Closed bendnorman closed 1 year ago
The build that finished on 9/28 passed despite running the code from 5264d2664ca664b397793d369cf4a2dec1de2c0a-dev
, the previous built that failed :/ This makes me think the PyGeos failure could maybe be an API issue? I can't think of any flaky APIs that are a part of the census processing though. Any ideas @zaneselvans?
I don't think there are any APIs involved. We've just got the one 2010 census DB, which should be the same as it's always been, and that's where the geometries are coming from. If something has changed with the geometries it's got to be in the Python GIS stack: GEOS, pygeos, shapely, fiona, geopandas. I don't think there's any random numbers in there either. Or you know, it could be cosmic rays! Not satisfying.
We decided to punt this issue unless it becomes a recurring problem right? No idea why it's not consistent.
Yup, we've decided to punt.
The
81d10c1dff2ae7da83593f70d694d694531a427e-dev
nightly build passed but the subsequent build5264d2664ca664b397793d369cf4a2dec1de2c0a-dev
failed with a single error:These were the PRs that were merged between the two builds. At first glance, these PRs shouldn't have changed anything with this ferc714 test.
Debugging: