GNOME Bugzilla – Bug 313474
problems with volume handling
Last modified: 2005-12-19 13:03:03 UTC
I have several problems with volume handling: 1) Once that i mount a volume using gnome-volume-manager or by double clicking a drive in the computer window the popup that i get clicking the right button of the mouse still says "mount volume" instead of "umount volume" 2) once that a volume is mounted if i double click on the drive (in the computer window) no window is open for viewing the contents, what i get is an error saying that the volume is already mounting 3) when a volume is mounted (CDROM or USB memory key) it should appear on the desktop, but it does not. I have the /apps/nautilus/volumes_visible set to true. All of this worked ok in gnome 2.10. I'm using a HAL enabled gnome-vfs and HAL and gnome-vfs from CVS HEAD.
This sounds more like an gnome-vfs/HAL issue than somthing in nautilus.
These are all pretty serious problems, especially if they are regressions. I've asked #bugs to try to reproduce; hopefully we can get some more data that way.
I can reproduce point 3) with 2.10 : when I put a USB memory key it doesn't even mount it and with a CDROM I have to go and open manually the cdrom in the computer window to mount it and still it doesn't show on the desktop... very frustrating when I know that it was working about 3 or 4 months ago (don't remember when this started to act this way). I'm on ubuntu hoary 5.04. Just a little idea to find the source of the problem: maybe by looking the bigs changes in the ubuntu package changelog we could see when this all started to happened. Ho and my system is fresh updated with the hoary ubuntu packages.
Diego, just to make sure that we try to reproduce the right behavior, did you compile gnome-vfs to include support for HAL ? Thank you for the information you provided with your bug report !
One more thing, did you make sure that FAM support is included in your gnome-vfs build ? (don't know if its necessary, but something tells me that it should help).
I don't trigger the three problems you mention by using DBUS, HAL and gnome-vfs head. So maybe this is more related to nautilus cvs HEAD ? I'm just building it at the moment, and will test this as soon as possible.
Yes, I have hal support enabled. I'm quite sure of it.
Yes, but note federico's blog post: http://primates.ximian.com/~federico/news-2005-08.html#27 You can test if it is actually connecting to hal (really, everyone in this thread probably should do this) with: grep 'system_bus' ~/.xsession-errors If it returns something like this: ** (gnome-volume-manager:6647): WARNING **: manager.c/1609: hal_initialize failed: Failed to connect to socket /home/louie/jhbuilt//var/run/dbus/system_bus_socket: No such file or directory then you're not actually hitting hal in your vfsd.
Looks like I hit hal in my vfsd since grep 'system_bus' ~/.xsession-errors give me no output.
Using nautilus, gnome-vfs, hal and dbus from cvs (HEAD branch), everything is fine for me, for the three points mentionned in the bug report. I have other GNOME components that are in 2.8.x and 2.10.x versions. Do you think that it can be a problem in this case to reproduce the bug ? Also, could you please tell us more about what could be specific to your configuration (what distribution you use, when was the last time you updated your working copy of nautils, dbus, hal and gnome-vfs CVS, the kernel you use, etc.) ?
I'm seeing this as well, with HAL-enabled gnome-vfs 2.11.92 (from Ubuntu breezy). Some information that may (or may not) help: I'm doing some stuff with GnomeVFSVolumes in Rhythmbox, and the mount/unmount signals don't seem to be getting sent from the volume monitor. As a result gnome_vfs_volume_is_mounted is always returning true for things like my CD drive, even if the drive is empty or the tray open.
I've looked a bit more into this and it appears that gnome-vfs-daemon is somehow getting wedged. If I kill and restart gnome-vfs-daemon the mount/unmount signals start being sent again; although it stops again after a period of time (anywhere from a few minutes to a several hours). Gamin appears to be fine, as Nautilus receives file notifications from it even when the volume ones aren't being sent. If there are debug logs or anything that gnome-vfs-daemon produces that could help, let me know.
Can you get a backtrace from the vfs daemon when its wedged? Just attach to the daemon pid using gdb.
Also using HAL-enabled gnome-vfs 2.11.92 but cannot reproduce this. (debian testing)
I fixed a couple of bugs in my RB code yesterday, and now the vfs daemon isn't getting stuck anymore. It looks as if the problem occurs when an application crashes while in a pre-unmount/unmounted signal handler. I'll go back to the old buggy RB code, and see if I can get a backtrace from a wedged vfs daemon.
James: Did you investigation reveal anything new?
Unfortunately the backtrace doesn't help much:
+ Trace 63027
However after I had a program delibrately crash in the "mounted" signal handler, any program that had handler installed would print the following when the signal would normally be emitted: (1527 is the pid of a process that should have had a handler triggered) 1527: arguments to dbus_message_new_method_call() were incorrect, assertion "path != NULL" failed in file dbus-message.c line 793. This is normally a bug in some application using the D-BUS library. libhal.c 1114 : Couldn't allocate D-BUS message
Created attachment 52263 [details] testcase.c testcase that can be compiled with "gcc -o testcase `pkg-config --cflags --libs gtk+-2.0 gnome-vfs-2.0` testcase.c" When this program is run the first time it will segfault with something is mounted. On sucessive runs (until gnome-vfs-daemon is killed and restarted) it does not receive "mounted" signals, and so doesn't segfault.
I haven't come across this. Can anyone still reproduce?
this has been fixed, at least it doesn't happend any more
Closing this report as per last comment.