Jigsaw "Unable to set coefficients vector" issue with 6.0.0

[volcanopele]

ISIS version(s) affected: 6.0.0

Description
Running jigsaw on CaSSIS images (could impact other spacecraft images but I haven't tested that) results in the following set of errors and bundle adjustment does not occur:

Group = Error
    Program = jigsaw
    Class   = "USER ERROR"
    Code    = 2
    Message = "Unable to bundle adjust network
               [MY36_016125_127_0_cnetextract_net.net]"
    File    = jigsaw.cpp
    Line    = 196
  End_Group

  Group = Error
    Program = jigsaw
    Code    = 1
    Message = "Could not solve bundle adjust"
    File    = BundleAdjust.cpp
    Line    = 1013
  End_Group

  Group = Error
    Program = jigsaw
    Code    = 1
    Message = "Unable to apply parameter corrections to IsisBundleObservation"
    File    = IsisBundleObservation.cpp
    Line    = 509
  End_Group

  Group = Error
    Program = jigsaw
    Class   = "PROGRAMMER ERROR"
    Code    = 3
    Message = "Unable to set coefficients vector. The size of the given vector
               [1] does not match number of coefficients in the basis equation
               [3]"
    File    = BasisFunction.cpp
    Line    = 48
  End_Group

Running jigsaw using the same observation with 5.0.2, using the same code shown below, does not produce this error and bundle adjustment proceeds correctly. So this seems to be a bug that has cropped up in 6.0.0.

Both 5.0.2 and 6.0.0 are installed on my MacBook Pro running macOS 12. Each installation has their own "base" data areas but they should be identical since they were pull via RSync within a day of each other. They use the same tgo data area. My CaSSIS scripts successfully outputs products with ISIS 6.0.0, and they look reasonable, just with color fringing due to the lack of bundle adjustment.

The output of the preceding steps in the jigsaw step of my code look identical between 5.0.2 and 6.0.

How to reproduce
The jigsaw step in my CaSSIS images processing script looks like this:

footprintinit from=\$1 -batchlist=../${IMAGE_NAME}_image.lis
autoseed fromlist=../${IMAGE_NAME}_image.lis overlaplist=${IMAGE_NAME}_overlaps.txt deffile=$AUTOSEEDDEF \
    onet=${IMAGE_NAME}_autoseed_strip.net errors=${IMAGE_NAME}_autoseed_grid.err networkid=$IMAGE_NAME.1 \
    pointid=$IMAGE_NAME.\?\?\?\?\?\?\? description="$IMAGE_NAME grid1"
pointreg fromlist=../${IMAGE_NAME}_image.lis cnet=${IMAGE_NAME}_autoseed_strip.net \
    onet=${IMAGE_NAME}_pointreg_net.net deffile=$AUTOREGDEF MEASURES=CANDIDATES \
    flatfile=${IMAGE_NAME}_pointreg_net.txt
cnetcheck fromlist=../${IMAGE_NAME}_image.lis cnet=${IMAGE_NAME}_pointreg_net.net \
    prefix=${IMAGE_NAME}_chk_
(some code to separate framelets with enough good points after pointreg from those that don't)
cnetextract fromlist=../${IMAGE_NAME}_image.lis cnet=${IMAGE_NAME}_pointreg_net.net prefix=new \
    tolist=${IMAGE_NAME}_newimage.lis onet=${IMAGE_NAME}_cnetextract_net.net \
    networkid=$IMAGE_NAME.new description="$IMAGE_NAME cnetextract" \
    cubelist=${IMAGE_NAME}_goodimages.txt cubes=yes cubemeasures=yes
jigsaw fromlist=${IMAGE_NAME}_newimage.lis cnet=${IMAGE_NAME}_cnetextract_net.net \
    onet=${IMAGE_NAME}_jigsaw_net.net model1=huber file_prefix=${IMAGE_NAME}.js \
    sigma0=1.0e-3 maxits=200 update=yes OBSERVATIONS=yes camera_angles_sigma=0.02

Here is how spiceinit is used, in case that helps

spiceinit from=\$1.cub attach=true spkpredict=true ckpredict=true -batchlist=../${IMAGE_NAME}_image_name.lis

Possible Solution

Additional context
Sorry to bringing this up so late. The CaSSIS team is now working to move from ISIS 4.4 to a more recent version of ISIS due to some label improvements. I have used 5.0.2 on my local system for the last few months with great success and started testing 6.0.0 this week. This testing was performed using my bash shell CaSSIS processing script. Testing with the production pipeline is on going.

[jessemapel]

Looks like a bug in observation mode. When we append a new image to an ISIS observation we need to make sure that we properly initialize its SpicePosition and SpiceRotation members.

[volcanopele]

thanks for taking a look at this bug. It looks like we will have to use 6.0.0 or greater for our reprocessing project due to recent CaSSIS label updates in tgocassisrdrgen showing up on 6.0.0.

[jessemapel]

Can you let us know what type of images you had in your processing list? was this a full CaSSIS observation?

[volcanopele]

We are processing full CaSSIS observations so the images that are going into jigsaw are the framelets that will, in later steps in the pipeline, be mosaicked together into color swaths.

[jessemapel]

If someone else works this, CaSSIS data can be pulled from the ESA site:

https://archives.esac.esa.int/psa/#!Table%20View/CaSSIS=instrument

[jessemapel]

Copying over here because I realize I sent this in a private chat:

Look at the code in BundleObservationVector and BundleObservation before #4537

Specifically, BundleObservationVector::addNew and where BundleObservation::initializeExteriorOrientation are being called

[jessemapel]

This has also been encountered with JunoCam imagery, which is very similar to CaSSIS.

https://astrodiscuss.usgs.gov/t/should-junocam2isis-fullccd-yes-produced-single-band-cubes/797