View previous topic :: View next topic |
Author |
Message |
jackgentoo n00b

Joined: 04 Apr 2003 Posts: 40 Location: Santa Barbara, CA, USA
|
Posted: Sun Dec 28, 2014 6:49 pm Post subject: KDE input hang |
|
|
I've recently rebuilt my system from scratch. X works, fvwm works, but kde hangs after it starts initializing. What I mean is that the KDE startup icons show up, the panel shows up at the bottom, but as the startup icons fade everything stops. The mouse moves but doesn't change pointer, and nothing responds to clicks. It was working before I enabled the nouveaux kernel module, albeit with one monitor. With the nouveaux kernel module it starts up as though it is properly dealing with two monitors, but the hang I mention occurs. With just bare X the two monitors show the same single screen.
Normally I know how to debug things like this. But I see nothing in my Xorg logs or at the console that give me much clue. I've verified that without KDE I can run glxinfo and glxgears with no errors, so it doesn't seem to be an acceleration issue.
What would you do to debug this? |
|
Back to top |
|
 |
steveL Watchman

Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Mon Dec 29, 2014 3:00 pm Post subject: |
|
|
You've likely tried it, but startkde from command-line, kill it and check dmesg. Then maybe set -x in any startup files, eg if you're using .xinitrc (paired with set +x if you're getting too much, though that shouldn't be an issue in xinitrc.)
It's a bit of a cop-out, but I've always found nvidia cards work best with the proprietary drivers, so just use them (and no vga console, as I'd rather not worry about any clashes.) Though I'd try #gentoo@chat.freenode.net as well, as there's a few people who use nouveau on IRC. |
|
Back to top |
|
 |
jackgentoo n00b

Joined: 04 Apr 2003 Posts: 40 Location: Santa Barbara, CA, USA
|
Posted: Mon Dec 29, 2014 8:47 pm Post subject: |
|
|
I'd love to skip the nouveau drivers. But I have an older NVidia card not supported by the latest binary drivers. I've tried masking the newer drivers, but somehow xorg-drivers insists that I need a 300 series nvidia driver, though I haven't figured out why yet:
Quote: |
# /etc/portage/package.mask
>x11-drivers/nvidia-drivers-174
|
Quote: |
The following mask changes are necessary to proceed:
(see "package.unmask" in the portage(5) man page for more details)
# required by x11-base/xorg-drivers-1.15[video_cards_nvidia]
# required by @selected
# required by @world (argument)
# /etc/portage/package.mask:
=x11-drivers/nvidia-drivers-343.36
|
|
|
Back to top |
|
 |
ct85711 Veteran

Joined: 27 Sep 2005 Posts: 1791
|
Posted: Tue Dec 30, 2014 12:25 am Post subject: |
|
|
the easiest way to find out why a certain package is being pulled in, is run Code: | equery d package_name |
This also works, to help figure out why a certain version is being pulled in.
A quick check on my system, these packages depend on a newer version of nvidia_drivers (based on dependencies will pull in 300 series or newer).
Code: |
virtual/opencl-0-r2 (video_cards_nvidia ? >=x11-drivers/nvidia-drivers-290.10-r2)
virtual/opencl-0-r3 (video_cards_nvidia ? >=x11-drivers/nvidia-drivers-290.10-r2)
virtual/opencl-0-r4 (video_cards_nvidia ? >=x11-drivers/nvidia-drivers-290.10-r2)
sci-misc/boinc-7.2.0 (cuda ? >=x11-drivers/nvidia-drivers-180.22)
sci-misc/boinc-7.2.0-r1 (cuda ? >=x11-drivers/nvidia-drivers-180.22)
sys-cluster/torque-4.1.7 (nvidia ? >=x11-drivers/nvidia-drivers-275)
sys-cluster/torque-4.2.9-r2 (nvidia ? >=x11-drivers/nvidia-drivers-275)
media-tv/xbmc-11.0 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-12.1-r1 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-12.2-r1 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-12.3 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-13.0_p2 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-13.1 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-13.2 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-tv/xbmc-9999 (vdpau ? >=x11-drivers/nvidia-drivers-180.51)
media-video/nvidia-settings-331.20 (=x11-drivers/nvidia-drivers-3*)
media-video/nvidia-settings-340.58 (=x11-drivers/nvidia-drivers-3*)
media-video/nvidia-settings-343.22 (=x11-drivers/nvidia-drivers-3*)
dev-util/nvidia-cuda-sdk-6.5.14 (examples ? >=x11-drivers/nvidia-drivers-340.32[uvm])
dev-util/nvidia-cuda-sdk-6.5.19 (examples ? >=x11-drivers/nvidia-drivers-343.22[uvm])
dev-util/nvidia-cuda-toolkit-6.5.14 (>=x11-drivers/nvidia-drivers-340.32[uvm])
dev-util/nvidia-cuda-toolkit-6.5.19 (>=x11-drivers/nvidia-drivers-343.22[uvm])
app-crypt/oclhashcat-lite-bin-0.15 (video_cards_nvidia ? >=x11-drivers/nvidia-drivers-310.32)
app-crypt/oclhashcat-plus-bin-0.15 (video_cards_nvidia ? >=x11-drivers/nvidia-drivers-319.37) |
|
|
Back to top |
|
 |
jackgentoo n00b

Joined: 04 Apr 2003 Posts: 40 Location: Santa Barbara, CA, USA
|
Posted: Tue Dec 30, 2014 1:19 am Post subject: |
|
|
ct85711 wrote: | the easiest way to find out why a certain package is being pulled in, is run Code: | equery d package_name |
|
What was the full command line you used? With just the above it still isn't saying why it needs the 300 series version:
Code: | # equery depends x11-drivers/nvidia-drivers
* These packages depend on x11-drivers/nvidia-drivers:
x11-base/xorg-drivers-1.15 (video_cards_nvidia ? x11-drivers/nvidia-drivers) |
which is what I see in the xorg-drivers-1.15.ebuild file, but no mention of why it won't use the 173 version which isn't masked. |
|
Back to top |
|
 |
ct85711 Veteran

Joined: 27 Sep 2005 Posts: 1791
|
Posted: Tue Dec 30, 2014 6:09 am Post subject: |
|
|
the second part, that had the large list of packages that will pull in 300 series drivers was with the -a flag/switch (all that does is search all packages, including those not installed; default is only those that is installed).
Just wondering, did you happen to compile xorg-drivers with a newer version of the nvidia-drivers before, and not trying to switch to the lower version? If so, it may be a case of portage is seeing a some old information that xorg-drivers was originally built with a newer version and trying to rebuild with that dependency. Typically, when that happens; I usually remove xorg-drivers Code: | emerge -C xorg-drivers | then reinstall xorg-drivers afterwards to start fresh. |
|
Back to top |
|
 |
jackgentoo n00b

Joined: 04 Apr 2003 Posts: 40 Location: Santa Barbara, CA, USA
|
Posted: Wed Dec 31, 2014 5:06 am Post subject: |
|
|
Portage made an interesting choice. The nvidia-drivers version I needed (173 something) was masked by the config, but the one portage wanted was masked by my package.mask. Anyway, unmasking the version I needed got it installed. Well, not really. First I had to back up a couple of kernel versions. Then it just didn't work at all -- blank screen, no VTs, and nothing interesting in the Xorg log or syslog. Argh.
So KDE hangs with nouveau (but leaves the console alive), and with the NVidia drivers I was using prior to doing this reinstall X hangs the console too.
I've been running Gentoo for 11 years. This may be the end. |
|
Back to top |
|
 |
steveL Watchman

Joined: 13 Sep 2006 Posts: 5153 Location: The Peanut Gallery
|
Posted: Wed Dec 31, 2014 6:38 pm Post subject: |
|
|
jackgentoo wrote: | I've been running Gentoo for 11 years. This may be the end. |
Bleh, get on IRC.
You'd only come back in a year or so, fed up with the constant reinstalls and weird dependency errors. ;)
If the machine works with a boot-disk from another distro, run lspci -k there, and check out how they're setup. |
|
Back to top |
|
 |
jackgentoo n00b

Joined: 04 Apr 2003 Posts: 40 Location: Santa Barbara, CA, USA
|
Posted: Wed Dec 31, 2014 6:57 pm Post subject: |
|
|
You are right, the reasons I've run Gentoo will probably bring me back. But it took me less than 15 minutes to bring up Fedora 21 on the same hardware that I struggled with for days just to get X to run again. Yes, I'll suffer with incompatibility that will result in reinstalls and crap, but it will be a different set of problems for a while. |
|
Back to top |
|
 |
|