Mule and Shumlib 2025.10.1 Release

[james-bruten-mo]

Dear All,

We're pleased to announce the release of mule and shumlib 2025.10.1 versions.
These represent the first release of these codes from github. Currently these are in private repositories within the Met Office organisation, however we plan to make these public shortly.

Some additional mule changes include,

• Compatibility with numpy2.0. This has also been backported to the 2024.11.1 version.
• Option to raise an error code when mule-cumf detects file differences.

These are now deployed on Met Office platforms. I will look to make mule 2025.10.1 the default version late next week. Please let us know if you find any issues.

[james-bruten-mo]

This is now the default mule for metoffice platforms

[pdearnshaw]

Do I take it that this verison of mule can only be used with a loaded scitools module? It appears not to work if I try to load without it saying it will not work with system python.

[james-bruten-mo]

Hi Paul, yes unfortunately the system python and numpy are already pretty old, so some of the recent changes include newer numpy functionality.

[paul-barrett]

this is breaking the regional nesting suite (RNS). When will the default python be upgraded to allow use of mule without a prior module load scitools? Does the suite need modifying or is it going to be a short-ish wait?

[james-bruten-mo]

Hi Paul, the suite will need updating. I'd suggest using scitools/production-os47-2. I'd also suggest using the explicit version of um_tools as then these changes won't immediately affect you, eg. um_tools/2025.10.1/openmp.

[paul-barrett]

so it's currently calling um_tools/2020.01.01/openmp but complaining it can't find mule-convpp

[james-bruten-mo]

The versions on the Quads don't go back that far - I guess it's been defaulting to using the whatever is the default instead of raising an error. That's not very helpful so hopefully there's a way for us to change that behaviour. But for now I'd update to one of the newer versions - I suspect there's no reason you can't use 2025.10.1.

[paul-barrett]

I think it was defaulting to that yes, but also it's calling scitools production-os46-3, so just trying to un-hardwire that