JIT test GenericCatchInterfaceProgram fails in Checked NativeAOT

[agocke]

Stack trace:

00:15:44.971 Running test: JIT/Generics/Exceptions/GenericCatchInterfaceProgram/GenericCatchInterfaceProgram.dll
Unhandled Exception: System.Diagnostics.DebugProvider+DebugAssertException: IsInstanceOfClass called with interface MethodTable
   at System.Diagnostics.DebugProvider.Fail(String, String) + 0x35
   at System.Diagnostics.Debug.Fail(String, String) + 0x30
   at System.Runtime.TypeCast.IsInstanceOfClass(MethodTable*, Object) + 0x75
   at System.Runtime.EH.FindFirstPassHandler(Object, UInt32, StackFrameIterator&, UInt32&, Byte*&) + 0x108
   at System.Runtime.EH.DispatchEx(StackFrameIterator&, EH.ExInfo&) + 0x13f
   at System.Runtime.EH.RhThrowEx(Object, EH.ExInfo&) + 0x4e
   at GenericCatchInterfaceProgram.Main() + 0x27
   at Program.<<Main>$>g__TestExecutor23|0_24(StreamWriter tempLogSw, StreamWriter statsCsvSw, Program.<>c__DisplayClass0_0&) + 0xa5
   at Program.<Main>$(String[] args) + 0x338
   at JIT!<BaseAddress>+0x477a6c

[ghost]

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

Issue Details

---

Stack trace:

00:15:44.971 Running test: JIT/Generics/Exceptions/GenericCatchInterfaceProgram/GenericCatchInterfaceProgram.dll
Unhandled Exception: System.Diagnostics.DebugProvider+DebugAssertException: IsInstanceOfClass called with interface MethodTable
   at System.Diagnostics.DebugProvider.Fail(String, String) + 0x35
   at System.Diagnostics.Debug.Fail(String, String) + 0x30
   at System.Runtime.TypeCast.IsInstanceOfClass(MethodTable*, Object) + 0x75
   at System.Runtime.EH.FindFirstPassHandler(Object, UInt32, StackFrameIterator&, UInt32&, Byte*&) + 0x108
   at System.Runtime.EH.DispatchEx(StackFrameIterator&, EH.ExInfo&) + 0x13f
   at System.Runtime.EH.RhThrowEx(Object, EH.ExInfo&) + 0x4e
   at GenericCatchInterfaceProgram.Main() + 0x27
   at Program.<<Main>$>g__TestExecutor23|0_24(StreamWriter tempLogSw, StreamWriter statsCsvSw, Program.<>c__DisplayClass0_0&) + 0xa5
   at Program.<Main>$(String[] args) + 0x338
   at JIT!<BaseAddress>+0x477a6c

Author:	agocke
Assignees:	-
Labels:	area-NativeAOT-coreclr
Milestone:	-

[jkotas]

Regression introduced by #90234

[VSadov]

Interesting. I did not realize it is possible to use non-class types in the catch (arrays, interfaces).

[VSadov]

What is the easiest/right way to run this test? It looks like it does not run as a part of regular src\tests\run.cmd

[jkotas]

The test was disabled in https://github.com/dotnet/runtime/pull/89421/files#diff-72d87bb5dc27f15b0bfbc2423cff89c3085f4109dc8cd5d83e1724d828ebbde0R703-R705 . It will run once you revert these lines - make sure to include the revert in your PR.

[VSadov]

If I remove the suppression lines in src/tests/issues.targets I do not see the test failure (with or without my fix).

The way I run tests is:

taskkill /f /im dotnet.exe
git clean -xdf
build clr+libs -rc Checked -lc Release & src\tests\build.cmd nativeaot Checked tree nativeaot &  src\tests\run.cmd runnativeaottests Checked

I assume it is some kind of recently enabled test (since it did not catch the regression).
Is there some other switch that is needed to enable it?

[jkotas]

tree nativeaot is only going to run the tests in nativeaot subdirectory. You need to remove that argument to run all tests.

[VSadov]

Right, that makes sense. Will try running without tree nativeaot.
The switch used to be necessary some time ago, so I added it to the list of commands that I use.

[VSadov]

Hmm. It takes a lot longer, but there are still no failures, other than SmokeTests\HardwareIntrinsics\X64Avx512, which always fails.

The command I used:

build clr+libs -rc Checked -lc Release & src\tests\build.cmd nativeaot Checked &  src\tests\run.cmd runnativeaottests Checked

[jkotas]

Hmm, you are right - the failure shows up in the console output:

 C:\runtime\artifacts\tests\coreclr\windows.x64.Checked\JIT\Methodical\Methodical_ro\Methodical_ro.cmd >C:\runtime\art
  ifacts\tests\coreclr\windows.x64.Checked\JIT\Methodical\Methodical_ro\Methodical_ro.log
  Unhandled Exception: System.Diagnostics.DebugProvider+DebugAssertException: IsInstanceOfClass called with interface M
  ethodTable
     at System.Diagnostics.DebugProvider.Fail(String, String) + 0x36
     at System.Diagnostics.Debug.Fail(String, String) + 0x32
     at System.Runtime.TypeCast.IsInstanceOfClass(MethodTable*, Object) + 0x72
     at System.Runtime.EH.FindFirstPassHandler(Object, UInt32, StackFrameIterator&, UInt32&, Byte*&) + 0xfb
     at System.Runtime.EH.DispatchEx(StackFrameIterator&, EH.ExInfo&) + 0x131
     at System.Runtime.EH.RhThrowEx(Object, EH.ExInfo&) + 0x4b
     at GenericCatchInterfaceProgram.Main() + 0x2a
     at Program.<<Main>$>g__TestExecutor23|0_24(StreamWriter tempLogSw, StreamWriter statsCsvSw, Program.<>c__DisplayCl
  ass0_0&) + 0xa7
     at Program.<Main>$(String[] args) + 0x337
     at JIT.Generics!<BaseAddress>+0x26a500
  C:\runtime\artifacts\tests\coreclr\windows.x64.Checked\JIT\opt\JIT.opt\JIT.opt.cmd >C:\runtime\artifacts\tests\corecl
  r\windows.x64.Checked\JIT\opt\JIT.opt\JIT.opt.log

But the test is not marked as failed in the summary at the end. Looks like a bug in the test infrastructure.

[VSadov]

I am also able to repro this by running directly
artifacts\tests\coreclr\windows.x64.Checked\JIT\Generics\JIT.Generics\native\JIT.Generics.exe

That is enough for me to verify the fix and the sensitivity of the test to the fix.

But the test is not marked as failed in the summary at the end. Looks like a bug in the test infrastructure.

Looks like it.
Also it is possible that it is a results reporting issue specific to local runs only, since somehow the bug was found in the CI.

[jkotas]

Could you please open a new issue on the test failure reporting problem?

[VSadov]

Yes, will open an issue on that.