Conversation
rettigl
left a comment
rettigl
left a comment
There was a problem hiding this comment.
Did you test that this actually does work and does the right thing when tests fail in the different fashions? Passing tests here do not execute the changed code...
See #695 (comment) for an example where the number of log lines are different (case 1 above) In addition, I also modified one line in another pynxtools-xps test (to trigger case 2) and I get the result that we had seen so far. We get the same result as we got before: So I think that covers both use cases. |
RubelMozumder
left a comment
RubelMozumder
left a comment
There was a problem hiding this comment.
Yeah, not that much difference, at least experiments are saying that.
Let's go with it and see how it affects us.
|
@RubelMozumder @rettigl thanks for the feedback. Just FYI, I slightly changed the logging functionality to match that of the case where only some lines are different. But that does not affect the functionality here. Merging now. |
|
There are still 1000s of log lines here: https://github.com/FAIRmat-NFDI/pynxtools/actions/runs/16939887832/job/48005872843 |
Okay, it's a lot less than before, but still a lot coming from the Nomad parsing. Maybe one disables the logging propagation there as well? |
I was thinking about that, but then the NOMAD parsing is essentially silent. At least now the number of log lines is small enough that the GitHub UI can display the actual error (previously it was often just loading and loading). I am fine with suppressing the NOMAD logs or maybe we only keep those that are at least a warning? |
That sounds like a good idea. Right now, one would not even find those because of all the output. |
3 participants
Currently, the whole
read_nexusoutput gets logged to the CLI when pytest is run. In the GitHub Actions interface, this clutters the output and it can be fairly hard to find the actual issues if the tests fail.Here, we use
logger.propagate = Falseso that the log is stored in the files, but not in the CLI output.Additionally, changes to the actual output of the log comparison is made. So far, if the line count mismatched, we would not get any info which lines are different. Now, we log those lines that are unique in both files, making it easier to understand how the two logs are different.