oester+@rchland.ibm.com (Bob Oesterlin) (03/21/90)
We're currently try to track down a core leak in Xibm, that seems to be related to backing store. If you run Andrew console , have backing store enabled, and the screen blanks, the server will quickly grow to a great size. Console seems to be doing a large number of "set clipped region", and for some reason this call will cause the server to grow when the window blanks (or another window overlays the console window). I'm not saying that it's ATK's fault, but what ever console is doing certainly makes it worse. If we don't run console, the server stays (about) the same size over long periods. Bob PS: My X server that I started yesterday was about 1 meg in size; now it's close to 4 meg.
tom@ICASE.EDU (Tom Crockett) (03/21/90)
Excerpts from info-andrew: 20-Mar-90 Memory leaks, etc Bob Oesterlin@rchland.ib (701) > We're currently try to track down a core leak in Xibm, that seems to be > related to backing store. If you run Andrew console , have backing > store enabled, and the screen blanks, the server will quickly grow to a > great size. ... > If we don't run console, the server stays > (about) the same size over long periods. This is the same problem I have been having with Xsun and xlock. I just got the following messages from Bob Scheifler at MIT this morning. I tried it out, and it seems to work just fine on our Sun 3's and 4's. No more leaks. There will be an official patch coming out for this at some point. Excerpts from mail: 20-Mar-90 Re: server leak under xlock Bob Scheifler@expo.lcs.m (160) > Try removing all of the code in the body of mfbCopyGC, cfbCopyGC, > apa16CopyGC, rebuild your server, see if it still works, and see if > the memory leak is gone. Excerpts from mail: 21-Mar-90 Re: server leak under xlock Bob Scheifler@expo.lcs.m (185) > CopyGC was repeating work already done by CopyClip, resulting a region > being discarded rather than freed. -Tom