Open ggardet opened 2 years ago
The hash fail has very likely the same root cause as the parsing test earlier, the Arc radius autoscaling.
(Failing hash test case is path = Path(line1, arc1, cub2, quad3)
).
As can be seen, the actual values for the autoscaled radius are the same, just with different accuracy depending on architecture and python version.
For the d()
method, the format string for each coorditate could be changed to {:16g}
, which rounds the value so the representation becomes less architecture dependent:
>>> "{:.16g}".format(27.950849718747367)
'27.95084971874737'
This of course only changes the parser test case, the hash test would still fail.
Another option would be to change the test cases so the arc radiuses are no longer autoscaled.
TestGeneration.test_path_parsing
andTestPath.test_hash
tests fail on aarch64 with version 1.4.4: