cDAC: Add GetGCDescSeries contract API and continuation pretty-printing by Copilot · Pull Request #127419 · dotnet/runtime

Copilot · 2026-04-25T08:51:25Z

Summary

Implements continuation pretty-printing in the cDAC, mirroring AsyncContinuationsManager::PrintContinuationName from asynccontinuations.h, and adds a new GetGCDescSeries API to the RuntimeTypeSystem contract.

Fixes the WebApp3 test failure seen in recent runtime-diagnostics legs such as this: https://dev.azure.com/dnceng-public/public/_build/results?buildId=1395576

Note

This PR was created with the assistance of GitHub Copilot (AI-generated content).

…ation pretty-printing Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6e988c50-2e06-4957-952d-9b624dd862ca Co-authored-by: rcj1 <77995559+rcj1@users.noreply.github.com>

jkotas · 2026-04-25T13:40:04Z

+
+        // NumSeries is stored one pointer-width before the method table. When interpreted as a signed
+        // value, a negative count indicates a value-class (repeating) series, which uses a different
+        // layout and is not supported by this API.


We may want to have a general purpose GCDesc enumeration, something that can replace
https://github.com/microsoft/clrmd/blob/main/src/Microsoft.Diagnostics.Runtime/GCDesc.cs

(GCDesc is one of those internal data contracts that CLRMD and SOS have hardcoded implicit dependency on that's not ideal.)

Copilot

Pull request overview

This PR extends the cDAC RuntimeTypeSystem contract surface to expose GCDesc “series” (GC pointer runs) and uses that information to pretty-print dynamically-created continuation types in a way that mirrors the runtime’s AsyncContinuationsManager::PrintContinuationName.

Changes:

Add IRuntimeTypeSystem.GetGCDescSeries and implement it in RuntimeTypeSystem_1 to enumerate GC pointer runs from the GCDesc preceding a MethodTable.
Add continuation synthetic-name formatting to TypeNameBuilder using GetGCDescSeries.
Add a new ContinuationObject data descriptor/type entry plus mocks, tests, and contract documentation updates.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/native/managed/cdac/tests/MockDescriptors/MockDescriptors.RuntimeTypeSystem.cs	Adds mock helpers to allocate MethodTables with GCDesc (regular + repeating/value-class layouts) and exposes `ContinuationObjectSize`.
src/native/managed/cdac/tests/MethodTableTests.cs	Adds unit tests covering `GetGCDescSeries` behaviors across MethodTable/non-MethodTable, no-GC, regular series, and repeating series cases.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Legacy/TypeNameBuilder.cs	Uses `IsContinuation` + `GetGCDescSeries` to generate the `Continuation_<dataSize>[_<gcOffset>_<gcCount>]*` synthetic name instead of `(dynamicClass)`.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Contracts/Contracts/RuntimeTypeSystem_1.cs	Implements `GetGCDescSeries` by reading GCDesc memory preceding the MethodTable and normalizing sizes.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Abstractions/DataType.cs	Adds `DataType.ContinuationObject` for retrieving continuation header sizing info from the contract descriptor.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Abstractions/Contracts/IRuntimeTypeSystem.cs	Adds the `GetGCDescSeries(TypeHandle, uint objectSize)` contract API.
src/coreclr/vm/datadescriptor/datadescriptor.inc	Adds a `ContinuationObject` type descriptor entry so the reader can query its size.
docs/design/datacontracts/RuntimeTypeSystem.md	Updates contract documentation to include the new API and describe GCDesc series layouts.

dotnet-policy-service · 2026-04-25T21:05:38Z

Tagging subscribers to this area: @steveisok, @tommcdon, @dotnet/dotnet-diag
See info in area-owners.md if you want to be subscribed.

hoyosjs · 2026-04-27T18:55:25Z

~~Might be missing https://github.com/hoyosjs/runtime/blob/9ef63b95ace0796ca363bc53687ddbb670f1119b/src/coreclr/gc/gcdesc.h#L227-L231.~~ NVM - looks like this is not stored inline with CGCDescSeries array

jkotas · 2026-04-27T20:54:46Z

+            long absNumSeries = -numSeries;
+            ulong startOffset = _target.ReadPointer(mtAddress - 2 * pointerSize).Value;
+
+            ulong currentOffset = startOffset;


If this is trying to replicate https://github.com/microsoft/clrmd/blob/main/src/Microsoft.Diagnostics.Runtime/GCDesc.cs exactly, this would need to take the passed in object size into account too and keep repeating the sequence.

I am not sure whether the API shape that takes object size works best for all cases. It assumes that you have object size that is not always the case, like for the Continuation pretty printing.

Ok. I don't see an issue with taking the object size; Continuation method tables do have the total object size stored on the method table flags, accessible through the GetBaseSize API.

runtime/src/coreclr/vm/asynccontinuations.cpp

Line 157 in 23d2e12

pMT->SetBaseSize(OBJECT_BASESIZE + startOfDataInInstance + dataSize);

Could we take an optional numComponents instead of objectSize

For non-array types, baseSize==objectSize, for array types objectSize can be found from the numComponents and TypeHandle.

This removes redundant lookups at the call sites. The non-array path in AppendContinuationName does GetBaseSize only to pass it back in the
contract that already knows BaseSize.

Suggested shape:

// Default 0 covers non-arrays cleanly: objectSize == BaseSize. IEnumerable<(uint SeriesOffset, uint SeriesSize)> GetGCDescSeries( TypeHandle typeHandle, uint numComponents = 0);

Inside the implementation:

uint baseSize = GetBaseSize(typeHandle); uint componentSize = GetComponentSize(typeHandle); uint objectSize = baseSize + numComponents * componentSize; // ...existing logic...

This also matches native's CGCDesc::GetNumPointers(MT*, ObjectSize, NumComponents) more closely in spirit, and lets AppendContinuationName
simplify to GetGCDescSeries(typeHandle) (no baseSize local needed).

Sounds good

IEnumerable<(uint SeriesOffset, uint SeriesSize)> GetGCDescSeries(TypeHandle typeHandle, uint numComponents = 0);

This shape works well for enumerating all GC references in the given object instance. It does not work well for dumping GCInfo without an object instance, or for efficient or partial enumeration of GC references.

We can start with this shape; I am just pointing its short-comings.

hoyosjs · 2026-04-27T21:19:20Z

dotnet/diagnostics#5815

Copilot

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 1 comment.

Copilot · 2026-04-28T23:31:17Z

+        /// </summary>
+        /// <param name="name">Descriptive name for the allocation.</param>
+        /// <param name="baseSize">Value to store in the method table's <c>BaseSize</c> field.</param>
+        /// <param name="startOffset">The <c>startoffset</c> field of the CGCDescSeries (offset from object start).</param>


The startOffset here is described as an offset from the "object start", but GCDesc startoffset is an offset from the managed object pointer (MethodTable) and excludes the syncblock/ObjectHeader prefix. Clarifying this in the parameter documentation (e.g., "offset from the object pointer / first byte of the Object", not including ObjectHeader) would help prevent callers from accidentally including OBJHEADER_SIZE (which would shift all series offsets).

Suggested change

/// <param name="startOffset">The <c>startoffset</c> field of the CGCDescSeries (offset from object start).</param>

/// <param name="startOffset">

/// The <c>startoffset</c> field of the CGCDescSeries, as an offset from the managed object

/// pointer (the first byte of the object). This does not include any sync block /

/// <c>ObjectHeader</c> prefix.

/// </param>

Add GetGCDescSeries API, OffsetOfContinuationData global, and continu…

36201f0

…ation pretty-printing Agent-Logs-Url: https://github.com/dotnet/runtime/sessions/6e988c50-2e06-4957-952d-9b624dd862ca Co-authored-by: rcj1 <77995559+rcj1@users.noreply.github.com>

Copilot AI assigned Copilot and rcj1 Apr 25, 2026

Copilot AI requested review from Copilot and removed request for Copilot April 25, 2026 08:51

Copilot created this pull request from a session on behalf of rcj1 April 25, 2026 08:52 View session

Copilot AI requested a review from rcj1 April 25, 2026 08:52

Copilot finished work on behalf of rcj1 April 25, 2026 08:52

github-actions Bot added the area-Diagnostics-coreclr label Apr 25, 2026

jkotas reviewed Apr 25, 2026

View reviewed changes

Comment thread docs/design/datacontracts/RuntimeTypeSystem.md

cleanup and code review

ecd3dd1

Copilot AI review requested due to automatic review settings April 25, 2026 18:43

Copilot started reviewing on behalf of rcj1 April 25, 2026 18:44 View session

Copilot AI reviewed Apr 25, 2026

View reviewed changes

Comment thread src/native/managed/cdac/tests/MethodTableTests.cs

rcj1 requested review from max-charlamb and noahfalk April 25, 2026 19:41

rcj1 marked this pull request as ready for review April 25, 2026 19:42

rcj1 added area-Diagnostics-cdac and removed area-Diagnostics-coreclr labels Apr 25, 2026

max-charlamb mentioned this pull request Apr 27, 2026

[cDAC] Fix GetMethodTableName empty type name mismatch #127405

Closed

jkotas reviewed Apr 27, 2026

View reviewed changes

rcj1 added 2 commits April 27, 2026 14:36

update outer loop

c21229e

adding component size param and adding test

1cc0f46

Copilot AI review requested due to automatic review settings April 28, 2026 23:20

Copilot started reviewing on behalf of rcj1 April 28, 2026 23:20 View session

Copilot AI reviewed Apr 28, 2026

View reviewed changes

-        /// <param name="startOffset">The <c>startoffset</c> field of the CGCDescSeries (offset from object start).</param>
+        /// <param name="startOffset">
+        /// The <c>startoffset</c> field of the CGCDescSeries, as an offset from the managed object
+        /// pointer (the first byte of the object). This does not include any sync block /
+        /// <c>ObjectHeader</c> prefix.
+        /// </param>

Conversation

Copilot AI commented Apr 25, 2026 • edited by rcj1 Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Uh oh!

jkotas Apr 25, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

dotnet-policy-service Bot commented Apr 25, 2026

Uh oh!

hoyosjs commented Apr 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jkotas Apr 27, 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 Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

max-charlamb Apr 28, 2026

Choose a reason for hiding this comment

Uh oh!

rcj1 Apr 28, 2026

Choose a reason for hiding this comment

Uh oh!

jkotas Apr 28, 2026

Choose a reason for hiding this comment

Uh oh!

hoyosjs commented Apr 27, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 28, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Copilot AI commented Apr 25, 2026 •

edited by rcj1

Loading

hoyosjs commented Apr 27, 2026 •

edited

Loading

jkotas Apr 27, 2026 •

edited

Loading