Confirmed by maintainers¶
While physverify was tested on public physics code, each suspected defect was checked by a run against a reference from the physics and reported to the project. These are reports where the project's maintainers confirmed the defect or merged a fix (checked 2026-11-14). The link in each row is the source; the one-line summaries are ours.
| Project | What was found | Outcome |
|---|---|---|
| SymPy | units: the astronomical unit, nautical mile and torr differ from their exact definitions |
confirmed by a core developer against primary sources |
| Astropy | the "surface tension" physical type is registered as J m² instead of J/m² | confirmed by a maintainer; fix in progress |
| PyGMT | sequence_join treats any string containing a space as a datetime |
fixed (PR #4961) |
| Pynite | BeamSegY.max_moment ignores P_delta |
confirmed by the author; fix planned |
| EinsteinPy | DeSitter() puts the scale factor on a spatial coordinate |
fixed (PR #707) |
| findiff | set_accuracy is ignored after the first call; 3 - Diff(0) has the wrong sign (#120) |
both fixed |
| TESPy | the pressure term of the incompressible-fluid entropy has the wrong units | fixed |
| openmmtools | the harmonic-oscillator test system reports the wrong standard deviation of its potential | confirmed by a developer |
| moltemplate | msifrc2lt.py: an undefined name in the out-of-plane branch |
confirmed by the author |
| PyAstronomy | pyasl.planck used CGS constants where SI were expected; the maintainer found the cause was wider: the two constants modules (with and without the quantities package) defaulted to different unit systems |
confirmed and fixed |
| qmat | three Runge-Kutta schemes labelled with a higher order than they have | fixed; qmat now checks its tables with physverify.butcher in CI |
Smaller projects whose authors fixed a reported defect include freepaths, ugropy, phlower, spin-model-transformers, finitevolume2-python, mhd-boris and StatMech_IsingModel.
This list shows that the defects were real; it is not a measure of how often physverify finds bugs. Most were found by a test script written for each repository, with physverify running the checks and the sandbox (see Results on real code).