Non identical extractions

[profxj]

If I perform the following test:

1. Full run of PypeIt from scratch on good ole shane_kast_blue, 600_4310_d55
2. Move Science/ to Sci_1
3. Repeat the extraction without remaking the calibrations; move that Science/ to Sci_2
4. Do it a 3rd time for good charm

I find the 1D extractions are all different and approaching 1% in a few cases.

A "simple" comparison to show are the FWHM saved to the spec1d files:

In [1]: ex1_spec1d_full.FWHM
Out[1]: array([71.65610607,  5.15995966, 24.19016042])

In [2]: ex2_spec1d_full.FWHM
Out[2]: array([71.80109709,  5.15968347, 11.29537328])

In [3]: ex3_spec1d_full.FWHM
Out[3]: array([72.06389084,  5.15963796, 22.18462915])

These differ at ~0.1% to even 100% when the S/N is low (1st and 3rd cases)
but less at higher S/N. These outputs are for the standard star exposure and the standard star itself is the 2nd FWHM value here.
To the best of my ability, I have confirmed the inputs into extraction are identical to 1e-8 precision

So, I think it is something within the extract.
There is plenty going on in there, e.g. bspline

[profxj]

This has been linked to the C-code for the bspline.

Current mitigation is to avoid that code and use the Python version of the bspline.
I believe the fix is as "easy" as removing the "compiler...gnu.so" file in pypeit/

But there will be a better solution/fix soon.

[tbowers7]

Now that the C bspline code has been axed, it would be helpful for someone to rigorously retest this again and close the issue if solved.