Mihai Cara

Correct. But that means we cannot completely get rid of this memory: `H x W x 1` is the minimum.

Most likely https://github.com/spacetelescope/jwst/issues/6091 is similar to https://github.com/spacetelescope/jwst/issues/5131. I have not seen any evidence of this.

Regression tests: JWST: https://github.com/spacetelescope/RegressionTests/actions/runs/17073575486 Roman: https://github.com/spacetelescope/RegressionTests/actions/runs/17073538185

The code for `region.py` in `gwcs` has better accuracy than the code that was used in `jwst`. @nden reported two failures in regression tests when #8892 was merged: https://github.com/spacetelescope/jwst/pull/8892#issuecomment-2429469191 I...


Upon further analysis, the differences between those 7 pixels in this PR compared to `main` is due to small differences in computed relative sky values. This results in different (by...

In comments: > Therefore, if `x.flag == WCSSET`, there is no need to call `wcsset()` and we can return immediately. I really would like to know Mark's opinion on this....

BTW, the unit test that's failing, `test_get_coord_range_nan_regression`, is an example of a bad way of setting CRVAL.

context is exactly the same as in https://github.com/astropy/astropy/pull/16279#issuecomment-2043541877: "setter does not get called if an element of the array is modified in-place". This was brought up numerous times in the...

Probably if this PR is merged, something like this will not work: ```python w = WCS() w.wcs.ctype = ['RA---TAN', 'DEC--TAN'] w.wcs.ctype[1] = 'GLON--TAN' w.wcs.ctype[0] = 'GLAT--TAN' assert w.wcs.lattyp == 'GLAT'...