Multiple qemu usermode tests disabled until a bug is fixed

[rmalmain]

#3650 disables tests for qemu_coverage and qemu_tmin because of a bug in the restarting manager.
they should be enabled again once the bug is fixed.

[rashyy17]

Hi,
I’d love to work on this issue if it’s available. Please let me know if I can take it up.

[Tanmay140906]

Hi @rmalmain , I’ve been digging deeper into this issue and spent past few of my days carefully reading through qemu_tmin and qemu_coverage fuzzers, events/restarting.rs and the LLMP restarting flow. From what I understand, the restarting layer itself looks architecturally correct. It serializes (Option, EM::RestartState) via StateRestorer, restores them properly in setup_generic_restarting_mgr, and also reconstructs the event manager using the saved inner state. I don’t see an obvious inconsistency in how state restoration is handled there.

At this point, it seems more likely that the issue is at the boundary between restored StdState and the freshly rebuilt runtime components in the QEMU fuzzers (coverage shared memory, observers, feedback instances, executor, etc.). I’m now reviewing MaxMapFeedback, ObserverEqualityFactory, and StdTMinMutationalStage to check whether any of them implicitly assume continuity of observer/coverage state across restarts.

Could you tell me if I am approaching it correctly or I should instead be looking elsewhere (e.g., QEMU module wiring or snapshot interaction) ?