You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As found in the investigation #790, when doing a permanent simple update model.update() with a batch dataset (batch size > 1), there is a bug that only happens on Mac and on certain architectures that made the feedstock CI crash (not our own).
To Reproduce
A detailed explanation with investigation can be found in #790, but to simply reproduce, make a branch from main and revert 795b7e9, then build and run using the same configuration on Mac as the failed azure pipelines.
After #790, it was found that permanent updates with batch datasets were not properly tested, and the necesary tests were added in the core in the same PR, showing that the core indeeds support it. Additionally, the root of this issue maybe related or similar to #769.
The text was updated successfully, but these errors were encountered:
mgovers
changed the title
[BUG] Simple update with batch datasets in Mac yields error
[BUG] Permanent update with batch datasets in Mac yields error
Oct 21, 2024
FYI: renamed Simple update to Permanent update because it carries the meaning better. This is not a big problem, just a minor thing. "Simple update" may appear more severe to a user than it actually is.
Describe the bug
As found in the investigation #790, when doing a permanent simple update
model.update()
with a batch dataset (batch size > 1), there is a bug that only happens on Mac and on certain architectures that made the feedstock CI crash (not our own).To Reproduce
A detailed explanation with investigation can be found in #790, but to simply reproduce, make a branch from main and revert 795b7e9, then build and run using the same configuration on Mac as the failed azure pipelines.
Expected behavior
Feedstock CI should not break as this feature is properly tested and passes in our CI.
Additional context
After #790, it was found that permanent updates with batch datasets were not properly tested, and the necesary tests were added in the core in the same PR, showing that the core indeeds support it. Additionally, the root of this issue maybe related or similar to #769.
The text was updated successfully, but these errors were encountered: