[CI] ScaledFloatFieldMapperTests#testSyntheticSourceMany failing about elasticsearch HOT 3 OPEN

@lkts wanna take a look at hardening ScaledFloatFieldMapper#decodeForSyntheticSource and updating the test as described above?

from elasticsearch.

Pinging @elastic/es-storage-engine (Team:StorageEngine)

from elasticsearch.

The reason that the assertion trips is that in ScaledFloatFieldMapper#decodeForSyntheticSource(...) method scaledValue / scalingFactor expression results into negative infinity. Then when v += Math.ulp(v) is executed, v variable becomes NaN. The assertion fails, because it can't recompute the scaledValue (which is -77).

In this case the scalingFactor is 4.275522857276975E-307. Using such tiny scaling factors are likely going to produce infinity as a result. In the context of synthetic source, synthesizing the original value or an equivalent isn't possible. Either the test should be adjust to use values that are Math.floor(value) != 0.0d or the ScaledFloatFieldMapper#decodeForSyntheticSource(...) should be adjust to not attempt adjusting the value if scaledValue / scalingFactor results into infinity.

from elasticsearch.