On Sun, 4 Nov 2001, Justin A wrote: > Hmm. It was better after I restarted X a few times. It was at 20M or > so for a while, but by the end of the day around 60M. I was away from > my computer for 2 hours or so and when I got back it was locked up(sysrq > worked) > > I disabled dpms(caused problems with nvidia driver before?) and am > running that while loop outputting to a file now.
Don't run these tests with the "nvidia" driver. It's too complicated and the Matrix of driver versions to server versions is too large. I'm only aware of one big memory leak in <= 4.1, that is XAA could leak some pixmaps that got purged from the pixmap cache. This would only be triggered by the "nv" driver if you were using Xv and the pixmap cache was full so Xv would have to purge it to make room for the YUV buffers. The "nvidia" driver works around this bug in some later versions, and earlier versions before the fix were caused alot of problems by the XAA bug because the driver could purge to make room for 3D buffers. If you are using the "nv" driver and not using Xv (or are using XFree86 CVS where the pixmap cache leak is fixed) there are no known leaks. > > If anything, gaim is the program causing the leak(budy icons)....that is > after using gaim for a bit the memory usage goes up. Closing gaim stops > the usage from going up, but it doesn't go back down. > > Is there any way to get a better breakdown of what the memory used by X > is made up of? I know how to hack the server to print out resouces left unfreed by the client when the client dies. I'm wondering if we should make an option to have the server print out that information. I suspect it would be useful to alot of developers if the X-server could tell them whenever the exiting client had alot of outstanding resources. Mark. _______________________________________________ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert