5-openmpi-hl-cpp-100 keep
Well, having different sonames for libhdf5_cpp.so and libhdf5_*mpi*_cpp.so
might be difficult and non-intuitive.
But at least you don't have to touch libhdf5*-fortran-102 ;-)
BUT: This would still only solve the co-installability mess of hdf5,
the gdal problem is still there.
Andreas
