Failed recon after recompiling UM in ACCESS-AM3

Atmosphere AM3
,
1

Hi Team. I tried to recompile UM while running ACCESS-AM3 using this repository: /home/563/qg8515/roses/am3-compile. I made minimal changes to this repository after created from a working template (SPACK_BUILD='', TASK_BUILD_UM=true, um_sources, and jules_sources). The UM and JULES repositories track the latest branches AM3-dev (e.g. https://github.com/ACCESS-NRI/UM.git). The model output can be found here: /g/data/gx60/experiments/2026-03-04_am3-control/cylc-run/am3-compile/.

The jobs (fcm_make_um get_jules_source get_um_source install_ainitial install_ancil) succeeded, but recon failed with the following error message:

**????????????????????????????????????????????????????????????????????????????????**

**???!!!???!!!???!!!???!!!???!!!**       ERROR        **???!!!???!!!???!!!???!!!???!!!**

**?**  Error code: 1

**?**  Error from routine: CHECK_IOSTAT

**?**  Error message:

**?**        Error reading namelist cable_nml

**?**        IoMsg: end-of-file during read, unit 11, file **/g/data/gx60/experiments/2026-03-04_am3-control/cylc-run/am3-compile/work/20070101T0000Z/recon/SHARED,** line 2217, position 0

**?**        Please check input list against code.

**?**  Error from processor: 0

**?**  Error number: 0

**????????????????????????????????????????????????????????????????????????????????**

Could you please let me know how to resolve or circumvent this issue? I am trying to recompile UM because my research involves modifying model source code.

2

@qinggangg A quick reply before the end of the day, there might be another solution tomorrow but I thought I’d get you something now.

Generally, if you need to change a source code, the best practice is to start your own branch (so not AM3-dev) and to stem it from the latest released version (so not the HEAD of AM3-dev). Otherwise, you could pick up a state that is temporarily not working because a model change has been put in and the config change is just about to be merged but not yet. Or you could pick up a bug that hasn’t been discovered yet during development. So do not do research based of the HEAD of the AM3-dev branch.

The latest release of AM3 was the alpha release, it is using UM version 2025.08.0 and JULES version 2025.08.0.

I would then advice you start by creating your own branch in the UM and JULES, in this way (change newbranch to the name you like):

git checkout -b newbranch 2025.08.0

Then, use the released configurations and your build and simulations should be successful.

3

Hi @clairecarouge

Thank you very much. The problem is solved after checking out the correct version 2025.08.0. The recon succeeded and it proceeds to atmos_main now. I’ll report back if any issue occurs with atmos_main.

Regards, Qinggang

4

@qinggangg We try and get the heads of the different models and config branches to work together but it’s not always possible (as I said, timing when cloning can make it incompatible). Even so, we are going to check and make sure the configs and model components are in sync and nothing is broken between the models and the configs.

5

Thank you, @clairecarouge. The effort is really appreciated. This simulation has run successfully for two years. So this issue is resolved for me now.

6

Just for my sanity, I’ve checked that the model works using the current heads of the models and config branches. But as said previously, it is better to add any research from a release state instead of the heads of the dev-* branches.