Debugging non-reproducibility of ACCESS-NRI 0.3.1 -> 0.4.0#47
Debugging non-reproducibility of ACCESS-NRI 0.3.1 -> 0.4.0#47dougiesquire wants to merge 7 commits intomainfrom
Conversation
|
🚀 Attempted to deploy 🖥️
|
|
🚀 Attempted to deploy 🖥️
|
|
🚀 Attempted to deploy 🖥️
|
|
🚀 Attempted to deploy 🖥️
|
|
🚀 Attempted to deploy 🖥️
|
…versions used in 0.4.0 Unfortunately, the new versions of CMEPS/CDEPS require updating ESMF and the old versions don't work with the updated ESMF. This makes it very difficult to test just updating ESMF
|
🚀 Attempted to deploy 🖥️
|
|
🚀 Attempted to deploy 🖥️
|
|
🚀 Attempted to deploy 🖥️
|
|
Closing as this has served its purpose. |
This is a PR to debug and document where we lost reproducibility when we updated from 0.3.1 to 0.4.0. The context is that our repro tests had a bug and was effectively always reporting reproducible runs. We thus updated components from 0.3.1 to 0.4.0 under the impression that the updates were not answer-changing when in fact they are.
Companion PR to ACCESS-NRI/access-om3-configs#173. See also https://github.com/COSIMA/access-om3/issues/266
🚀 The latest prerelease
access-om3/pr47-11at 9382009 is here: #47 (comment) 🚀