Documentation

Miri.Interpreter.GlobalAllocationMachine

Lazy compiler-global conversion, pinned machine.rs:967–1032,1678–1693. Byte backing is acquired before relocation conversion. Global race clocks start at zero on the main thread, independently of the accessing thread. The returned owned allocation is cached by MemoryLifecycle, never by this callback.

Real storage effects; native preparation must reserve the actual backing identity used by global-root address resolution. Managed storage explicitly returns no host owner. Release runs on failed conversion, like Rust's drop.

Instances For

    The managed backend's semantic bytes have no native ownership. Receiving a native owner here is an inconsistent backend construction, not a no-op.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      Instances For
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For

          Global and TLS kinds share this source zero/main-thread race category; both may leak and neither stores an allocation span (machine.rs:222–239).

          Equations
          Instances For
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For

              Driver used by ensureTreeExtra. It follows the same lazy conversion/cache path as ordinary reads; typed callback failures survive the string boundary.

              Equations
              • One or more equations did not get rendered due to their size.
              Instances For