Test NEDwaves with replayed problem data (+ synthetic data?)

[lauralindzey]

Based on how the 3 floats all got stuck within NEDwaves_memlight, Phil thinks the issue is likely to be that some inputs cause it to run out of memory. Combined with our discovery that the code does not consistently check the that the return value of calloc_replacement is not null before using it, this could lead to stack corruption, which would match some of the symptoms we saw in stepping through the assembly.

There are two steps that would make sense for confirming this + hopefully point to better fixes:

1. Try to reproduce this using the logged data from floats 190, 195, 196 (last file before Nov 2)
2. Write a fuzzer to find other inputs that cause issues

[lauralindzey]

#1 has been done. Reposting from Slack:

I've managed to get my microswift stuck using input data from the microswifts that have been stuck at sea. I think this is as good a confirmation as we're going to get that the problem is input data, and I'm hoping that being able to get them stuck on command will help speed up our ability to test potential fixes.

To use this for testing, you'll need to cherry-pick commits from the gnss_test_data branch, then remove them when you're done testing.

• git cherry-pick f892927 # This adds the test data, no changes to functionality
• git cherry-pick 8620fc7 # This will freeze NEDwaves in the normal loop; this is probably what you'll want to use for testing fixes to unfreezing the code.
• git cherry-pick b31bc71 # Adding this commit will have NEDwaves run immediately and freeze, without having to wait for the GNSS sample period to end -- this was what I used to test which inputs freeze the process, since I didn't want to wait 17 minutes each time.

I took the GNSS logs from the at-sea swifts, and generated a header file defining arrays for the data. I pushed that script to:
https://github.com/lauralindzey/microswift_test_gnss
If you want to add additional test files:

• copy a microswift GNSS log file in with an appropriate name
• rerun python3 python3 generate_test_include.py
• and copy the resulting data_include.h to NEDwaves/test_data_includes.h
• (optional: send me a PR to add them to the repo!)

To test different inputs, change the #define at the top of the NEDwaves/test_data_includes.h file and re-compile to change the desired test case.

So far, I've tested with:

• MICRO190_GOOD_110125_220000 -- happy NEDwaves thread
• MICRO190_STUCK_110225_060000 -- landed in FPU_IRQHandler
• MICRO192_GOOD_110225_140000 -- happy NEDwaves thread
• MICRO192_STUCK_110225_230000 -- landed in FPU_IRQHandler
• MICRO195_STUCK_110225_061400 -- landed in FPU_IRQHandler