Re: XFree dramatic memory leak?
On Wed, 2003-09-10 at 14:04, Michel Dänzer wrote:
> On Tue, 2003-09-09 at 07:54, Soeren Sonnenburg wrote:
> > On Tue, 2003-09-09 at 01:47, Michel Dänzer wrote:
> > > On Mon, 2003-09-08 at 06:52, Soeren Sonnenburg wrote:
> > > >
> > > > I just realized that the daily updatedb run caused this 1G Ti 15"
> > > > powerbook to swap (with just some apps open under X).
Ok, finally the problem reoccured and I found one client which seems to
be responsible (taking 940MB as pixmaps) :
Client 12 (base = 0x1a00000, mask = 0x1fffff): 7 resource types
WINDOW: 4
PIXMAP: 44 (940768384 bytes)
GC: 21
FONT: 7
CURSOR: 33
Unregistered resource 20: 2
Unregistered resource 34: 1
however the last action I did was popping up an xmms playlist window
(which never appeared, but xmms quit instead)... then I found way to
large window sizes in the .xmms/config file for that playlist window...
so I don't know whether what I saw beforehand was a cause of this. So I
better take back everything I said about memoryleaks until I see the
problem reoccur without xmms. Sorry about that false report.
However how can I find out who client 12 is ?
> > > BTW, 2.6 kernels handle this much better than 2.4 kernels. :)
> >
> > But these don't happen to boot already on ppc machines right ?
>
> But sure they do. :) I'm using the linuxppc-2.5 tree via rsync, there's
> also linuxppc-2.5-benh if you don't mind using bk. Beware that there are
> still many minor problems, but the only major problem for me is that
> sleep doesn't work with the linuxppc-2.5 tree yet.
it looks like I will also jump on the 2.5/2.6-test track as soon as I
find the time.
> You should probably ask the following questions on an XFree86 mailing
> list, but I'll do my best to answer some nonetheless.
>
> > BTW, is there a way to limit the movement of the mouse cursor, i.e. I
> > want to limit it when I am disconnected from the second head to :0 and
> > then when reconnected allow the full width of :0 and :1 ?
>
> There's no possibility to do this in the server yet, but it might be
> possible to achieve this with a client to some extent.
that client would need to hang in the message queue and discard
movements over the border... hmmhh could work :-)
[moving windows from :0 to :1 and vice versa]
> Not transparently to clients, as non-Xinerama heads are basically
> separate displays, but as I said earlier, GTK+ as of version 2.2 can do
> this, though unfortunately there's no app-independent support for it
> yet, see e.g.
> http://mail.gnome.org/archives/wm-spec-list/2003-May/msg00029.html,
> http://mail.gnome.org/archives/gtk-devel-list/2002-November/msg00103.html and http://www.spinics.net/lists/xf-xpert/msg04877.html .
nice links, I will investigate that.
Soeren.
Reply to: