On Tue, 27 Mar 2012 14:25:12 +1000 David Seikel <onef...@gmail.com> said:
svn update. fixed. see my log. working for me in fb at the moment anyway (keys, mouse - though no mouse pointer - would have to turn it on manually in elementary - not there currently). > I'm developing an embedded app under X, but the real app runs on the > Linux frame buffer. Part of the real hardware pretends to be a USB > keyboard. Some of the fake key events it sends need to be discarded if > they are too old by the time the app gets around to process them. Some > need to be processed no matter how late they are. So I'm relying on > the Evas_Event_Key_Down->timestamp field that evas sends to my > callback. The idea is to compare that to the current time, and discard > it if it's too old, if it's one of the ones that need to be filtered > for age. > > From what I can tell, frame buffer sets timestamp to ecore_time_get(), > which is a double, the number of seconds since some undefined point in > time. Though the timestamp is an unsigned int, so that gets implicitly > rounded or truncated in whatever manner the compiler decides. > > X stuffs XKeyEvent->time into timestamp, which looks like microseconds > as an integer. Orders of magnitude bigger than what ecore_time_get() > returns. I dunno if it's using the same time base though. My tests so > far seem to show it is the same time base. > > This presents two problems - I need to write two versions of "get > current time", one for X, and one for frame buffer, so that it will > have the correct number for the comparison. Frame buffer version can > use ecore_time_get(), but I dunno where X get's it's timestamp from. > > I want the code to have as little special casing as possible, so prefer > the X and frame buffer code paths to be identical, or only as little > difference as possible. > > The second problem is that the frame buffer timestamp only has a > resolution of whole seconds, which is not good. Microseconds is good. > > Now I understand why the X timestamp is being used, it's best if that > timestamp is as close to the real time the key was actually pressed, > for exactly the reasons I need, filtering out old key events. X is > supplying that, so it's a bit closer to the actual event time than what > we can guess by the time it gets to us. I guess the frame buffer does > not supply a similar timestamp, so we have to generate one when we can. > What I don't understand is why X is microseconds and frame buffer is > seconds? > > Can we change frame buffer to multiply ecore_time_get() by 1000 before > it rounds it by stuffing it into an int? And any other similar input > type that uses ecore_time_get(). That will make things more consistent > for people like me that are writing code that has to work the same > across different canvas types. As a bonus, a more useful resolution > for frame buffer key event timings. > > -- > A big old stinking pile of genius that no one wants > coz there are too many silver coated monkeys in the world. -- ------------- Codito, ergo sum - "I code, therefore I am" -------------- The Rasterman (Carsten Haitzler) ras...@rasterman.com ------------------------------------------------------------------------------ This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure _______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
