[SOLVED] Extremely frustrating install / actually just a (few) bug(s)

quequotion · 2014-02-14 04:23:24

I'm having a really hard time getting to a stable, usable desktop.
I've followed the wiki pretty much to the letter, but got caught on a couple of issues it didn't cover along the way (ipheth, fakeRAID). Then I've spent a few days trying to get a graphical desktop.

At first, I hit a wall because of a problem with the nvidia drivers that prevented the module from loading, thus also preventing lightdm and xorg from starting. That was apparently a bug that had been fixed, regressed, and is perhaps fixed again now (resolved by upgrade). **EDIT** Which was the same issue as this year-old forum post and a bug report I can't seem to find again, but with newer versions of nvidia and linux. Not a problem with the latest versions.

Next I made some software choices and set about getting to a usable desktop. I spent a day or two trying out this and that but finally settled on most of the same things I'd been using in Ubuntu before: ibus-anthy, nautilus, gedit, firefox, plank, super-wingpanel, slingshot-launcher, skype, vlc, compton; plus a few new things: indicator-session-pantheon pantheon/indicator-sound pantheon/indicator-datetime.

Most of that seemed to be working, except I can't seem to find any way to convince X to use "jp" for my keyboard layout other than running "setxkblayout jp" from (openbox) autostart (tried all the suggestions i could find in the wiki). **EDIT** Perhaps the "right" way (with lightdm at least) is to have the layout specified in the file /etc/X11/Xkbmap.

Then, suddenly, pulseaudio (which i don't recall installing but i'l take it anyway) stopped working, possibly due to a fixed bug in systemd that applies the wrong permissions to its run directory at reboot, so no more audio. Nautilus can't get authorization to unlock or mount an encrypted drive, or mount an unencrypted drive either. Also lightdm shows a few daemon users, another bug that is supposed to be fixed. **EDIT** which is this bug as mentioned below, but with newer versions of pulseaudio and systemd.

At the next reboot everything was broken. Well, almost everything: lightdm sometimes fails to start but can be restarted with systemctl; no xsessions would load at first but after opening the lightdm config file in nano, and making absolutely no changes, that problem went away; openbox loads but only the desktop menu and gnome-terminal are operational, everything else freezes on open and refuses to repaint.

I also have compiz-bzr installed, but it just crashes the desktop and sends me back to lightdm. **EDIT** installed; not running as my default desktop; not yet anyway. mostly there to verify if problems are with the WM and to prepare for moving to compiz in the future once I have a stable, usable desktop.

**EDIT** Original Title: Extremely frustrating install: every reboot something else is broken.
Programs freezing on load is solved.

Also, sorry for the meandering post. Since I am new to arch, I intended to explain how I got to where I am and I had hoped someone might notice if there was a connection between the things I had done and the problems I was having. There wasn't.

Last edited by quequotion (2014-02-15 03:40:00)

jasonwryan · 2014-02-14 06:51:56

It is difficult to provide any meaningful advice as there are so many unrelated issues in your post. Please see
https://wiki.archlinux.org/index.php/Fo … ow_to_Post

You might want to start by uninstalling some stuff and getting a basic X setup working, before adding in a DM, etc. As for automounting, if you are using a standalone WM, you'll need to set that up yourself, the wiki page has details.

Not an Installation issue, moving to NC...

quequotion · 2014-02-14 11:06:47

yeah, i know its hard to pin down what's wrong because so much is wrong.

i've beed following the guidelines of the wiki all along the way. i had a basic X setup working after getting around the trouble with nvidia, so i proceeded to install desktop software. once i had made a few selections it seemed a good time to reboot.

then i found that X wasn't working anymore, so i tried to find out what was wrong there but nothing appeared to be wrong with X or lightdm and just as mysteriously as they stopped working they are now working again.

I've found ways to work around most of the issues (can open and mount luks drive manually, can run setxkbmap at every reboot, etc) and the ones I haven't aren't urgent (pulseaudio can wait), except for the most recent:

Most graphical applications freeze on open. The only application windows that will respond to input are gnome-terminal, alsamixergui, and openbox's desktop menu. Terminal applications are wokong fne

Because nearly everything freezes immediately, it's hard to diagnosethe problem. What could be happening here? This didn't happen for the first two reboots, but that was before i had lightdm, nvidia, x, and openbox working. Now that those packages are correctly installed and configured, everything else seems to be broken. Since all of the applictions are broken the sme way, it's likely there's a common cause.

Kartious · 2014-02-14 11:17:18

You will need to post outputs if you would like anyone to help you in the most effective manner, hense why Jason linked you how to post properly.

You will need to show logs, configuration files and system information. Using the [ code] tags to display them

x33a · 2014-02-14 11:17:44

It seems a driver issue.

If you are using the binary nvidia driver, try switching to nouveau. Also disable compiz for a while.

Neburski · 2014-02-14 11:52:48

x33a wrote:

It seems a driver issue.
If you are using the binary nvidia driver, try switching to nouveau. Also disable compiz for a while.

Why propose him to switch to the open source driver without first asking the user to post the GFX card(s) and the driver(s) he's using. I rarely see a GFX driver issue being resolved just by switching to the OS driver. Most of the time it's the user not having read the documentation properly and thus misconfigured his setup.

jasonwryan wrote:

You might want to start by uninstalling some stuff and getting a basic X setup working, before adding in a DM, etc.

That is exactly what quequotion needs to do. First get a very basic X setup working before adding additional features. For every feature that's being added quequotion needs to ensure that it is configured and working properly before adding the next feature.

x33a · 2014-02-14 12:05:26

Neburski wrote:

Why propose him to switch to the open source driver without first asking the user to post the GFX card(s) and the driver(s) he's using. I rarely see a GFX driver issue being resolved just by switching to the OS driver. Most of the time it's the user not having read the documentation properly and thus misconfigured his setup.

I am not asking him to switch permanently, just for testing.

Kartious · 2014-02-14 12:24:59

Neburski wrote:

x33a wrote:
It seems a driver issue.
If you are using the binary nvidia driver, try switching to nouveau. Also disable compiz for a while.
Why propose him to switch to the open source driver without first asking the user to post the GFX card(s) and the driver(s) he's using. I rarely see a GFX driver issue being resolved just by switching to the OS driver. Most of the time it's the user not having read the documentation properly and thus misconfigured his setup.

Yes this

The user should post the output of

 lspci -v | grep VGA

or anything of the equivalent ..

rodyaj · 2014-02-14 13:32:42

quequotion wrote:

I'm having a really hard time getting to a stable, usable desktop.
I've followed the wiki pretty much to the letter, but got caught on a couple of issues it didn't cover along the way (ipheth, fakeRAID). Then I've spent a few days trying to get a graphical desktop.
...
Next I made some software choices and set about getting to a usable desktop. I spent a day or two trying out this and that but finally settled on most of the same things I'd been using in Ubuntu before: ibus-anthy, nautilus, gedit, firefox, plank, super-wingpanel, slingshot-launcher, skype, vlc, compton; plus a few new things: indicator-session-pantheon pantheon/indicator-sound pantheon/indicator-datetime.

None of this is relevant information towards helping us fix your problems.

Most of that seemed to be working, except I can't seem to find any way to convince X to use "jp" for my keyboard layout other than running "setxkblayout jp" from (openbox) autostart (tried all the suggestions i could find in the wiki).

Make a 10-keyboard.conf as documented in: Using X configuration files.

Then, suddenly, pulseaudio (which i don't recall installing but i'l take it anyway) stopped working, possibly due to a fixed bug in systemd that applies the wrong permissions to its run directory at reboot, so no more audio.

If a misbehaving app has messed with the ownership of /run/user sub-directories then your pulse-audio server may not start. Running pulseaudio -D will set the ownership back to 'user' for pulse-audio and allow it to start.

Nautilus can't get authorization to unlock or mount an encrypted drive, or mount an unencrypted drive either.

A lot of file managers need an authentication agent to deal with permissions. Desktop environments like Gnome set this up automatically, but stand-alone window managers such as Openbox don't. Installing polkit-gnome might be enough to fix this issue, but also check what 'optional dependencies' pacman lists when you install nautilus.

Also lightdm shows a few daemon users, another bug that is supposed to be fixed.

Linking to the relevant bug reports would help.

At the next reboot everything was broken. Well, almost everything: lightdm sometimes fails to start but can be restarted with systemctl

When dealing with systemd there are many ways to check for errors. Off the top of my head:

sudo journalctl -xn # show recent errors
sudo systemctl --state=failed # show list of services that failed to start
sudo journalctl -u # show errors from specific service

I would do a 'journalctl -u' on lightdm to get some more output.

no xsessions would load at first but after opening the lightdm config file in nano, and making absolutely no changes, that problem went away; openbox loads but only the desktop menu and gnome-terminal are operational, everything else freezes on open and refuses to repaint. I also have compiz-bzr installed, but it just crashes the desktop and sends me back to lightdm.

I doubt just opening a file for editing fixed your problems. The fact that compiz crashes and that apps refuse to repaint suggests that your video card is to blame for most of your problems. I would do what x33a suggested:

x33a wrote:

If you are using the binary nvidia driver, try switching to nouveau. Also disable compiz for a while.

Also, does your computer have switchable graphics e.g., between an integrated intel and nvidia?

quequotion wrote:

What could be happening here? This didn't happen for the first two reboots, but that was before i had lightdm, nvidia, x, and openbox working.

Well there is the key: everything was working, so one of those things has broken it. My bet is nvidia, but you could also try an alternative display manager such as slim, or an alternative window manager.

Because nearly everything freezes immediately, it's hard to diagnosethe problem.

Try skipping the display manager by switching to another console with Ctrl-Alt-F2 and logging in. Put a line in ~/.xinitrc that will log X errors to a file:

exec openbox >> ~/.wm-errors 2>&1

Then run 'startx' and let the .wm-errors file populate. Post the output here.

Most graphical applications freeze on open. The only application windows that will respond to input are gnome-terminal, alsamixergui, and openbox's desktop menu. Terminal applications are wokong fne

Launch the graphical apps from a terminal so that when they freeze you can see any error messages in the terminal.

If that doesn't work then you need to do this:

jasonwryan wrote:

... start by uninstalling some stuff and getting a basic X setup working, before adding in a DM, etc.

Last edited by rodyaj (2014-02-14 14:27:46)

Neburski · 2014-02-14 16:38:54

rodyaj wrote:

If that doesn't work then you need to do this:
jasonwryan wrote:
... start by uninstalling some stuff and getting a basic X setup working, before adding in a DM, etc.

No, that is exactly what needs to be done first. He should _not_ try out the open source drivers, nor another DM, nor another WM/DE, ....
He first needs to identify what the root of the issue is before taking action.

@quequotion: first disable lightdm, compiz and whatever other fancy packages you are using. Then it would be nice to know what GFX card(s) and drivers you have (see the comment by Kartious). Also the contents of his Xorg log would be good and if you have any configuration files in /etc/X11/xorg.conf.d please post them as well.

You should also check if your GFX card(s) are supported by the driver that you are using.

You mentioned that you had issues with installing the nvidia drivers, you need to be more specific about those issues and if you claim there was a bug involved then you need to provide a link to the bug report.

quequotion · 2014-02-14 19:06:44

It's this: https://bugzilla.mozilla.org/show_bug.cgi?id=833117

I wasn't getting consistent error output (and most of the time no output), not to mention having no means of copying and pasting from the disabled computer's screen to my phone (all my earlier posts in this thread were typed on a phone) so providing helpful logging was a bit of a challenge. I suppose i could have tried to access the forums with lynx.

Apparently this bug can affect any program that relies on glib, not only firefox, and the symptoms are programs that open and immediately freeze.

workaround;

export G_SLICE=always-malloc

I think I'll add this to .bashrc for now.

Nevermind, that only worked once.

Since I wasn't getting anywhere with this problem, I went back to working on pulseaudio and made an interesting discovery: everything works fine when pulseaudio isn't running.

So now I'm considering this: https://bugzilla.redhat.com/show_bug.cgi?id=753882 as the actual culprit (same bug as the launchpad pulseaudio/systemd bug, redhat version cites pam_systemd specifically).

Last edited by quequotion (2014-02-16 04:47:00)

rodyaj · 2014-02-14 19:53:34

Neburski wrote:

rodyaj wrote:
If that doesn't work then you need to do this:
jasonwryan wrote:
... start by uninstalling some stuff and getting a basic X setup working, before adding in a DM, etc.
No, that is exactly what needs to be done first. He should _not_ try out the open source drivers, nor another DM, nor another WM/DE, ....

No, the first thing that needs to be done is a cursory look at the log files for anything immediately telling. If there is nothing obvious, then all quequotion will lose by following my advice is the time it takes to reinstall/uninstall a couple packages and reboot. I'd rather first quickly swap two low-level things well known to cause breakage (the WM and graphics driver) than the drastic measures you advocate of removing multiple software and configs to get back to a base X system (only to start trawling Xorg log files). If the replacement pacakges fix things, then nothing stops quequotion from more in depth bug tracking to get the first choice driver and WM working.

He first needs to identify what the root of the issue is before taking action.

Why does there have to be a "root" cause. From the looks of it quequotion has a plethora of problems. Ultimately, you might be right that starting from scratch will help trace the problems, but it could just as easily send quequotion right back to square one if there is more than one app, or a certain combination of apps causing conflicts. I'm simply suggesting that if there is nothing immediately useful in the logs, then looking to the usual suspects and temporaily replacing the core things such as drivers (which are well known to cause the types of glitching and freezes the OP describes) could save a lot of time uninstalling and debugging everything else.

And I'm just going by my own experience here; countless times my fresh install problems were solved by just changing the driver or going back to a stable version of a window manager.

Last edited by rodyaj (2014-02-16 03:37:02)

quequotion · 2014-02-14 20:06:55

rodyaj wrote:

None of this is relevant information towards helping us fix your problems.

I really hoped it would be, such that someone might notice if I had installed anything known to cause problems such as this. Apparently not.

Make a 10-keyboard.conf as documented in: Using X configuration files.

Done that. I've tried all the methods in the wiki. My xorg log even shows this particular one taking effect, but it doesn't actually work. None of the other methods do either. There's another I'm still looking into: lightdm has its own built-in calll to setxkbmap, in /etc/lightdm/Xsession, which is looking for a setting in /etc/X11/Xkbmap

If a misbehaving app has messed with the ownership of /run/user sub-directories then your pulse-audio server may not start. Running pulseaudio -D will set the ownership back to 'user' for pulse-audio and allow it to start.

Still need to find out why this is happening (every reboot). I read this bug report where it was traced back to systemd, but that was supposedly fixed.
**EDIT** This is very likey the cause of most of my troubles; as it seems the fix was only applied to the ubuntu package and systemd's git. I am a little weary of installing systemd-git to find out if it's fixed there, but there is a commit in place.

A lot of file managers need an authentication agent to deal with permissions. Desktop environments like Gnome set this up automatically, but stand-alone window managers such as Openbox don't. Installing polkit-gnome might be enough to fix this issue, but also check what 'optional dependencies' pacman lists when you install nautilus.

I'll give that a shot; sounds right.

Linking to the relevant bug reports would help.

Yes, but very cumbersome on a phone, particularly on an iPhone with its "innovative" copy-and-paste mechanism... anyway, I'll probably be able to fix this one on my own when i get around to it. the fix was something like making sure user numbers under 1000 are hidden and if i recall it's just a configuration option. I also kind of want to see if the daemon users will appear as actual daemons if i use crowd-greeter...

When dealing with systemd there are many ways to check for errors. Off the top of my head:
sudo journalctl -xn # show recent errors
sudo systemctl --state=failed # show list of services that failed to start
sudo journalctl -u # show errors from specific service
I would do a 'journalctl -u' on lightdm to get some more output.

yeah, I'm a complete newb on systemd. Thanks for the advice; I'm sure I'll be making use of it somewhere before the day is done.

I doubt just opening a file for editing fixed your problems. The fact that compiz crashes and that apps refuse to repaint suggests that your video card is to blame for most of your problems. I would do what x33a suggested

I'm not saying it makes sense, just that I can't think of another explanation based on what I recall. Also, I said compiz is installed, not that I'm using it. I intend to, once I have a stable desktop running, but I've been working almost exclusively with openbox so far.

Also, does your computer have switchable graphics e.g., between an integrated intel and nvidia?

No, custom-built desktop here. I am curious though if I could create some kind of relationship between the i7's integrated GPU and the GT630 pci-e GPU, but that's an experiment for a later time.

Well there is the key: everything was working, so one of those things has broken it. My bet is nvidia, but you could also try an alternative display manager such as slim, or an alternative window manager.

Well, you never really know in OSS eh? Everything was working, and none of those things had broken anything, but something completely under the radar had. Maybe glib had been upgraded during the last xsession; until today it wasn't on my watch-out-for-upgrades list. **EDIT** I also suspect the problem could have been caused by firefox, hypothetically considering that if firefox's memory allocation were broken then it could break allocation for any currently or subsequently opened programs. I have not yet explicitly tested opening programs in a variety of different orders to verify this theory.

Try skipping the display manager by switching to another console with Ctrl-Alt-F2 and logging in. Put a line in ~/.xinitrc that will log X errors to a file:
exec openbox >> ~/.wm-errors 2>&1
Then run 'startx' and let the .wm-errors file populate. Post the output here.

Next time I find myself with nothing but ttys to work with I will give this a shot.

Launch the graphical apps from a terminal so that when they freeze you can see any error messages in the terminal.

This wasn't producing consistent or reliable output, but in the end it was how I was able to find a workaround. I also noticed I have a tendancy to disregard Glib-Blah-Blah-Blah errors, probably because most of them are harmless. This one was not (though some users have reported seeing this error and having no trouble).

If that doesn't work then you need to do this

I did, like mad, but it wasn't the right thing to do. The packages were all installed correctly the first time and configured correctly. It's not my first time to work with most of these packages. The problem was this bug (more recent report): https://bugzilla.mozilla.org/show_bug.cgi?id=833117

Last edited by quequotion (2014-02-16 05:00:06)

quequotion · 2014-02-14 20:18:59

Neburski wrote:

disable lightdm, compiz and whatever other fancy packages you are using.

Perhaps the arch froums are not the best place to mention the word "compiz". It's not a "fancy package", none of them are. Just because the console can do what ever I want doesn't mean I want to do everything by console. Anyway, you do have a point and this is also usually how I do things:

1. Try something
2. Find problem
3. Try again.
4. Confirm consistent problem.
5. Reconfigure
6. Confim reconfiguration does/n't resolve problem
7. Reinstall
8. Confirm reinstall does/n't resolve problem
9. Research known unknowns (google)
10. Attempt suggested workarounds or fixes if available
11. Confirm workarounds or fixes do/n't resolve problem
12. Research uknown unknowns (forum post)
13. Try alternatives.

**EDIT** That said, I don't always do things strictly in this order. One of the reasons I have compiz installed is to have an alternative WM to test if openbox is causing problems.

Last edited by quequotion (2014-02-15 02:49:52)

quequotion · 2014-02-15 02:54:14

The arch forums are really refreshing after spending years in ubuntu's. Users who know what they're talking about actually respond and give helpful advice here!

rodyaj · 2014-02-15 13:22:05

quequotion wrote:

rodyaj wrote:
None of this is relevant information towards helping us fix your problems.
I really hoped it would be, such that someone might notice if I had installed anything known to cause problems such as this. Apparently not.

Yes I can see now that you were just trying to be helpful.

quequotion wrote:

Make a 10-keyboard.conf as documented in: Using X configuration files.
Done that. I've tried all the methods in the wiki. My xorg log even shows this particular one taking effect, but it doesn't actually work. None of the other methods do either. There's another I'm still looking into: lightdm has its own built-in calll to setxkbmap, in /etc/lightdm/Xsession, which is looking for a setting in /etc/X11/Xkbmap

To tell the truth, I've just set my kemap through ~/.xinitrc for a long time now, and it saves worrying about config files becoming "depracted" (thinking back to HAL to dbus). There are still far too many different ways to interact with Xorg, and it makes it confusing to use.

quequotion wrote:

rodyaj wrote:
Linking to the relevant bug reports would help.
Yes, but very cumbersome on a phone, particularly on an iPhone with its "innovative" copy-and-paste mechanism...

Fair enough.

quequotion wrote:

I also suspect the problem could have been caused by firefox, hypothetically considering that if firefox's memory allocation were broken then it could break allocation for any currently or subsequently opened programs. I have not yet explicitly tested opening programs in a variety of different orders to verify this theory.

You could be on to something. My window manager (Sawfish) has been locking up when Firefox is running the last week or so. I need to do some experiments and look at that export line you posted.

Last edited by rodyaj (2014-02-15 13:25:34)

quequotion · 2014-02-16 05:07:40

rodyaj wrote:

To tell the truth, I've just set my kemap through ~/.xinitrc for a long time now, and it saves worrying about config files becoming "depracted" (thinking back to HAL to dbus). There are still far too many different ways to interact with Xorg, and it makes it confusing to use.

Perhaps one day X itself will be deprecated for something easier to work with....

You could be on to something. My window manager (Sawfish) has been locking up when Firefox is running the last week or so. I need to do some experiments and look at that export line you posted.

I tried opening programs in various orders, but I was not able to distinguish any difference in their behavior. Everything locked up immediately no matter what order they started in. Furthermore, I found that it's more likely a problem with systemd & pulseaudio in my case, but that doesn't rule out the issue with firefox, which is also a more recently reported bug.

rodyaj · 2014-02-16 15:54:04

quequotion wrote:

I tried opening programs in various orders, but I was not able to distinguish any difference in their behavior. Everything locked up immediately no matter what order they started in. Furthermore, I found that it's more likely a problem with systemd & pulseaudio in my case

As you say, it is probably something lower-level such as systemd. It sometimes takes pure luck to open things in the right order that you can notice a relationship. This is why I don't usually bother fresh installing to solve problems, as it won't necessarily help you notice these relations. You could look through the output of dmesg for hardware isssues too. In the past I've had various hardware such as faulty wireless dongles lock up my system.

quequotion wrote:

... but that doesn't rule out the issue with firefox, which is also a more recently reported bug.

When I ran pacman -Syu yesterday I noticed Firefox and glibc upgraded. So far I haven't had any more crashes. Might be worth upgrading.

Last edited by rodyaj (2014-02-16 16:19:05)

Arch Linux

#1 2014-02-14 04:23:24

[SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#2 2014-02-14 06:51:56

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#3 2014-02-14 11:06:47

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#4 2014-02-14 11:17:18

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#5 2014-02-14 11:17:44

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#6 2014-02-14 11:52:48

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#7 2014-02-14 12:05:26

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#8 2014-02-14 12:24:59

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#9 2014-02-14 13:32:42

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#10 2014-02-14 16:38:54

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#11 2014-02-14 19:06:44

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#12 2014-02-14 19:53:34

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#13 2014-02-14 20:06:55

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#14 2014-02-14 20:18:59

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#15 2014-02-15 02:54:14

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#16 2014-02-15 13:22:05

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#17 2014-02-16 05:07:40

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

#18 2014-02-16 15:54:04

Re: [SOLVED] Extremely frustrating install / actually just a (few) bug(s)

Board footer