infinite waiting

[Yanjiachan]

When using a message_queue, if a process exits or crashes and then restarts, it can cause another process to enter into an infinite wait state.
waiting at:
p_hdr->m_cond_send.notify_one();

[igaztanaga]

This is a known issue, if your platform does not support robust synchronization primitives there is no possible correct emulation. You can define BOOST_INTERPROCESS_ENABLE_TIMEOUT_WHEN_LOCKING, so that waits longer than BOOST_INTERPROCESS_TIMEOUT_WHEN_LOCKING_DURATION_MS (10.000 milliseconds by default) will return error.

What platform are you using? Windows?

[Yanjiachan]

Thanks for your reply, I am using message_queue on linux.
Is there any relevant documentation about these macro definitions?

[igaztanaga]

Sadkly, there is no public information that's why I will left this bug open until I document it. In any case, since you are using Linux, starting from Boost 1.78, the implementation should use Robust Mutexes (see #65) , so you should not have these infinite loop problems. Which Boost version are you using?