mir_buffer_stream_release returns an unusable MirWaitHandle

[Chris Halse Rogers]

mir_buffer_stream_release returns a MirWaitHandle that will be freed before it becomes ready. It is unsafe to use the return value of mir_buffer_stream_release in any way.

[Chris Halse Rogers (raof)]

Failing actually fixing our whole WaitHandle infrastructure, stashing BufferStreams' wait handles in the Connection alongside the Surfaces' wait handles is probably the best bet.

[Daniel van Vugt (vanvugt)]

Why would any release function need to give the client a wait handle? It's irrelevant to the client when the release is completed...

[Chris Halse Rogers (raof)]

This is a reasonable question.

At least one answer is that the client may have some per-object state
that it needs to clean up, and it wants guarantees that such state has
been cleaned up before proceeding.

[Chris Halse Rogers (raof)]

Also - it is guaranteed that post release completion the object will
not have any callbacks generated for it. There's no point before
release completion with that guarantee.

[Alan Griffiths (alan-griffiths)]

I've not reviewed the changes over the past two years, but in the current codebase:

1. the wait handle isn't released before becoming ready
2. this function has been deprecated as part of the MirWaitHandle cleanup