Closed mondracek closed 2 months ago
Tested on CH3Br, and FFPB. The resulting *LCPD*.png
files are virtually the same as with the previous implementation. They are not exactly bit-wise identical if different --Vrange
are used, but that is expected.
Continuous integration fails with test-package v 3.9 and higher. Probably nothing to do with the part of the code I have modified. @yakutovicha, you probably understand the best how this continuous integration is supposed to work and how to manage it. Can you give me a hint what may be going on?
Edit: Sorry, the problem was likely just that the main branch on which I have had based this was not up to date. I have force-pushed an amendment. Another edit: Tests still failing, even after re-basing on an up-to-date main branch.
What I have found about the continuous-integration tests:
ci.yml
to test only one version of python, the test passes.Run pytest
stage, the error message (at lines 31-33) is the following
/home/runner/work/_temp/6ab19913-178e-4b8f-92bf-dae56f3545bf.sh: line 1: 1858 Segmentation fault (core dumped) PPAFM_RECOMPILE=1 pytest tests examples -v --cov --cov-report json
examples/PTCDA_single/example_ptcda.py::example_ptcda_single
Error: Process completed with exit code 139.
test
branch which was an exact copy of the current main
branch.What happened as a side effect of me finding out and what I am going to do now:
ci.yml
, solely for the debugging purpose. I am going to force-push this branch again from my local repository now in order to remove this commits.@mondracek I merged the PR for fixing the CI tests. If you merge main into this branch, the tests should pass.
Sorry, I rebased and force-pushed instead of merging with main, which could be undesired in some cases and I am going to avoid it next time, but that's a side point here ... The final result should be the same. What is unfortunately the important point is that _The tests are still failing, even with only examples/PTCDA_Hartreedz2/ being tested.
@mondracek There is something odd going on with the tests. I simply reran them and they passed, so it seems not to be deterministic. If you are confident that this PR cannot be influencing this, then I would just go ahead with it now.
Okay. I will wait for one approving review before merging. Preferably from @aureliojgc , as he knows the details of the LCPD calculations, but if Aurelio does not have time for it, anyone who feels what I did makes sense can approve.
@mondracek Shouldn't this be merged or is there some problem?
I thought I should wait for @aureliojgc to finish the tests he mentions above but then I decided to just go ahead and merge it.
Resolves #272 and resolves #245 too.
Note that from now on, the files
LCPD.xsf
and*LCPD*.png
will be placed in aQ?.??K?.??/Amp?.??/
rather than directly in the working directory, because, in general, more LCPD results for different Q, k, A can be calculated in one run (consequence of implementing issue #272). No restriction on--Vrange
now (implementing issue #245), except that of course at least 3 different bias values are needed to evaluate LCPD . Regression of quadratic polynomial (for the Kelvin parabola) with the least-square-of-error criterion will be used when more than three values of bias are available. The bias range inppafm-plot-results --Vrange
still needs to be compatible with the range inppafm-relaxed-scan --Vrange
, of course.