[mono] FullAOT build asserts at `class1 != ARG_CLASS_NO_CLASS' for EmptyStructs.dll

[matouskozak]

Build Information

Build: https://dev.azure.com/dnceng-public/public/_build/results?buildId=763399
Build error leg or test failing: AOT compile CoreCLR tests
Afffected CI: linux-x64 Release AllSubsets_Mono_MiniFullAot_RuntimeTests minifullaot (runtime-extra-platforms), linux-x64 Release AllSubsets_Mono_LLVMFULLAOT_RuntimeTests llvmfullaot (runtime-llvm)
Range of commits: 03b4c79...1337553

Crash dump

2024-08-01T22:17:44.4437150Z   Mono Ahead of Time compiler - compiling assembly /__w/1/s/artifacts/tests/coreclr/linux.x64.Release/JIT/Directed/StructABI/EmptyStructs/EmptyStructs.dll
2024-08-01T22:17:44.4437819Z   AOTID E87160E2-BE64-3040-7216-8519D54F4C66
2024-08-01T22:17:44.4438379Z   * Assertion at /__w/1/s/src/mono/mono/mini/mini-amd64.c:806, condition `class1 != ARG_CLASS_NO_CLASS' not met
2024-08-01T22:17:44.4438707Z   
2024-08-01T22:17:44.4438904Z   
2024-08-01T22:17:44.4439160Z   =================================================================
2024-08-01T22:17:44.4439382Z   	Native Crash Reporting
2024-08-01T22:17:44.4439612Z   =================================================================
2024-08-01T22:17:44.4439878Z   Got a SIGABRT while executing native code. This usually indicates
2024-08-01T22:17:44.4503811Z   a fatal error in the mono runtime or one of the native libraries 
2024-08-01T22:17:44.4504078Z   used by your application.
2024-08-01T22:17:44.4507375Z   =================================================================
2024-08-01T22:17:44.4507639Z   
2024-08-01T22:17:44.4507841Z   =================================================================
2024-08-01T22:17:44.4508094Z   	External Debugger Dump:
2024-08-01T22:17:44.4508310Z   =================================================================
2024-08-01T22:17:44.4508528Z   
2024-08-01T22:17:44.4508730Z   =================================================================
2024-08-01T22:17:44.4508995Z   	Native stacktrace:
2024-08-01T22:17:44.4510175Z   =================================================================
2024-08-01T22:17:44.4515383Z   	0x77209e3f97df - Unknown
2024-08-01T22:17:44.4515852Z   	0x77209e39f1be - Unknown
2024-08-01T22:17:44.4516289Z   	0x77209e3f9091 - Unknown
2024-08-01T22:17:44.4516991Z   	0x77209e746c90 - Unknown
2024-08-01T22:17:44.4517301Z   	0x77209e7970d4 - Unknown
2024-08-01T22:17:44.4517605Z   	0x77209e746bde - Unknown
2024-08-01T22:17:44.4517854Z   	0x77209e72f832 - Unknown
2024-08-01T22:17:44.4518019Z   	0x77209e49d904 - Unknown
2024-08-01T22:17:44.4518400Z   	0x77209e4ad256 - Unknown
2024-08-01T22:17:44.4518545Z   	0x77209e49dd6f - Unknown
2024-08-01T22:17:44.4518692Z   	0x77209e49ded5 - Unknown
2024-08-01T22:17:44.4518833Z   	0x77209e49df17 - Unknown
2024-08-01T22:17:44.4518980Z   	0x77209e3f376f - Unknown
2024-08-01T22:17:44.4519131Z   	0x77209e3d77f7 - Unknown
2024-08-01T22:17:44.4519275Z   	0x77209e3da791 - Unknown
2024-08-01T22:17:44.4519429Z   	0x77209e37b9fa - Unknown
2024-08-01T22:17:44.4519586Z   	0x77209e36f7e6 - Unknown
2024-08-01T22:17:44.4519737Z   	0x77209e36b59f - Unknown
2024-08-01T22:17:44.4519894Z   	0x77209e36182f - Unknown
2024-08-01T22:17:44.4520032Z   	0x77209e360378 - Unknown
2024-08-01T22:17:44.4520164Z   	0x77209e3d68a7 - Unknown
2024-08-01T22:17:44.4520293Z   	0x5d103e91b6d6 - Unknown
2024-08-01T22:17:44.4520417Z   	0x77209e730efb - Unknown
2024-08-01T22:17:44.4520551Z   	0x77209e730fbb - Unknown
2024-08-01T22:17:44.4520677Z   	0x5d103e9187e9 - Unknown
2024-08-01T22:17:44.4520769Z   
2024-08-01T22:17:44.4520885Z   =================================================================
2024-08-01T22:17:44.4521014Z   	Basic Fault Address Reporting
2024-08-01T22:17:44.4521146Z   =================================================================
2024-08-01T22:17:44.4521348Z   Memory around native instruction pointer (0x77209e7970d4):0x77209e7970c4  05 00 44 89 f2 89 de 89 c7 b8 ea 00 00 00 0f 05  ..D.............
2024-08-01T22:17:44.4521550Z   0x77209e7970d4  89 c3 f7 db 3d 00 f0 ff ff b8 00 00 00 00 0f 47  ....=..........G
2024-08-01T22:17:44.4521712Z   0x77209e7970e4  c3 48 83 c4 18 5b 41 5c 41 5d 41 5e 41 5f 5d c3  .H...[A\A]A^A_].
2024-08-01T22:17:44.4521859Z   0x77209e7970f4  0f 1f 40 00 4c 89 ef e8 90 ae ff ff e9 55 ff ff  ..@.L........U..
2024-08-01T22:17:44.6457478Z /__w/1/s/src/mono/msbuild/aot-compile.proj(19,9): error MSB3073: The command "/__w/1/s/artifacts/tests/coreclr/linux.x64.Release/Tests/Core_Root/corerun /__w/1/s/artifacts/tests/coreclr/linux.x64.Release/JIT/Directed/StructABI/EmptyStructs/EmptyStructs.dll" exited with code 134.

Error Message

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

{
  "ErrorMessage": "artifacts/tests/coreclr/linux.x64.Release/JIT/Directed/StructABI/EmptyStructs/EmptyStructs.dll\" exited with code 134.",
  "BuildRetry": false,
  "ExcludeConsoleLog": false
}

Known issue validation

Build: 🔎 https://dev.azure.com/dnceng-public/public/_build/results?buildId=763399
Error message validated: [artifacts/tests/coreclr/linux.x64.Release/JIT/Directed/StructABI/EmptyStructs/EmptyStructs.dll" exited with code 134.]
Result validation: ✅ Known issue matched with the provided build.
Validation performed at: 8/7/2024 11:37:07 AM UTC

Report

Summary

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

[matouskozak]

I think the failures got triggered by #104237 @tomeksowi .

It looks like Mono doesn't support this scenario:

runtime/src/mono/mono/mini/mini-amd64.c

Lines 804 to 806 in cb65d11

	/* Empty structs have a nonzero size, causing this assert to be hit */
	if (sig->pinvoke)
	g_assert (class1 != ARG_CLASS_NO_CLASS);

Do you know what changes would be required to make this work for Mono?

cc: @steveisok

[tomeksowi]

I think the failures got triggered by #104237 @tomeksowi .

Sorry about breaking the build. #104237 JIT changes should be confined to RISC-V and LoongArch.

Do you know what changes would be required to make this work for Mono?

I'm not familiar with Mono but If I understand the code correctly quad stands for what System V ABI calls eightbyte. If so, it's the problem described in #104098. I made ActiveIssues to disable on System V the test cases in JIT/Directed/StructABI/EmptyStructs which have NO_CLASS eightbytes (the ones with standalone Empty arguments or Empty8Float) so it shouldn't get there. Is the problem Mono fails already when it sees a function signature with a NO_CLASS eightbyte? If so, I think only conditional compilation of these test cases would help.

[tomeksowi]

One more thing, in VM we simply reclassify NO_CLASS eightbytes to INTEGER, see https://github.com/dotnet/runtime/pull/103799/files#diff-8b4d839010e3c7e363eb430d3efed40b0f9a01a7661e16f753374598ded88f8cR2695

Maybe Mono could do something similar, I'll try to build Mono and reproduce the issue.

[matouskozak]

Is the problem Mono fails already when it sees a function signature with a NO_CLASS eightbyte? If so, I think only conditional compilation of these test cases would help.

I think that's the case.

One more thing, in VM we simply reclassify NO_CLASS eightbytes to INTEGER, see https://github.com/dotnet/runtime/pull/103799/files#diff-8b4d839010e3c7e363eb430d3efed40b0f9a01a7661e16f753374598ded88f8cR2695

Maybe Mono could do something similar, I'll try to build Mono and reproduce the issue.

Thank you, let me know how it goes.

[tomeksowi]

I tried to build, the NO_CLASS to INTEGER reclassification did solve the test compilation of EmptyStruct but the build fails on other tests on my machine so I couldn't test it.

Another problem is in #105800 I'm adding reflection variants of the tests which also fail on Mono. Unfortunately I'm not familiar enough with Mono to investigate now.

I think it's best to turn off this test for System V Mono for now, these cases are needed mostly for RISC-V and LoongArch to confront passing structs according to hardware FP calling convention with native compilers, empty structs make for interesting cases because they add different field paddings.