#1095955 openmm: please package v8.2.0

#1095955#5
Date:
2025-02-14 09:53:04 UTC
From:
To:
Hello,

I am filing this bug to document the advantages, issues and progress of
updating openmm to v8.2.0.

Advantages
==========

python-pdbfixer/1.11 requires openmm/8.2.0. This blocks #1093240.

Issues
======

openmm v8.2.0 stops versioning its shared libraries. I wrote about this
on debichem-devel@ a while ago [1]. TL;DR: for openmm/8.2.0 we will have
to switch to providing static libraries unless a solution to having
shared libraries versioned is found; otherwise 3 reverse dependencies
will have to be adapted to link statically. What is more, static library
building has been reported to FTBFS [2].

Decision for trixie
===================

I would better leave packaging openmm/8.2.0 for post-trixie. Good thing
is that openmm/8.1.2 seems to be immune to python3.13-add and numpy2
transitions. However, time until freeze is scarce to resolve the
aforementioned issues.

[1]
https://alioth-lists.debian.net/pipermail/debichem-devel/2024-April/016396.html
[2] https://github.com/openmm/openmm/issues/4406

Andrius

#1095955#12
Date:
2025-08-11 16:27:43 UTC
From:
To:
One option is to just add a debian patch that reverts upstream PR#4498.

mdtraj would use openmm (>= 8.2) if it were available, but can work
around the problem by just ignoring openmm.

#1095955#17
Date:
2025-09-01 10:07:15 UTC
From:
To:
Hi,

Yes, that is an option. However, tampering with soversions against
upstream wishes is generally frowned upon.

I too would like to use openmm (>= 8.2). I am leaning towards fixing
myself the upstream issue blocking the static builds.

Best wishes,
Andrius