On 2022-01-05 13:37, Neil Williams wrote:
> On Fri, 17 Dec 2021 12:57:48 +0200 Andrius Merkys <mer...@debian.org>
> wrote:
>> I see you have switched from having all libraries in cctbx/
>> subdirectory [1] to public libs and libdevel locations. Personally I
>> do think Debian as a downstream should be setting SONAMEs. This does
>> not seem to be forbidden by the policy, but in case the upstream
>> introduces SONAMEs, clashes may occur. Moreover, caring for ABI
>> compatibility is a lot of work.
> 
> When trying to package libobjcryst, it became obvious that a SONAME had
> to be applied to cctbx, just as libobjcryst itself needs to patch in a
> SONAME. It is an extra amount of work but C++ symbols based on a
> package using lots of templates are fuzzy at best and with so many
> different modules in cctbx, the only practical way to handle it seems
> to be a new SONAME each time. The header files are also problematic. We
> might be able to restrict SONAME changes to upstream versions which
> make changes to the header files included in libcctbx-dev rather than
> every new upstream release. Until we've got cctbx through NEW, it is
> going to be hard to tell.

Understood. So it boils down to uploading to NEW with each and every
upstream release. But I guess we will have to live with that as there
are no better alternatives.

>> On a separate thread, I managed to package reduce [2] locally.
>> However, as you have also noted it, the name of source, binary and
>> executable is problematic. Maybe it is worth trying to talk to
>> upstream about renaming it to avoid clashes.
> 
> Maybe package it as pdb-reduce or pdb-hydrogen-reduce ?
> 
> /usr/bin/reduce does not exist in Debian yet but it may be worth
> packaging the script as pdb-reduce with a note in README.Debian -
> anyone switching from using the upstream build to Debian packages will
> need to make other changes anyway.

pdb-hydrogen-reduce sounds good to me.

>> [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=679905#98
>> [2] https://github.com/rlabduke/reduce

Best,
Andrius

Reply via email to