Closed lwellerastro closed 2 months ago
Based on some chats that transpired this week, I'm wondering if updates to geos is causing the problem:
conda list | egrep "isis|geos|ale"
# packages in environment at /anaconda3_linux/envs/isis8.2.0-RC1:
ale 0.10.0 py312h8572e83_1 conda-forge
geos 3.12.1 h59595ed_0 conda-forge
isis 8.2.0_RC1 0 usgs-astrogeology/label/RC
conda list | egrep "isis|geos|ale"
# packages in environment at /anaconda3_linux/envs/isis8.0.3:
ale 0.9.1 py39h7633fee_1 conda-forge
geos 3.11.1 h27087fc_0 conda-forge
isis 8.0.3 np126_0 usgs-astrogeology
geos version varies across isis versions tested, but PolygonTools.cpp (from caminfo failure message) uses geos so I thought it worth mentioning.
PR merged, re-open if fix doesn't work.
ISIS version(s) affected: isis8.2.0-RC1 This is a very recent bug. All other versions of isis work with the same input image and same command line arguments. This includes isis8.0.3 which I thought was the same as 8.2.0-RC1 based on the changelog, but apparently they differ.
The one difference I have noted is isis8.2.0-RC1 uses a newer version of ale whereas 8.0.3 and others uses an older/different version (leading path info stripped):
vs
So perhaps there is something about the latest ale helping to generate the error I'm seeing?
Description
Most Kaguya images that cross the 0/360 longitude boundary fail caminfo polygon=true (uselabel=true or when calculated on the fly) with the following error:
All of my images failed that have the 0/360 lon boundary and lat>+/-65 (polar quads) and about half of the images between lat 65S to 65N and 0/360 lon failed. All of these had previously been processed under isis5.0.2 and ran through caminfo without error.
How to reproduce
See my user work area Isis3Tests/Caminfo/isis8.2.0RC1/ for an example image. The image was processed in the following way under isis8.2.0-RC1 (shape model is under a work directory):
Successful caminfo output run on this via isis8.0.3 is in the same directory and named TC1W2B0_01_00849N757E0018.isis803.pvl
My Kaguya data are currently processed under 5.0.2, but I wanted to re-process everything under a more recent version of isis/ale prior to jigsaw updates and kernel creation. I chose the latest version because it picked up changes to camstats that haven't been available since isis7.0 or earlier which contains important information for a database which will be used for mosaic creation.
Possible Solution
I don't see anything in the changelog for isis8.2.0-RC1 or in the ale changelog that would explain the change.
Additional context
Only images having the 0/360 boundary fail (thousands), though there are several hundred that don't (have no idea what's special about them). The version of ale has changed and it is only under this version that the data fail.