dbekaert / RAiDER / 1e5c40f1-c20a-4d00-81d0-bfc94c35b76c

55%

dev: 55%

Pull #822

circleci

regenerate processed ERA-5 fixtures with the half-level pressure fix

The three processed ERA-5 fixtures in test/weather_files still stored the
pre-#826 pressure: #816 recomputed only their derived wet/hydro cubes from
stored t/p/e, and test_intersect/test_slant passed only because
WeatherModel.load() returns early when a processed file already exists.

Regenerate them from raw model-level data with dev merged in, so the
pressure fix, the height-datum fix and #823's scratch-dir fixtures are all
present. The raws are the 2023 CDS downloads each fixture was cut from; the
LA one comes from 5f5d3807, since the tracked ERA-5_2020_01_30_T13_52_45.nc
was later replaced by a Mexico domain. Each was rewritten into the layout
ECMWF.fetch writes today (full model-level geopotential from calcgeoh, lnsp
on the first level) and processed onto exactly the fixture's existing grid.

Reprocessing the same inputs with the pre-#826 code leaves t bit-identical
and moves the goldens only 1.6-2.6 mm, so the shift is the pressure fix:
  test_gnss_intersect  2.332483 -> 2.377665  (+43.6 mm from #826)
  test_slant_proj      2.332479 -> 2.377760  (+43.2 mm)
  test_ray_tracing     2.975697 -> 3.033324  (+55.0 mm, slant)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

169 of 185 new or added lines in 7 files covered. (91.35%)

3645 of 6643 relevant lines covered (54.87%)

0.55 hits per line