[AaronRobinsonMSFT]

Changes

Replace the fixed-bucket debugger and profiler hash tables with growing SHash implementations.

• Migrate debugger object/field caches and the EE pending-evaluation, module, method-info, and patch tables.
• Migrate profiler allocation-by-class counts, preserving NOTHROW/OOM handling and reusing table capacity between collections.
• Use owning COM enumeration snapshots, preserving reference ownership, locking, and clone/reset/skip behavior.
• Keep patch slots in a separate contiguous array, preserving live slot indices across both SHash rehashing and array growth, duplicate breakpoint chains, and controller ordering.
• Keep LookupPtr read-only and use callback-based Update for existing entries in the shared and NativeAOT SHash implementations. Use native iteration, DAC backing-storage enumeration, and retained-capacity clearing; document borrowed lookup-pointer lifetimes.
• Pair interop-safe nothrow array allocation with matching deallocation, value-initialize reused patch slots, and retain the trivial-copyability assertion required by array relocation.

Removed

Delete CHashTable, CHashTableAndData<>, HASHENTRY, FREEHASHENTRY, HASHFIND, their pointer typedefs, the unused CNewData/CNewDataNoThrow allocators, and the profiler's CHashTableImpl wrapper.

Validation

Built Windows x64 Checked CoreCLR with Release libraries/host. Ran native debugger/SHash checks, cDAC unit tests, runtime loader and Hot Reload suites, managed stack-trace tests, live debugger scenarios, and profiler allocation/GC tests.

Observed performance

A native Windows x64 Checked /O2 lookup diagnostic compared the migrated DBI table with a synthetic legacy 101-bucket table: one million lookups per round, three rounds. Medians below are ns/lookup; lower is better.

Entries	Legacy	SHash	Change
100	5.97	8.47	2.50 ns slower
10,000	184.33	11.34	16.3x faster
100,000	1,590.45	35.41	44.9x faster

These are local synthetic lookup measurements.

Known limitations

• Collectible-module live debugging still reproduces the Debugger::TrapAllRuntimeThreads GC contract assertion previously reproduced on the original baseline. The profiler GC test also prints the same GetObjectGeneration diagnostics as the baseline, despite passing.
• The Windows x86 GCBridge CI failure was not reproduced locally with pre-migration, migrated, or original CI runtimes. Whether the migration caused that failure remains unestablished.
• Actual VS/WinDBG clients, debugger-driven ApplyChanges, and NativeAOT GC-stress execution were not exercised locally.

Related to #11056; the measurements above do not establish a fix for current ordinary Release exception-formatting performance. This broader migration supersedes #132264.

Resolves #11466

Note

This description includes content generated by GitHub Copilot.

[azure-pipelines]

Azure Pipelines:
Successfully started running 3 pipeline(s).
13 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.

[max-charlamb]

This looks good to me from a cDAC perspective. I queued a runtime-diagnostics pipeline on this branch to verify.

[jkotas]

/azp run runtime-coreclr outerloop

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).

[hoyosjs]

I like this better than the old design - that being said to document:

Enumerations will now report the data as seen by the snapshot - so it may miss new or see deleted cordb objects. To me this is ok and a design point. Debuggers might need to adjust to it though.

[jkotas]

LGTM!

It may be a good idea to run some debugger tests if you have not done it already.

[AaronRobinsonMSFT]

@steveisok and @hoyosjs I've shared this PR with our VS peers as an FYI. I've tried a slew of scenarios locally, but obviously not every possible scenario.

[tommcdon]

Changes look good to me. I didn't observe any negative impact to debugger scenarios.