GCI_marquardt_fitting_engine incorrectly shifts input when no instrument response provided

[marktsuchida]

flimlib/src/main/c/EcfSingle.c

Lines 1893 to 1898 in 40b1fb8

	if (ninstr <= 0) {
	trans = &(trans[fit_start]); // we can change the pointer since it has been passed 'by value'
	fit_end -= fit_start;
	ndata -= fit_start;
	fit_start = 0;
	}

These lines don't work. The fitted parameters end up being for the data shifted to the left by fit_start, and no correction is performed. The amplitude of the fit ends up being scaled down as a result.

The workaround of providing an ideal instrument response of length 1, containing 1.0, appears to work.

One question is why the GCI_marquardt (as opposed to GCI_marquardt_instr) function is not used in the case of no instrument response, despite being present.

[marktsuchida]

(It turns out that this problem was already pointed out in the commit that introduced it.)