Skip to content
R2Kernel™ Runtime Kernel for Robotics 1.0.0rc10
R2Kernel™ Runtime Kernel for Robotics 1.0.0rc10

    R2Kernel 1.0.0rc5

    RC5 includes the runtime audit fixes, clearer Control measurements, and Python Control connection lifecycle improvements since the published RC4 distribution. It remains a release candidate.

    Changes

    • Package replacement validates the new tree before committing it and retains recovery information. Managed device, process, lease, and IPC access checks preserve the admitted identities and reject stale or unauthorized access.
    • Network frames, generic codec decoding, and deferred work have explicit resource bounds. Unsupported compatibility syscall ABIs are rejected.
    • Federation dispatch checks the current authenticated peer, registration, session, interface, codec, and binding together. Remote lifecycle information remains associated with its authenticated source.
    • SDK preparation, Event lifecycle, and cold Operation registration avoid publishing readiness prematurely or blocking unrelated state queries during cold topic preparation.
    • Installer and upgrade checks prepare the managed Python environment and validate SDK downloads, checksums, archives, and native loading before service cutover. Unsupported legacy HTTPS helper paths fail explicitly.
    • The node graph distinguishes Core receipt rate, protocol completion, failure, deadline overruns, and measurement scope. Long descriptions can be expanded without hiding the rate.
    • Python local Control improves connection lifecycle handling. Authentication, deadline, and failed-exchange checks remain enforced; a failed call is not permission to replay an application effect.
    • The incinerator tutorial reports HIGH transport rejection and failed safe-off acknowledgements. Pulse timing, leases, watchdog behavior, and mandatory safe-off are unchanged.

    Validation and known limitations

    The fixed-application Linux Release comparison used Python controller and gateway fixtures with 8-byte requests and responses, a 1 ms period and deadline, and 1,000 measured calls per run. The corrected SDK succeeded in 911 and 887 calls; the restored original SDK succeeded in zero calls between those runs. Successful RTT medians were approximately 0.30 and 0.29 ms. Failed calls and late cycles remain: this does not satisfy the 1 ms / 1 kHz requirement and does not qualify physical-device or hard real-time performance.

    The trace investigation found connection startup costs and runnable scheduling delays in failed exchanges. Some failures occur while consuming a response or waiting for buffer release. The cause of every remaining miss has not been isolated. Core completion counters describe protocol release completion, not handler acknowledgement or physical output completion.

    The historical incinerator fired:false failure was not reproduced in thirteen fixed Release collections. Separate cold smoke checks with the logging changes also passed. Its exact historical cause remains unconfirmed; the additional failure logs improve diagnosis rather than proving it resolved.

    See Installation, Resource Control, and Time Sync Operations for installation and platform-specific operating requirements.