Michael Hope wrote:
> > > I meant getting the window handle directly from the innards of Java. The
> > > current program has some code ripped directly from xprop which searches by
> > > name, but it takes a good 3-4 sec on a remote X server.
> >
> > Could part of the time issue be due to network latency?
>
> I'm not actually that sure. The time increases with the load on the
> server (well, duh!), but it is always at least two seconds. The network
> is a bog standard 10baseT, and the number of clients running has little
> effect on it. According to xnetload, the average transfer from the server
> is about 50kB/s, so thats not it directly. Comparing the time of a native
> C program changing another existing window versus a linked in JNI changing
> a Java window near its time of creation would be an interesting test...
Latency is a good suspect -- it's certainly not the bandwidth that's killing
you. There's a lot of two-way X protocol traffic involved in the sort of query
you're doing, and a little latency can go a long way :-(. I suspect you're also
performing the time-consuming task of opening a new connection, since there's
probably no way for you to get the (Display *) pointer from Java. That may be
the worst performance sink of all.
Unfortunately, I've been poking around and looking at the peer classe APIs and
haven't seen any way to get your hands on the window ID.
Nathan
---------------------------------------------
This message was sent using Endymion MailMan.
http://www.endymion.com/products/mailman/
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
