interprocess::message_queue priority forces linear insert time

[defaultryan@…]

interprocess::message_queue uses a priority queue for ordering messages, which is fine, except that many applications don't have any use for the priority feature, and would rather use a straight FIFO queue.

The reason this matters is that when using messages with all the same priority, each addition to the queue requires a linear complexity insertion.

Seems that there should be an option to not provide a message priority, and let the message queue use something more like a ring/circular buffer with low overhead and constant insertion time.

[Ion Gaztañaga]

I've just committed (revision 79898) an implementation that uses a circular buffer as index and optimizes common lookups for uses cases without priority usage. It should be much faster.

[Ion Gaztañaga]

Fixed in Boost 1.52.