mirror of
https://github.com/gryf/wmaker.git
synced 2026-01-02 20:04:15 +01:00
- Added readme about handling ConnectionDidDie or ConnectionDidTimeout
notifications.
This commit is contained in:
@@ -2,5 +2,5 @@
|
||||
|
||||
AUTOMAKE_OPTIONS = no-dependencies
|
||||
|
||||
EXTRA_DIST = wm.png wm.html
|
||||
EXTRA_DIST = wm.png wm.html README.connection
|
||||
|
||||
|
||||
47
WINGs/Documentation/README.connection
Normal file
47
WINGs/Documentation/README.connection
Normal file
@@ -0,0 +1,47 @@
|
||||
Methods of handling WMConnectionDidDieNotification notification events
|
||||
(same for WMConnectionDidTimeoutNotification)
|
||||
----------------------------------------------------------------------
|
||||
|
||||
Once your program got this notification (you need to install an observer for
|
||||
it), there are some ways to handle it:
|
||||
|
||||
1. Make your observer enqueue a new notification in the ASAP queue, and the
|
||||
observer for this new notification (it can be the same function if you
|
||||
arrange to distinguish between the two cases), should remove the connection.
|
||||
You can also close the connection before enqueuing the new notification to
|
||||
the ASAP queue, but is not strictly necessarily, since it will be closed
|
||||
when the observer for the new enqueued notification will be called and you
|
||||
will call the close/remove function there. This is just to make sure your
|
||||
connection will be silent, and won't generate new events until you reach
|
||||
that point.
|
||||
This is by far the best method, since it will assure you that if you
|
||||
enqueue more than one notification to remove the same connection, they will
|
||||
be coalesced, and called only once.
|
||||
2. In your observer, put the died/closed connection in a bag, and destroy all
|
||||
the connections present in the bag, in your main loop, after you call the
|
||||
WHandleEvents()/WMHandleEvent(). Also closing the connection can be done
|
||||
before putting the connection in the bag, but is optional as noted above.
|
||||
In this case you need to make sure you don't put in the bag the same
|
||||
connection more than once, in case the DieNotification is sent more that
|
||||
once to you. This is automagically solved by method 1.
|
||||
3. In case it's your only connection, and you plan to exit if it was closed or
|
||||
died, then you can safely close/remove it, and exit. As long as you no
|
||||
longer access it, there is no problem.
|
||||
|
||||
4. Make you observer remove the connection. Then make sure that after that
|
||||
point your code no longer tries to access that connection (this usually
|
||||
means until your code gets back to the main loop). This is almost always
|
||||
very hard to achive and subject to hidden errors. I do not recommend this
|
||||
way of handling the died notification. It is ugly and very complicated to
|
||||
handle. If you use it and get plenty of SIGSEGVs then you know why.
|
||||
This method was not presented here to be used, but to show what should
|
||||
be avoided in dealing with the died notification, in case someone gets the
|
||||
idea to try it this way.
|
||||
|
||||
|
||||
Note: read/write operations means to use our read/write functions (like
|
||||
WMGetMessage()/WMSendMessage()), not the C library ones read()/write().
|
||||
Note2: removing a connection is done by WMDestroyConnection(), while
|
||||
WMCloseConnection() only closes the socket, and removed any pending
|
||||
queues and timers on the connection.
|
||||
|
||||
Reference in New Issue
Block a user