JIT: Assertion failed '!"Inconsistent profile data"' during 'Importation'

[amanasifkhalid]

Build Information

Build: https://dev.azure.com/dnceng-public/public/_build/results?buildId=1004148&view=results
Build error leg or test failing:

Error Message

Fill the error message using step by step known issues guidance.

{
  "ErrorMessage": "Inconsistent profile data",
  "BuildRetry": false,
  "ExcludeConsoleLog": false
}

Introduced by #114016. cc @dotnet/jit-contrib

Known issue validation

Build: 🔎 https://dev.azure.com/dnceng-public/public/_build/results?buildId=1004148
Error message validated: [Inconsistent profile data]
Result validation: ✅ Known issue matched with the provided build.
Validation performed at: 4/3/2025 5:53:50 PM UTC

Report

Build	Definition	Test	Pull Request
1004148	dotnet/runtime	Methodical_do.WorkItemExecution	#113977

Summary

24-Hour Hit Count	7-Day Hit Count	1-Month Count
0	1	1

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[amanasifkhalid]

I haven't been able to reproduce this locally, or with the build from #113977. Hopefully the next run will be more enlightening.

[amanasifkhalid]

This still hasn't hit in any subsequent CI runs. Removing blocking tag.

[amanasifkhalid]

I finally got a repro with a Debug build of the runtime. The default epsilon for checking if two weights are equal is large enough that we fail to detect changes in small weights, so we end up with consistency issues during flow propagation through call-finallies like so:

Checking Profile Weights (flags:0x1c)
  BB71 - block weight 0.00012 inconsistent with incoming likely weight 1e-05
  BB75 - block weight 0.00013 inconsistent with incoming likely weight 1e-05
  BB79 - block weight 0.00014 inconsistent with incoming likely weight 1e-05
  BB83 - block weight 0.00015 inconsistent with incoming likely weight 1e-05
  BB87 - block weight 0.00016 inconsistent with incoming likely weight 1e-05
  BB91 - block weight 0.00017 inconsistent with incoming likely weight 1e-05
  BB95 - block weight 0.00018 inconsistent with incoming likely weight 1e-05
  BB99 - block weight 0.00019 inconsistent with incoming likely weight 1e-05
  BB102 - block weight 0.0002 inconsistent with incoming likely weight 1e-05

Using fgProfileWeightsConsistentOrSmall seems to work better.