Closed keflavich closed 6 years ago
Confirmed that the tests now fail. The problem was actually with beams specified with degree major/minor axes (e.g., as from a FITS header) and the comparisons to small numbers being done on numbers with unspecified units (could be deg^2 or arcsec^2 or steradians!)
(thanks to @e-koch for helping out with this)
What the hell. I thought we solved this. https://travis-ci.org/radio-astro-tools/radio-beam/jobs/336352673
hooray! The failures are now legit.
The tolerance for a certain difference-of-numbers was set to 1e-8 by default, which in deg^2 corresponds to 0.36 arcsec - i.e., a very usable beam size! This PR makes the tolerance smaller.