From discussions on slack, it seems that inter-chip crosstalk is still only possible in the two-step process; Step 1 processes the entire focal plane through overscan to create crosstalkSources, Step 2 then runs the standard ISR processing, but will now find the crosstalkSources.

For testing, I used the same toy repository created for DM-23985.

 pipetask qgraph  -d "exposure IN (292404)" -b ./gen3 \

          -i DECam/raw/all,DECam/calib \

         -o czw_ct_passA -p ./IsrCt.yaml --qgraph-dot ./pipeline_partA.dot

pipetask run -j 4 -d "exposure IN (292404)" -b ./gen3 \

         -i DECam/raw/all,DECam/calib \

         -o czw_ct_passA -p ./IsrCt.yaml --register-dataset-types

pipetask qgraph  -d "detector=1 AND exposure IN (292404)" -b ./gen3 \

          -i czw_ct_passA \

         -o czw_ct_passB -p ./IsrCtB.yaml --qgraph-dot ./pipeline_partB.dot

pipetask run -j 4 -d "exposure IN (292404)" -b ./gen3 \

         -i czw_ct_passA \

         -o czw_ct_passB -p ./IsrCtB.yaml --register-dataset-types

I've attached the pipeline qgraph files to this ticket (they are large and wide).

This line of text here to make it clear that these are two separate pipeline qgraph files, which isn't initially obvious.

Sorry for the long delay on this review! I deeply appreciate you doing this, as it will help a lot as AP continues adopting gen3.

Two main things. First, I don't see any changes to crosstalk tests on this ticket, so please convince me they're not needed or add some. Second, I appreciate the example above for how to run this, but this workflow isn't reflected clearly in the obs_decam/pipelines directory.

One bigger meta-question, which I think I know the answer to... Are two pipetask run calls absolutely needed? I fear this will throw a wrench in ap_pipe, which runs from raws through difference imaging.

Finally, since it's been a bit (sorry!!) I'd appreciate one final "integration test" in gen2 and gen3 to make sure this works as anticipated. Happy to help with this; I have repos you are welcome to borrow at /project/mrawls/hits2015 (gen2) and /project/mrawls/hits2015-3 (gen3).

The code changes look great! The pipeline situation is much clearer now, thanks.

Please make sure all tests pass when testdata_decam is setup, and as I said before, that you do a small DECam isrTask "integration test" in both gen2 and gen3. I'll be happy to mark this as reviewed when you confirm all that works! I'm willing to pair-test it with you, if you'd like.

I've attached the list of test commands run (DM-27833.txt), and that should cover all the integration tests I could think of.  Briefly:

• confirm obs_decam builds with testdata_decam setup
• Run }}{{ap_verify on both ticket and w10, and confirm they yield pixel-identical outputs.
• Create fresh gen2 repo, ingest the crosstalk calibs, and confirm it runs cleanly, and that the outputs have crosstalk features removed.
• Do the same in gen3, running the two step processing.

Outstanding! Thanks for testing all that, and for your hard work implementing this.