`FitManager` with uncertainties should not be "pre-declared"

[kif]

While trying to follow this tutorial:
https://www.silx.org/doc/silx/latest/Tutorials/fit.html#fitting-gaussians

I stumbled upon a bug, or at least a very unexpected behaviour:

from silx.math.fit.functions import sum_gauss
from silx.math.fit import fittheories
from silx.math.fit.fitmanager import FitManager
result = results["CdTe:Si"]
print(len(result.radial), len(result.intensity), len(result.sigma))
>>> 4000 4000 4000
fit = FitManager()
# fit.fitconfig["WeightFlag"] = True
fit.setdata(result.radial, result.intensity, sigmay=result.sigma, xmin=8)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[56], line 1
----> 1 fit.setdata(result.radial, result.intensity, sigmay=result.sigma, xmin=8)

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/fitmanager.py:792, in FitManager.setdata(self, x, y, sigmay, xmin, xmax)
    790         self.xdata = self.xdata[bool_array]
    791         self.ydata = self.ydata[bool_array]
--> 792         self.sigmay = self.sigmay[bool_array] if sigmay is not None else None
    794 self._finite_mask = numpy.logical_and(
    795     numpy.all(
    796         numpy.isfinite(self.xdata), axis=tuple(range(1, self.xdata.ndim))
    797     ),
    798     numpy.isfinite(self.ydata),
    799 )

TypeError: 'NoneType' object is not subscriptable

To me, setdata with some uncertainties should automatically set the flag "WeightFlag"

[kif]

Another bug a couple of lines later:

fit.runfit()

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[63], line 1
----> 1 fit.runfit()

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/fitmanager.py:879, in FitManager.runfit(self, callback)
    876 xwork = self.xdata[self._finite_mask]
    878 try:
--> 879     params, covariance_matrix, infodict = leastsq(
    880         self.fitfunction,  # bg + actual model function
    881         xwork,
    882         ywork,
    883         param_val,
    884         sigma=self.sigmay,
    885         constraints=param_constraints,
    886         model_deriv=self.theories[self.selectedtheory].derivative,
    887         full_output=True,
    888         left_derivative=True,
    889     )
    890 except LinAlgError:
    891     self.state = "Fit failed"

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/leastsq.py:373, in leastsq(model, xdata, ydata, p0, sigma, constraints, model_deriv, epsfcn, deltachi, full_output, check_finite, left_derivative, max_iter)
    361 """
    362 I cannot evaluate the initial chisq here because I do not know
    363 if some parameters are to be ignored, otherways I could do it as follows:
   (...)
    370 ignored or not between calls.
    371 """
    372 iteration_counter += 1
--> 373 chisq0, alpha0, beta, internal_output = chisq_alpha_beta(
    374     model,
    375     fittedpar,
    376     x,
    377     y,
    378     weight,
    379     constraints=constraints,
    380     model_deriv=model_deriv,
    381     epsfcn=epsfcn,
    382     left_derivative=left_derivative,
    383     last_evaluation=last_evaluation,
    384     full_output=True,
    385 )
    386 n_free = internal_output["n_free"]
    387 free_index = internal_output["free_index"]

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/leastsq.py:762, in chisq_alpha_beta(model, parameters, x, y, weight, constraints, model_deriv, epsfcn, left_derivative, last_evaluation, full_output)
    760     yfit = last_evaluation
    761 deltay = y - yfit
--> 762 help0 = weight * deltay
    763 for i in range(n_free):
    764     derivi = numpy.resize(deriv[i, :], (1, nr))

ValueError: operands could not be broadcast together with shapes (4000,) (3256,) 

Since the WeightFlag has to be set, this calls probably the enableweight method that resets the value of self.sigmay to its full length, not the trimmed one.

[kif]

And yet another:

fit = FitManager()
fit.setdata(result.radial, result.intensity, sigmay=result.sigma, xmin=8)
fit.loadtheories(fittheories)
fit.setbackground("Strip")
fit.settheory('Split Pseudo-Voigt 2')
fit.estimate()
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
Cell In[124], line 1
----> 1 fit.estimate()

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/fitmanager.py:441, in FitManager.estimate(self, callback)
    439 # estimate the function
    440 try:
--> 441     fun_params, fun_constraints = self.estimate_fun(xwork, ywork)
    442 except LinAlgError:
    443     self.state = "Estimate failed"

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/fitmanager.py:1008, in FitManager.estimate_fun(self, x, y)
   1006 if hasattr(estimatefunction, "__call__"):
   1007     if not self.theories[self.selectedtheory].pymca_legacy:
-> 1008         return estimatefunction(x, y)
   1009     else:
   1010         # legacy pymca estimate functions have a different signature
   1011         if self.fitconfig["fitbkg"] == "No Background":

File ~/.venv/py313/lib/python3.13/site-packages/silx/math/fit/fittheories.py:759, in FitTheories.estimate_splitpvoigt2(self, x, y)
    757 newcons[6 * i, 0] = cons[3 * i, 0]
    758 # position
--> 759 newcons[6 * i + 1, 0] = cons[3 * i + 1, 0]
    760 # fwhm1
    761 newcons[6 * i + 2, 0] = cons[3 * i + 2, 0]

IndexError: index 85 is out of bounds for axis 0 with size 85

Very likely that estimate_splitpvoigt2 is initializing array newcons with the wrong shape (5 instead of 6)