Use a ReaderWriterLockSlim in RcwCache

[jkoritzinsky]

This should fix #117775

[Copilot]

Pull Request Overview

This PR replaces the Lock implementation with a ReaderWriterLockSlim in the RcwCache class to address concurrency issues reported in GitHub issue #117775. The change aims to improve thread safety by providing appropriate read/write locking semantics for cache operations.

Key changes:

• Replace Lock with ReaderWriterLockSlim for better read/write concurrency control
• Update method synchronization to use appropriate read/write locks based on operation type
• Convert from lock statements to explicit lock acquisition with proper try/finally blocks

[jkoritzinsky]

@MihuBot benchmark Interop.ComWrappersTests.ParallelRCWLookUp

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/interop-contrib
See info in area-owners.md if you want to be subscribed.

[MihaZupan]

The bot uses an allow list for running commands, added you now.

Method	Toolchain	Mean	Error	Ratio	Allocated	Alloc Ratio
ParallelRCWLookUp	Main	395.9 ms	8.10 ms	1.00	1.66 KB	1.00
ParallelRCWLookUp	PR	209.1 ms	2.94 ms	0.53	1.66 KB	1.00

[jkoritzinsky]

Looks like this fixes the regression!

Thanks @MihaZupan for adding me to the allow-list!

[jkoritzinsky]

/ba-g timeouts