DF-2605 / run.2.log
=== DF-2605 BASELINE run: radix=32 (UB shift path) on UNPATCHED kernel (#0) === On gcc x86-64 -O2, `1U << 32` is UB; the SHL masks the count to 5 bits so bytes wraps to 1 (instead of the intended 4GB). This is a mis-sizing bug (implementation-defined), not a clean panic on this compiler. === Phase 4: Run PoC as unprivileged user maxx (triggers INDIRECT load) === [*] getdents(/mnt/h2test/testdir) -> triggers INDIRECT chain load (DF-2605) [*] getdents returned -1 (errno=33 Numerical argument out of domain) === Phase 5: Cleanup (if we got here, no panic happened) === RUN_DONE (Guest stayed UP. No panic, but chain->bytes is bogus = 1 -> wrong I/O sizing.)