JIT: Assertion failed '!compIsForInlining()' during 'Morph - Inlining' in `jit64_2` tests

[amanasifkhalid]

Build Information

Build: https://dev.azure.com/dnceng-public/public/_build/results?buildId=945874&view=results
Build error leg or test failing:

Error Message

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

{
  "ErrorMessage": "Assertion failed '!compIsForInlining()'",
  "BuildRetry": false,
  "ExcludeConsoleLog": false
}

Configuration:

set DOTNET_TieredCompilation=0
set DOTNET_JitStress=1a7

Example log. cc @dotnet/jit-contrib

Known issue validation

Build: 🔎 https://dev.azure.com/dnceng-public/public/_build/results?buildId=945874
Error message validated: [Assertion failed '!compIsForInlining()']
Result validation: ✅ Known issue matched with the provided build.
Validation performed at: 2/10/2025 7:39:24 PM UTC

Report

Build	Definition	Test	Pull Request
945874	dotnet/runtime	jit64_2.WorkItemExecution

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.

[EgorBo]

Looks like the assert complains that a function with varargs was attempted to be inlined

[EgorBo]

Curious if it could be inliner-related changes in #111782 or #111948

[JulieLeeMSFT]

CC @hez2010 if he can take a quick look at it and guess what might have caused this failure.

[hez2010]

It seems to be unrelated to my change.
The assertion here is to assert any callee of a method that being inlined cannot be vararg, which seems to be incorrect.
Here we are inlining:

[000030] I-C-G------                         *  CALL      int    MCCTest.MyClass:Main():int (exactContextHandle=0x00007FFB4FC2AEC1)

which makes perfect sense, but then in the body of MCCTest.MyClass:Main we have a

               [000193] --CXG------                         *  CALL ind  struct
               [000192] ----------- calli tgt               \--*  FTN_ADDR  long   MCCTest.MyClass:Sum(double,int,long,float,short,double):MCCTest.VType3

which is vararg, so the assertion triggers while importing MCCTest.MyClass:Sum during inlining MCCTest.MyClass:Main.

I don't think we have any way to detect whether a method call inside an inlinee is vararg or not before actually import it, so this check seems wrong to me. Maybe we should simply remove this check and instead block inlining vararg calls.

[hez2010]

Here is the test binary: jit64_2.zip
To repro, set environment vairables

DOTNET_TieredCompilation=0
DOTNET_JitStress=1a7

and run jit64_2.dll with corerun.

[EgorBo]

@hez2010 was it just exposed by your changes (.e.g. we know inline what we did not) ?

[hez2010]

@hez2010 was it just exposed by your changes (.e.g. we know inline what we did not) ?

Yeah, it seems to be a new valid inlinee that we didn't inline before. The actual issue should already exist in the JIT before my change.

[AndyAyersMS]

I have a Jan 22 build lying around that also fails this way. Could just be random inline stress working its magic.