Fix misc memory bugs [main]

[badrishc]

Summary

Two callsites that invoke ZADD internally to populate a destination sorted set were leaking the IMemoryOwner<byte> that the ZADD backend rents from MemoryPool<byte>.Shared.

Root cause

The ZADD backend (SortedSetAdd) writes its integer reply through RespMemoryWriter. When the caller passes a default GarnetObjectStoreOutput (whose SpanByte length is 0), the very first WriteInt32 triggers ReallocateOutput, which rents a MemoryPool<byte> buffer (minimum ~512 bytes) and assigns it to
zAddOutput.SpanByteAndMemory.Memory.

Both callsites had:

var zAddOutput = new GarnetObjectStoreOutput();
RMWObjectStoreOperationWithOutput(destinationKey, ref zAddInput, ..., ref zAddOutput);
// zAddOutput never disposed

Neither disposed the resulting Memory. Under heavy GEO*STORE/ZUNIONSTORE/ZINTERSTORE traffic this caused real MemoryPool churn and GC pressure.

Fix

Wrap each call in try/finally that disposes the Memory if !IsSpanByte:

var zAddOutput = new GarnetObjectStoreOutput();
try
{
    RMWObjectStoreOperationWithOutput(destinationKey, ref zAddInput, ..., ref zAddOutput);
    // ... use zAddOutput as needed ...
}
finally
{
    if (!zAddOutput.SpanByteAndMemory.IsSpanByte)
        zAddOutput.SpanByteAndMemory.Memory?.Dispose();
}

Files changed

• libs/server/Storage/Session/ObjectStore/SortedSetGeoOps.cs (GEO*STORE)
• libs/server/Storage/Session/ObjectStore/SortedSetOps.cs (ZUNIONSTORE, ZINTERSTORE)

Validation

• All 312 RespSortedSet + RespSortedSetGeo tests pass
• dotnet format --verify-no-changes clean

Note on related fixes on dev

This is a subset of the fixes on the companion dev branch (badrishc/memory-fixes). The other fixes from that branch were intentionally not ported here because they do not apply to main:

• The BITOP overflow-pointer fix relies on the ISourceLogRecord / LogRecord / OverflowByteArray model that exists only on dev. On main the BITOP backend operates on SpanByte values that are always pinned in log memory, so the use-after-fixed bug fixed on dev does not exist on main.
• The PFCOUNT/PFMERGE bounds-check tightening from dev is unnecessary on main because the main backend already validates value.Length <= dst.Length before the Buffer.MemoryCopy, so the wrong-capacity argument is gated.

[Copilot]

Pull request overview

Fixes pooled-buffer leaks when internal ZADD operations are invoked to populate destination sorted sets, by ensuring heap-backed outputs are properly disposed.

Changes:

• Wrap internal RMWObjectStoreOperationWithOutput(... ZADD ...) calls in try/finally blocks.
• Dispose GarnetObjectStoreOutput.SpanByteAndMemory.Memory (when heap-backed) to return rented MemoryPool<byte> buffers.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
libs/server/Storage/Session/ObjectStore/SortedSetOps.cs	Disposes internal ZADD output for ZUNIONSTORE/ZINTERSTORE-style range-store flow to prevent MemoryPool buffer leaks.
libs/server/Storage/Session/ObjectStore/SortedSetGeoOps.cs	Disposes internal ZADD output for GEO*STORE flow to prevent MemoryPool buffer leaks.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.