From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.3) Gecko/20010801 Description of problem: My roswell system has become completely useless. It might very well be the kernel, the VM in 2.4.6 seems to be pretty bad. Basically overnight, after updatedb has run, and disk I/O buffers have 'polluted' memory, linux really doesn't want to give it,resulting in overmassive I/O and incredible slow performance. My P600 laptop has 256M with 512M swap, and it behaves like a 486/100 or worse. First I thought it was sawmill + nautilus, so i switched back to enlightenment. There the menus were all corrupted and not present I should see (e.g. no panel , no desktop icons).So after i figured out the VM problem was not sawfish, wanted to go back to Sawfish. switchdesk itself (the cmdline utility) didn't want to let me choose sawfish (tried various spellings), so tried the gnome version,that one caused X to terminate back to console.... I'm sorry, i hope the final version will be better, I've spent a lot of time this week on this beta, making it my daily machine, but it has become totally useless, and I will switch back to rh62 or 71... How reproducible: Always Steps to Reproduce: 1.switchdesk-gnome 2.select Sawfish, while Enlightnement is active 3.X terminates Additional info:
so, ok, i'm a whimp, i switched back to redhat 7.1, but also learned another thing about this bug: The same problem is present on rh71 !!!! In fact, it's partially a pilot error due to not thinking of the difference between window manager and desktop manager? The bug still remains i believe, if you switch to sawfish, X dies. However, i found if it did switchdesk GNOME on the console, and started X, it was happily back in sawfish. I will now patiently await the nightly crontab to see if memory becomes locked up again. Although this is kernel 2.4.2 (i didn't upgrade that part of rh71 yet). So far however, rh71 has been a blessing. Knock on wood of course.
switchdesk isn't supposed to let you use plain sawfish. Switchdesk supports: KDE GNOME (this uses sawfish) Enlightenment Fvwm Windowmaker Twm switchdesk doesn't do anything that could cause an X server crash (actually it just changes your .Xclients file), so if X is crashing when switchdesk is running, you're either facing an X problem or a problem with your X session (e.g. when you're in enlightenment, your .Xclients file does "exec /usr/bin/enlightenment", so if enlightenment crashes, your X session will terminate. Can't reproduce it here; assigning to XFree86 because it's a possible source of the problem. Please provide more info (e.g. which X driver are you using, do you see any messages when X exits, ...)
I created a new clean account, and could not reproduce it either, so after your info i conclude (despite that I thought I had cleaned up everything) that I forgot one of the .Xclients and for some reason it wound up with the wrong contents. I would also like to report that redhat7.1 survived the night beautifully, it's like day and night (no pun intended) difference. The beta just didn't work for me. I still suspect it's the kernel, the rh71 kernel is after a little earlier and I did hear some trouble about VM, but I never would have thought my P600/256M would behave like a 486 with 32M.... I still wonder why nobody at redhat noticed this behavior. Is nobody using their beta machines with a fair number of windows open and letting it run overnight?
We do; this doesn't happen on every machine though. (and most of us are running a much more recent kernel by now - the post-beta1 changes need testing (check out rawhide).
I should also add another laptop difference between rh71 and roswell! Roswell runs a newer version, 4.1, whereas rh71 runs 4.0.3. Like on rh62 on this laptop (where i ran a patched 3.3.6) coming out of suspend my X display would be ready for use in about 10-15 seconds. Under whistler this was about 40-50 seconds!!!!! I don't know if this is a redhat problem, but I would like to hear if other redhat laptop users have experienced this. 10-15 secs i find ok, but the 40-50 sec wait is very very long. I might as well do suspend to disk, or just reboot for that matter.