[daip] FIXBX bug?

[Eric Greisen]

Andy Biggs wrote:
> Hi Amy. Sorry to bother you again, but I may have found a bug in FIXBX. 
> Firstly, I should say that this is a wonderful task that's made my life 
> a lot easier. However, after changing my faceting scheme (moving to a 
> larger cellsize and therefore covering a larger area of the sky) and 
> using FIXBX to re-set my clean boxes, I've found that certain fields in 
> my map are not being cleaned by IMAGR, despite them containing sources.
> 
> On closer examination it turns out that the output boxfile from FIXBX 
> contained clean boxes for these fields which didn't correspond to any 
> sources in the field. I've done some tests that show that the problem is 
> probably occurring when FIXBX comes across a situation where, after 
> mapping the clean boxes to the new faceting scheme, a field does not 
> contain any 'new' clean boxes. If that field number has 'old' clean 
> boxes from the previous faceting scheme then it leaves them in place. 
> These naturally don't correspond to any emission in the new field 
> resulting in IMAGR never trying to clean that field even if it contains 
> bright sources. FIXBX should, of course, either write out no boxes for 
> that field or write out one as large as the field itself. At least, I 
> think this is what is going on. Otherwise, FIXBX is working perfectly.
> 
> In some ways, this problem mimics that in CCEDT where the old clean 
> components are left in place if the output file would contain none.

What FIXBX does at present is read in the boxes in BOXFIL.  It reads in 
the old boxes in INFILE.  If a facet in BOXFIL receives some boxes from 
INFIL then all old boxes are erased.  If no boxes go to a facet then the 
old boxes specified in BOXFIL are retained.  This was really based on 
the assumption that BOXFIL would contain no boxes or only the default 
SETFC output circular inscribed boxes.  Thus the "old" boxes that are 
retained are from the new faceting scheme not the old one since BOXFIL 
should contain the new faceting scheme only.

Upon reflection, this seems inconsistent and boxes in BOXFIL should 
probably be completely ignored although there really should be no boxes 
other than those written by SETFC.  FIXBX would then add a default 
circle to otherwise empty facets.

If you really believe that FIXBX is doing something else (as you suggest 
above), please send me the 3 box files involved.

Eric Greisen