pmSubtractBias won't return a psSpline1D

[Paul Price]

I tried calling pmSubtractBias with and without pre-allocating a psSpline1D as
the fitSpec, but couldn't get it to return a non-NULL fitSpec.

This might be related to the alterations made to the psSpline1D about a month ago.

[gusciora@…]

This should be done.

[Paul Price]

What was wrong? How was the bug fixed?

[gusciora@…]

If pmSubtractBias() is called with fitSpec being an already allocated spline
then I wrote a routine which explictly copies the newly generated spline in
pmSubtractBias() to the fitSpec argument. I verified this works in the current
CVS tree, however if you check that out, there are still some memory leak problems.

Speaking of the memory stuff, when I create an entire FPA->chip->cell->readout
hierarchy, should I be able to free the whole thing with a simple psFree(FPA)?
Or must I free some of the sub trees as well?

[Paul Price]

Freeing the FPA should free all the dependent elements.

Shouldn't the fix to pmSubtractBias be as simple as:

pmSubtractBias(..., void fitSpec, pmFit fit, ...)
{

....
psSpline1D *spline = NULL; Spline to use in fitting overscan
if (fit == PM_FIT_SPLINE) {

if (fitSpec && *fitSpec) {

We're given a spline; use that.
spline = *fitSpec;

} else if (fitSpec && ! *fitSpec) {

We're given a pointer to NULL; create a spline to pass back
spline = psSpline1DAlloc();
*fitSpec = spline;

} else {

Told to use a spline, but the user doesn't want it back.
...

}

}
....

}