Doing something bad to X.org is a good thing. uriel
On Thu, Apr 8, 2010 at 9:26 PM, Benoit T <benoit.triq...@gmail.com> wrote: > Hello, > > Debian testing has replaced Xorg 1.7.5 with 1.7.6 about 3 days ago. > Because i had not logged out of my X session in between, i picked up the > new X server only today. > > dwm does something very wrong to it, causing it to consume a lot of cpu > (say 25-50% according to top with 3s interval) and making X clients > sluggish to unusable. > > I would not bet my life that the effect was not there with 1.7.5, but > I had noticed nothing at all, I mean it's only today that firefox got > very more sluggish than usual. > > Experimental evidence: > - It is clearly dwm and not another client: the effect is there even when > no other clients run but a lame xterm with top or ps inside. > - It happens not only with my personal build and patches, but with just > about every revision, from tag 0.1 to pristine tip. > - I disabled the debug traces we get in the latest revisions from hg, to > no avail (btw. thanks for nothing for leaving that on, i had not noticed > how huge my .xsession-errors had grown) > - It does not happen with any other WM i had on hand (icewm, twm, olvwm, > mwm) > > At this point I don't know if it has anything to do with the graphics > driver (intel in X, i915 in kernel) as I have only my laptop at hand. I > shall try on my server later tonight. > Don't know if that has anything to do either, but with the most recent > 2.6.32 kernel in debian testing, I had to disable kernel mode setting in > i915 as it was a cause of memory corruption after resume from disk. > Because I did not bother to reenable a vga= argument in Grub, I have > been running consoles in real text mode for the last few weeks. Oh, and > X.org locks up the display from time to time when switching between X > and consoles, anyway. > > At any rate there is no clear indication that there is a new frank bug > in X.org, rather there must be something that dwm does or quite possibly > does not do (wrt. other WMs) on the X wire, that keeps X.org busy. At > least, the X.org folks say that when the X server eats a lot of cpu > time, it's not necessarily an indication of an X.org bug (Uriel, *shut > up*) but rather that some X clients bomb the server with more or less > appropriate requests. What's for sure is that I am not embarking on a > gdb/profiling session of the X.org process on my own - but advice and > procedures are welcome. > > I will also try another bunch of tiling WMs among those readily > available in debian testing, including dwm derivatives. We'll see who > triggers the "bug". > > In the meantime i have no other option but to revert to icewm, my > former favorite, and, man, does it suck compared with dwm... > Please help me bring the arch-WM back to its senses !!! ;-) > > Cheers > > -- > Benoit Triquet <benoit.triquet at gmail.com> > .''`. > : :' : We are debian.org. Lower your prices, surrender your code. > `. `' We will add your hardware and software distinctiveness to > `- our own. Resistance is futile. > >
