Crash in cryo_project

[janden]

Specifically, running it with double precision seems to crash it for certain NUFFT codes:

f = load(fullfile('projections', 'simulation', 'maps', 'cleanrib.mat'));
cryo_project(f.volref, eye(3), size(f.volref, 1), 'double');

says the imaginary component is too large. Here, it's about 13 times larger than precision and since imagtol is 5*precision, it fails.

What is the reason for having such a strict check on the imaginary components? When would this happen? We end up taking the real part afterwards anyhow.

[yoelsh]

This was supposed to be a safeguard against bugs.
I have simply never tested it in double precision...
In single precision and lower I never had problem.
I guess it should be fine even if we lose 2 digits, that is 100*precision.

[janden]

Ok. Will make this a bit looser then.