Re: Help, help, help!!!

(I approved this, but if you're going to post to gnome-list it does
help to be subscribed, or all your posts are held in the same queue
with the spams and klezes and so on until someone goes through the

On Tue, Jul 30, 2002 at 07:34:35AM +0200 or thereabouts, Damian Dobroczynski wrote:
>  Nautilus is dead & my icons on the desktop gone!
>  But this is only the beginning of the story. Nautilus stopped working after 
>  n-th crash of my GNOME environment. For several months GNOME has been hanging 
>  without warning!!! Verbosely, everything stops (including kernel I think, 
>  because I couldn't reboot my computer via net (telnet) --- tcp is dead!). The 

Okay. This is a good indicator that it's something more than X :) 
When it happens, can you even ping the box?

Kernel crashes are the kernel's fault. Some app may have told it stuff
which confused it, but the kernel is not supposed to get confused and

X crashes and hangs are X's fault. Some app may have told it something
which confused it (Netscape and Xscreensaver are examples I've seen do 
this), but X is not supposed to get confused and hang or crash.

So whatever's going on, it sounds a little more wide-ranging than
Nautilus and GNOME. Nautilus or GNOME may be provoking it, of course,
but this sounds like something else.

>  whole system stops responding. This happens only during _GNOME_ session! KDE 
>  sessions are OK till some program based on Mozilla engine runs. This moment 
>  is very risky because THIS happend several times during KDE session. I can't 
>  remember if I did something special --- THIS may happend any time, any place 
>  (for ex. during clicking on a link, or writing something in Emacs). 
>  Some words about Nautilus: icons are gone, the program launched by hand 
>  starts working and nothing more. Nothing popups, no windows, no messages in 
>  var/log/messages, no messages on terminal (besides "GnomeVFS-WARNING **: 
>  Unable to convert MIME info from UTF-8 to the current locale pl. MIME info 
>  will probably display wrong").

There is a way to tidy up Nautilus stuff. Unhappily, I don't know it.
Someone else here is bound to.
>  Evolution is nearly dead (this may be related to the Nautilus silence): 
>  everything is ok till I browse a file on may IMAP account. Then, the program 
>  stop working. Hangs.

This sounds like a separate bug from the whole thing crashing.
Can you browse an IMAP email with other mailers? (You could try 
mutt: I know that has IMAP support.) 
>  I beg for help. I do not want to reinstall everything (which probably will 
>  not help). Some missing files? After the computer hangs I have to reset it 
>  in  very brutal way... I did it "n" times...

Never a good sign. You mention Mandrake 8.1. Does this have a 
journalling filesystem? What kind of output are you getting on
bootup following this?

>  system info:
>  PC Pentium IV, 1.6 GHz.
>  Mandrake 8.1
>  Ximian-GNOME 1.4.x (the latest)
>  dynamic IP (this does not matter, because when the IP was fixed all of the 
>  calamities happened also)
>  XFree 4.x with GeForce II (NVidia-made driver)

Is this the nVidia binary drivers? Some people have good luck with
these, but I know that some people have met kernel lockups and X
lockups with the binary drivers. I also know that the typical 
response from kernel hackers to a hang with nVidia drivers is
"it's binary software, we can't fix it, and we don't want the
report". Some X hackers feel this way too. So if you have the
binary drivers, you will have to replicate it without their
being loaded before you can get a bug report off and be confident
it will be looked at.

Things I would suggest to narrow it down:

  * If this started happening after you installed something in
particular, remove that something. See what happens.

  * I presume this is your home machine or one you have root
access on. Create another user called 'test' (or something).
Log in as this and mess around with GNOME as this user. Does the
weirdness happen? If not, try the next idea. If it does, it's
probably not something in your usual account's setup.

  * Move (or remove) your directories which have your account's
GNOME defaults in. Start over. If everything appears to work,
start moving things back and see which one causes trouble. The
directories things live in are ~/.gnome/, ~/.gtkrc/, the dotfile
for your window manager (~/.sawfishrc at a guess), and perhaps
a few other ~/.gnomesomethings. In /tmp you might nuke the
contents of /tmp/orbit-username and /tmp/.ICE-unix. If they
belong to your account name, at least.

  * Run rpm -V (or rpm -Vv(v)) on the rpms which provide the
packages you're having trouble with and on the packages on which
they depend. Since you have Ximian, you can get a list of the
relevant GNOME ones with "rpm -qa | grep ximian". That will
spit out the majority of the GNOME packages. Then rpm -V gnome-core,
rpm -V gnome-libs, etc etc. rpm -V verifies the files the rpm
provided are as expected. It only reports errors. If you get
no output, there are no errors. This is good. (You'll find an
explanation of the output in 'man rpm' -- big page, but it is
in there.) 

  * I'm sorry, but I would suggest removing the nVidia binaries
completely and seeing whether it still happens. If you can't/won't
do that, stop them being loaded, shut down (politely, not with the
big button), and restart. Check they're not loaded. Then see whether
it still happens.

  * If you have a second computer and the two are networked, ssh
(or telnet, if you must) into the dodgy one before weirdness 
starts. Run 'top' and leave it running. This draws a constantly
updated picture of what's running on the box and how much CPU
etc is being used by each process. (Be careful with the memory 
numbers though: they can be misleading.) When the weirdness
starts, see whether your telnet session is still active and 
whether X or an X client is spinning in 'top' and eating 90% of
the CPU. When X does this, the machine -looks- like it's crashed,
but in fact it's just X that's stuck: the rest is fine. It's just
too busy attending to X to notice that any other process wants
time to run. It would be useful to know this. It does all sound 
like something more wide-ranging than "Gnome is borked", but best 
to check.

  * In fact, whilst in "rule out the obvious" mode, here's another.
Have you a floppy drive and disk? Find memtest86 and stuff it on
the floppy (you'll need to read the docs on this: I'm summarising
here and forget how I did it). Boot from that and run memtest86
overnight (at least). This checks your memory. The longer you
can run it the better. Half an hour is not enough. My experience
is generally that you run it for hours and hours and it all turns
out to be fine: but at least you know it for sure! 

  * Try grabbing your IMAP mail with a different mail client.
I rather suspect it's a separate issue from all the rest, but
it would be good to confirm it.

There may be Mandrake-specific or Ximian-specific things you
can do too (in fact, if Red Carpet will run, that will probably
verify your rpms for you). But I don't know them.

When you have done at least some of those, post the results :) 


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]