-
Notifications
You must be signed in to change notification settings - Fork 125
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Test: make CI test fails with warnings #4493
Conversation
While updating the thresholds for 58 warnings, I encountered a serious bug in the current Autotest workflow. The current mechanism for determining warnings and fatal errors is flawed. For instance, let's take an example that has three indices: energy, force, and stress. The correct logic should be: if none of the three indices is a fatal error, and there is at least one warning, then it should be classified as a warning; however, if any one of the indices is a fatal error, then the whole test should be classified as a fatal error. But the current logic is: as soon as any index is deemed as a warning, the workflow exits immediately without checking the subsequent indices, even though they may be completely wrong. In reality, we actually have 21 force indices with fatal errors hidden behind a warning for the total energy index:
|
It turns out that the failure of stress comparison has been a long-standing problem muted in the past. I traced all the way back to #3675 (version update to v3.5.4) and the comparison already failed back then. To nail down the specific PR that introduces the problem, I think we need to resort to developers that are familiar with the recent changes in the stress calculation in PW. I think the PR that first exposed this problem was #4496 : #4496 was merged on July 1. Four other PRs were merged that day afterwards, but their checks were done prior to the merge of #4496 , therefore no warnings from the stress test were displayed in their respective workflows. |
The List of Changes
and 3 warnings for GPU test:
Please Notice that