Runtime-async Exception.ToString() #122722

rcj1 · 2025-12-24T05:50:09Z

This PR implements exception stack trace collection in runtime-async stacks for both coreclr and nativeAOT.

Augmenting stack unwinding logic in AsyncHelpers.CoreCLR.cs to append frames from runtime-async async-await chain to exception stacktrace using DiagnosticIP of Continuation.
Implements ThrowExact helper to throw exception that has been stored in Continuation, while keeping its existing stack trace intact. Accomplished by creating throw helpers that set ExKind.RethrowFlag. Would be open to rename "RethrowFlag" and companions to "NoEraseFlag" or something similar, or leave as is.
ILC changes: Ensuring that stack trace records are emitted for runtime-async methods, and hidden for resumption stubs.
Testing: Added tests for line/method/document correctness including v1-v2 chaining.

@dotnet/ilc-contrib

- Add new JIT-EE API to report back debug information about the generated state machine and continuations - Refactor debug info storage on VM side to be more easily extensible. The new format has either a thin or fat header. The fat header is used when we have either uninstrumented bounds, patchpoint info, rich debug info or async debug info, and stores the blob sizes of all of those components in addition to the bounds and vars. - Add new async debug information to the storage on the VM side - Set get target method desc for async resumption stubs, to be used for mapping from continuations back to the async IL function that it will resume.

…mental APIs in async tests

…ostics

jkotas · 2026-01-24T04:53:53Z

src/coreclr/vm/debugdebugger.cpp

+    _ASSERTE(gc.pException != NULL);
+
+    // populate exception with information from the continuation object
+    OBJECTHANDLE handle = AppDomain::GetCurrentDomain()->CreateHandle(gc.pException);


Where is this GC handle going to be deleted? It looks like a leak. We may need StackTraceInfo::AppendElement overload that takes OBJECTREF. It would be wasteful to create GC handle just to pass an object reference to a method as an argument.

src/coreclr/vm/excep.cpp

src/libraries/Common/tests/TestUtilities/System/PlatformDetection.cs

jkotas · 2026-01-24T05:13:06Z

src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/StackTraceEmissionPolicy.cs

            if (method.HasCustomAttribute("System.Diagnostics", "StackTraceHiddenAttribute")
-                || (method.OwningType is MetadataType mdType && mdType.HasCustomAttribute("System.Diagnostics", "StackTraceHiddenAttribute")))
+                || (method.OwningType is MetadataType mdType && mdType.HasCustomAttribute("System.Diagnostics", "StackTraceHiddenAttribute"))
+                || (method is ILCompiler.AsyncResumptionStub)


@MichalStrehovsky Could you please double check changes in this method?

We have all sort of stubs. It looks odd that we check for AsyncResumptionStub and no other stubs here.

In coreclr, we filter these out of stacktraces with IsDiagnosticsHidden() in StackTraceInfo::AppendElement, which filters out IL stubs and async thunks. Since we do not pre-filter the stacktrace in nativeaot, we use the hidden attribute; I do believe we need this check, but we may want more.

For example, with the following test app

using System; class Program { static void Main() { Action a = () => Console.WriteLine("A"); Action b = () => throw new Exception("This i"); Action m = a + b; m(); } }

We get the following stacktrace

at Program.<>c.<Main>b__0_1() + 0x27 at aot!<BaseAddress>+0x22046d at Program.Main() + 0x9f

where aot!+0x22046d is the delegate multicast stub. How should these look - maybe we broaden this to ILStubMethod to match coreclr?

rcj1 · 2026-01-24T05:56:37Z

The new tests are failing on NativeAOT on Unix, e.g.:

2026-01-24T04:26:55.9401240Z   Illegal text-relocations:
2026-01-24T04:26:55.9405630Z     text-relocation in '_System_Diagnostics_StackTrace_Tests_System_Diagnostics_V1Methods__<AsyncCallable>System_Diagnostics_StackTrace_Tests_System_Diagnostics_V1Methods__Test0'+0x160 (/Users/runner/work/1/s/artifacts/obj/System.Diagnostics.StackTrace.Tests/Release/net11.0/native/System.Diagnostics.StackTrace.Tests.o) to '_System_Diagnostics_StackTrace_Tests_System_Diagnostics_V1Methods__<AsyncCallable>System_Diagnostics_StackTrace_Tests_System_Diagnostics_V1Methods__Test0'

They need to be disabled or we need to wait for @jakobbotsch fix (#123433).

Ok I can add that into the platform detection condition with a comment on this issue

jkotas · 2026-01-24T21:57:54Z

/azp run runtime-nativeaot-outerloop

azure-pipelines · 2026-01-24T21:58:11Z

Azure Pipelines successfully started running 1 pipeline(s).

jkotas · 2026-01-25T15:02:11Z

src/coreclr/vm/excep.cpp

        // Fetch the stacktrace and the keepAlive array from the exception object. It returns clones of those arrays in case the
        // stack trace was created by a different thread.
-        ((EXCEPTIONREF)ObjectFromHandle(hThrowable))->GetStackTrace(gc.stackTrace.m_pStackTraceArray, &gc.pKeepAliveArray, pThread);
+        ((EXCEPTIONREF)pThrowable)->GetStackTrace(gc.stackTrace.m_pStackTraceArray, &gc.pKeepAliveArray, pThread);


A lot of CI legs failed due to infra issues, but the few that completed has test failures that suggest GC heap corruption.

I think you need to GCPROTECT the throwable. You may want to move the GCPROTECT up towards the start of the method so that there is still just one GCPROTECT in the whole method.

jkotas · 2026-01-25T15:04:53Z

/azp run runtime-nativeaot-outerloop

azure-pipelines · 2026-01-25T15:05:09Z

Azure Pipelines successfully started running 1 pipeline(s).

src/coreclr/vm/excep.cpp

jkotas · 2026-01-25T15:13:16Z

src/coreclr/vm/excep.cpp

+    _ASSERTE(IsException(pMT));
+
+    PTR_ThreadExceptionState pCurTES = pThread->GetExceptionState();
+    // Check if the flag indicating foreign exception raise has been setup or not,


@janvorli Could you please review these changes? In particular, the changes around IsRaisingForeignException flag for runtime async frames. Do we need to worry about it at all for runtime async frames?

I do not believe so. This flag is only set during ExceptionDispatchInfo.Throw(), so the only place it would matter is if Edi.Throw() is immediately above a runtime-async frame in a stacktrace. However, above a runtime-async frame we will either have another runtime-async frame, or DispatchContinuations, the only call site of this method.

rcj1 · 2026-01-25T18:41:19Z

/azp run runtime-nativeaot-outerloop

azure-pipelines · 2026-01-25T18:41:28Z

Azure Pipelines successfully started running 1 pipeline(s).

rcj1 · 2026-01-26T01:02:11Z

/azp run runtime-nativeaot-outerloop

azure-pipelines · 2026-01-26T01:02:25Z

Azure Pipelines successfully started running 1 pipeline(s).

jkotas · 2026-01-26T01:47:35Z

src/libraries/System.Diagnostics.StackTrace/tests/StackTraceTests.cs

        }
    }
+
+#if !(NATIVEAOT && !WINDOWS)


I doubt that this is going to work. The build of libraries partition is runtime flavor agnostic. The only place under libraries where you can use NATIVEAOT ifdef is CoreLib - that's because the CoreLib is not actually built from under libraries, but from runtime flavor specific location (e.g. src\coreclr\System.Private.CoreLib or src\coreclr\nativeaot\System.Private.CoreLib).

Building libraries tests for NativeAOT can be disabled at project granularity https://github.com/dotnet/runtime/blob/main/src/libraries/tests.proj#L401

rcj1 · 2026-01-26T02:01:15Z

/azp run runtime-nativeaot-outerloop

azure-pipelines · 2026-01-26T02:01:26Z

Azure Pipelines successfully started running 1 pipeline(s).

rcj1 · 2026-01-26T02:46:18Z

/azp run runtime-nativeaot-outerloop

azure-pipelines · 2026-01-26T02:46:27Z

Azure Pipelines successfully started running 1 pipeline(s).

janvorli · 2026-01-26T16:05:21Z

src/coreclr/nativeaot/Runtime/amd64/ExceptionHandling.S


+        mov     r8d, 1                  // r8d = ExKind.Throw
+
+LOCAL_LABEL(RhpThrowExImpl):


This should not be marked as LOCAL_LABEL but rather ALTERNATE_ENTRY as it is called from outside of this function. For all architectures.

janvorli · 2026-01-26T16:22:45Z

src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/StackTraceEmissionPolicy.cs

            if (method.HasCustomAttribute("System.Diagnostics", "StackTraceHiddenAttribute")
-                || (method.OwningType is MetadataType mdType && mdType.HasCustomAttribute("System.Diagnostics", "StackTraceHiddenAttribute")))
+                || (method.OwningType is MetadataType mdType && mdType.HasCustomAttribute("System.Diagnostics", "StackTraceHiddenAttribute"))
+                || (method is Internal.IL.Stubs.ILStubMethod)


I wonder why the IL stubs were not a problem before this change.

See #122722 (comment), they did show up as unknowns but this should filter them out.

janvorli · 2026-01-26T16:25:32Z

src/coreclr/vm/debugdebugger.cpp

+    _ASSERTE(pException != NULL);
+    // populate exception with information from the continuation object
+    EECodeInfo codeInfo((PCODE)diagnosticIP);
+    if (codeInfo.IsValid())


Is there a valid case when the codeInfo is not valid here? If not, it would be nice to add _ASSERTE here.

janvorli · 2026-01-26T16:44:56Z

src/coreclr/vm/excep.cpp

        return;
    }

+    AppendElementImpl(pThrowable, currentIP, currentSP, pFunc, pCf, pThread, fRaisingForeignException);


While it cannot happen now, if some GC triggering call gets included between the point we get the pThrowable and this call, GC gets triggered and moves the object, we could end up passing stale pThrowable to the call. It would be safer to use ObjectFromHandle(hThrowable) as argument here.

janvorli · 2026-01-26T16:57:21Z

src/coreclr/vm/exkind.h

+#ifndef HAVE_EXKIND_H
+#define HAVE_EXKIND_H
+
+#include <cstdint>


Is seems that since you've moved it to the common.h, it is not needed here anymore, according to @am11's comment.

I thought that was just regarding @am11

jakobbotsch and others added 30 commits October 1, 2025 14:54

Set target method desc for resumption stubs

b7bb968

Add JIT-EE boilerplate

d5d0864

Update managed view

6addd0e

Remove TODOs

6a5bf19

Leaf transition

262260a

Comment

4e2a829

Delete unused enum

820afa9

Restore comment

3621eae

Fix osx build

f597424

Fix GCC build

6bb6cee

Change sentinel value, fix contract

f580e3f

Bump R2R

55b9522

Clean up

c9c64be

Expose async debug info accessor APIs

db77446

Missed bumping R2R version for naot

c808390

Fix reverse mapping to IL local nums

3ec5160

Fix monotonicity for async vars

bb12c77

Code style

bf3364b

Fix JIT-EE prompt tools from Egor's instructions, allow use of experi…

b49f38f

…mental APIs in async tests

Address feedback, make comment less misleading

d1bf49f

Add AsyncContinuationVarInfo.GCIndex

579714e

Publish NextContinuation in TLS

c056793

Rename ThunkTask -> RuntimeAsyncTask

9679f91

testing

abd7d8b

Merge branch 'main' of github.com:dotnet/runtime into jit-async-diagn…

7a34475

…ostics

Report native offsets instead

e839409

Run jit-format

d8ad0c1

Print reported async debug info, always report it

9fdd8a7

exception tostring

1da796e

jkotas reviewed Jan 24, 2026

View reviewed changes

src/libraries/Common/tests/TestUtilities/System/PlatformDetection.cs Outdated Show resolved Hide resolved

jkotas reviewed Jan 24, 2026

View reviewed changes

code review

3bd93bd

rcj1 added 2 commits January 24, 2026 17:09

Update StackTraceEmissionPolicy to include ILStubMethod

4b68b77

e

19c9248

jkotas reviewed Jan 25, 2026

View reviewed changes

src/coreclr/vm/excep.cpp Outdated Show resolved Hide resolved

jkotas reviewed Jan 25, 2026

View reviewed changes

gc etc

c8d340d

disable tests

6155e83

jkotas reviewed Jan 26, 2026

View reviewed changes

disable tests

0d8d237

janvorli reviewed Jan 26, 2026

View reviewed changes


		mov r8d, 1 // r8d = ExKind.Throw

		LOCAL_LABEL(RhpThrowExImpl):

Runtime-async Exception.ToString() #122722

Are you sure you want to change the base?

Runtime-async Exception.ToString() #122722

Conversation

rcj1 commented Dec 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rcj1 Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rcj1 commented Jan 24, 2026

Uh oh!

jkotas commented Jan 24, 2026

Uh oh!

azure-pipelines bot commented Jan 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jkotas commented Jan 25, 2026

Uh oh!

azure-pipelines bot commented Jan 25, 2026

Uh oh!

Uh oh!

jkotas Jan 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rcj1 commented Jan 25, 2026

Uh oh!

azure-pipelines bot commented Jan 25, 2026

Uh oh!

rcj1 commented Jan 26, 2026

Uh oh!

azure-pipelines bot commented Jan 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rcj1 commented Jan 26, 2026

Uh oh!

azure-pipelines bot commented Jan 26, 2026

Uh oh!

rcj1 commented Jan 26, 2026

Uh oh!

azure-pipelines bot commented Jan 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

rcj1 commented Dec 24, 2025 •

edited

Loading

rcj1 Jan 24, 2026 •

edited

Loading

jkotas Jan 25, 2026 •

edited

Loading