Bug in lock-free sofi about iceoryx

[budrus]

@elfenpiff This also assumes that there are multiple pushers. SOFI was designed for SPSC. All we said way that by accident is could also be SPMC, but we never had MPMC in mind

[elBoberido]

Yes, this is not an issue for iceoryx1 since there is a Mutex in the ChunkDistributor which takes care of the synchronization. The queue is used as expected and guarded by a Mutex in the non-spsc case.

I would not complicate the code just catch a misuse.

[elBoberido]

Well, there is one valid use case and we need to decide whether we want to support it. If one could guarantee that no concurrent pushes happen but still multiple producer threads access the SOFI, e.g. by exchanging an access token via a relaxes atomic, the synchronization of m_data is required. The question is whether this should be the task of the SOFI of the task of the mechanism which ensures that there is no concurrent access of multiple producer/consumer.

[elBoberido]

@elfenpiff @budrus do you agree to close this issue. From my point of view this is not a bug since the SpscSofi is, as the name suggests, a single producer single consumer queue. If used in a multi pusher context, it needs to be guarded with a mutex (or different synchronization mechanism), like it is done in the ChunkDistributor where a mutex is used to synchronize the access to the data.