[max-charlamb]

Summary

Return retained stack-context snapshots as ReadOnlySpan<byte> instead of allocating a new byte[] for every output request.

This changes only the context-output APIs:

• IStackWalk.GetRawContext returns a read-only view of the frame's retained context snapshot.
• IPlatformAgnosticContext.GetBytes returns a read-only view over the serialized platform context.
• COM and debugger consumers copy directly from the span where possible.
• APIs that require owned or writable context storage continue to materialize arrays explicitly.

The span remains backed by the context snapshot owned by the stack frame handle. Advancing the walk does not mutate that retained snapshot.

Performance

Measured frozen NativeAOT readers with provider identities reversed between two passes. Each pass used 3 launches with 30 observations per case, with no outlier removal. The primary run retained 2,520 observations across 14 timing cases.

Scenario	Before	After	Change	Reversed-order changes
Enum CodeHeavy Heap	4.665 ms	4.554 ms	-2.4%	-0.7% / -4.0%
Enum DeepStacks Heap	12.585 ms	12.182 ms	-3.2%	-5.4% / -0.8%
Enum SmallSimple Heap	2.787 ms	2.738 ms	-1.8%	-4.8% / +0.9%
Stack DeepStacks Heap Cold	9.040 ms	8.668 ms	-4.1%	-2.6% / -5.7%
Stack DeepStacks Heap Warm	8.645 ms	8.504 ms	-1.6%	-1.5% / -1.8%
Stack DeepStacks Mini Cold	11.450 ms	11.640 ms	+1.7%	+2.3% / +1.0%
Stack DeepStacks Mini Warm	11.013 ms	10.913 ms	-0.9%	+3.5% / -5.4%

A separate cold-walk confirmation used two more reversed passes and 720 additional observations:

Scenario	Before	After	Change	Reversed-order changes
Stack DeepStacks Heap Cold	8.692 ms	8.199 ms	-5.7%	-4.2% / -7.2%
Stack DeepStacks Mini Cold	11.185 ms	10.637 ms	-4.9%	-4.2% / -5.6%

The change removes one temporary 1,232-byte array for each AMD64 context output. Dump-enumeration timing does not directly exercise the migrated output APIs, and shared-host timing differences are not guaranteed gains.

Correctness checks confirmed identical serialized bytes for all seven platform context layouts, exact stack parity across 18 native/baseline/current cases, identical Mini/Heap/Triage dump ranges and updates, and matching context bytes for six real-dump cases in cold and warm states.

Testing

• dotnet test src\native\managed\cdac\tests\UnitTests\Microsoft.Diagnostics.DataContractReader.Tests.csproj -c Release --no-restore -m:1 -p:BuildInParallel=false -p:RuntimeConfiguration=Release -p:LibrariesConfiguration=Release -- passed, 3,914 tests.
• dotnet test src\native\managed\cdac\tests\DumpTests\Microsoft.Diagnostics.DataContractReader.DumpTests.csproj -c Release --no-restore -m:1 -p:BuildInParallel=false -p:RuntimeConfiguration=Release -p:LibrariesConfiguration=Release -p:SkipDumpVersions=net10.0 --filter 'FullyQualifiedName~StackWalkDumpTests|FullyQualifiedName~IXCLRDataFrameDumpTests|FullyQualifiedName~DacDbiStackWalkDumpTests' -- passed, 63 tests with 19 expected skips.

Note

This pull request description was generated with GitHub Copilot.

[azure-pipelines]

Azure Pipelines:
Successfully started running 5 pipeline(s).
10 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

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