[daip] more IMAGR problems

[Eric Greisen]

Lynn D. Matthews wrote:
> I've encountered another failure with IMAGR (messages appended below). I 
> had it running overnight with autoboxing and filtering turned on, plus a 
> flux limit set.
> 
> It got through 161 channels successfully, none of which had detectable 
> emission in them. It then choked on channel 162, the first channel with 
> emission present. The emission is fairly weak, but strong enough 
> that it should exceed the boxing threshold.
> 
> I did a couple of experiments. Setting FLUX=0 and turning off filtering 
> had no effect. But when I turned autoboxing off, and the failure no longer 
> occurs. With autoboxing on, I found that if I try to clean only one 
> channel, failures occur in the earlier channels of the cube where there 
> was no previous problem. Tweaking the IM2PARM parameters seemed to have no 
> effect.

I have done a couple of things that will help.  It should detect that 
nothing was found and even that there were no boxes and continue on 
normally.  Instead, OVERLAP=1 with ONEBEAM false, kept going.  In my 
default case, it did a gridded (non) subtraction which generated no 
error so it continued n in an infinite loop.  Since you asked for DFT, 
it went to VOSDFT instead and that routine takes exception to being 
asked to do nothing.  I am not sure what the right answer is on that 
point, but I am certain it should not be called in such cases.  I have 
put in a way to detect that Clean finds nothing left to do (or nothing 
at all because of no boxes) in all 3 ways through the code.  Tomorrow's 
MNJ should fix the problem.

Eric Greisen