v0.30.1-rc1

[dehann]

Hi @Affie , did you want to add anything to v0.30.1, or can the next PRs go to v0.30.2?

[dehann]

I am worried about the regression

• "Basic 3 door, trinary multihypothesis tests" often fails numerically #1570

So will look into that bit later today if can.

[dehann]

Caesar tests also seem to be failing more often on IIF master. I get the sense that some of the recent changes may have upset the multihypo code with silent bug somewhere.

[Affie]

It can go into the next, but master broke from something upstream (i think)

[codecov]

Codecov Report

Merging #1572 (5a736ae) into release/v0.30 (d8c8257) will increase coverage by 0.32%.
The diff coverage is 88.00%.

@@                Coverage Diff                @@
##           release/v0.30    #1572      +/-   ##
=================================================
+ Coverage          77.34%   77.67%   +0.32%     
=================================================
  Files                 71       71              
  Lines               4988     5093     +105     
=================================================
+ Hits                3858     3956      +98     
- Misses              1130     1137       +7     

Impacted Files	Coverage Δ
src/Factors/LinearRelative.jl	82.35% <ø> (-0.99%)	⬇️
src/IncrementalInference.jl	87.50% <ø> (ø)
src/ManifoldsExtentions.jl	68.00% <ø> (+24.00%)	⬆️
src/entities/FactorOperationalMemory.jl	100.00% <ø> (ø)
src/entities/JunctionTreeTypes.jl	48.14% <0.00%> (ø)
src/entities/SolverParams.jl	100.00% <ø> (ø)
src/services/GraphInit.jl	91.79% <ø> (ø)
src/ParametricUtils.jl	66.66% <87.73%> (+6.66%)	⬆️
src/Factors/GenericFunctions.jl	59.72% <100.00%> (ø)
src/ManifoldSampling.jl	88.88% <100.00%> (-0.59%)	⬇️
... and 10 more

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[Affie]

Technically parametric has a bug in with the return types, but if you are in a hurry for something you can tag.

[dehann]

Looks like it has something to do with getSolverParams(fg).useMsgLikelihoods = true. MWE:

using IncrementalInference

fg = generateGraph_CaesarRing1D()
getSolverParams(fg).useMsgLikelihoods = true

solveGraph!(fg)

Code completes fine if useMsgLikelihoods=false

[dehann]

useMsgLikelihoods uses approxDeconv, and I wonder if all that code got updated for Tangents vs Points in the new ArrayPartition work.