try quick fix on #1136

[dehann]

Might need backports

[codecov-io]

Codecov Report

Merging #1137 (35a3d0c) into master (19afa75) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #1137   +/-   ##
=======================================
  Coverage   73.20%   73.20%           
=======================================
  Files          55       55           
  Lines        4673     4673           
=======================================
  Hits         3421     3421           
  Misses       1252     1252           

Impacted Files	Coverage Δ
src/CliqStateMachineUtils.jl	81.25% <100.00%> (ø)
src/GraphProductOperations.jl	94.11% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 19afa75...35a3d0c. Read the comment docs.

[Affie]

Looks good, it needs something similar for multiprocessor.
What bothers me a bit is that fixing this will hide another potential error.

[dehann]

needs something similar for multiprocessor

I'd have to look for it, or please add if you have easy recall on where those lines of code are.

another potential error

oops, which error would that be?

[Affie]

oops, which error would that be?

See #1136 (comment)
There were no factors on the specific frontal.

[Affie]

I'd have to look for it, or please add if you have easy recall on where those lines of code are.

I added as I understand, please review.

[Affie]

I added same on multiproc

[Affie]

ping @dehann

[dehann]

hi @Affie , we going back and forth. I think we can merge although I am not sure about the concerns you raised. It looks like you cleared them with the later commit. If you are happy, please go ahead and merge this PR.

[Affie]

The concerns are about an underlying issue that causes the point to be zero. (that should not happen in this case)
There was a warning a while back that I think is related. I don't know if it was taken out or redirected to the logger, I'll try and find it. The problem is to find MWE that shows this. Give me some time on this one as I don't want to potentially hide an issue.

[dehann]

okay, understood thanks!

[Affie]

Ok, I kind of remember (70%) the waning being from here:

IncrementalInference.jl/src/GraphProductOperations.jl

Line 81 in 4bad101

@warn "Unknown density product on variable=$(vert.label), lennonp=$(lennonp), lenpart=$(lenpart)"

[dehann]

error #1136 was caused by keyword arguments in AMP.manifoldProduct not having enough content for broadcasting. A few alternate PRs should have fixed the issue. The attempted fix in this PR is not valid, and would introduce an unwanted error.