Re: _NET_WM_USER_TIME reconsidered



Elijah Newren wrote:

I don't think having the WM being woken up on every keypress is all
that big of a problem.  What I'm worried about is having lots of other
apps woken up on every keypress.  We have a library called libwnck
that is seeing strongly increased usage (both directly and through
e.g. python language bindings).  libwnck can be used to build
full-blown pager-like applications, but is also useful for many other
cases.  Since libwnck watches toplevel windows to get its job done
(e.g. needs to know hidden state changes, icon changes, name changes,
etc.) every process using this library will be woken up on every
keypress.  The number of context switches is simply going to increase
over time as more and more apps use this library or others like it.

If people are just setting a property or something else small, using libwnck is kind of insane because it will do the whole "mirror and monitor complete state of all toplevel windows" song and dance whether you use all that state or not. The original intent was to have just one instance of libwnck really, shared by all the pager-like apps in the panel process.

i.e. for libwnck, context switches are likely only a small part of the problem ;-) this is NOT a library intended to be used by general purpose apps, because what it does (monitor every app on the system) is a crazy thing for most apps to do.

Havoc




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]