Doing an RC release
hp at redhat.com
Mon Sep 11 09:34:53 PDT 2006
John (J5) Palmieri wrote:
> The second is the add a new return code from dbus_connection_dispatch()
> to avoid deadlock issues. We pretty much punted this but left it on the
> TODO just in case someone wanted to address it. I personally don't
> think it is worth holding up 1.0. A future fix could be setting a flag
> that says a binding can handle IN_PROGRESS. If this flag is set
> dispatch will be able to deffer a dispatch is one is already in progress
> (i.e. recursive dispatch). In other words this can be fixed post 1.0.
I changed my proposal on this though; at the end of the thread I had
decided just having recursive mutexes was right. See the second part of
the todo item.
A simple fix for 1.0 might be to add recursive mutex funcs to the thread
vtable, allow them to be optional using the flags stuff, and then within
the 1.0.x series we could fix this bug by using recursive mutexes when
they are available. When a binding doesn't provide recursive mutexes
we'll either implement them on top of the nonrecursive, or just leave
the bug there for that binding.
More information about the dbus