I added some debugs of my own from the source code.
Looks like "try_mount" fails, with "server"=NULL and "share"=NULL.
I got the code write out the complete mount_spec:

smb-share:host=pgbulfon,user=gbulfon,prefix=0.000000Users0.000000gbulfon

....code is looking for "server", while mount_spec contains "host".
And what is that prefix???

I suspect that the problem arises because of this:

GLib-GObject-WARNING **: Two different plugins tried to
register 'GVfsUriMapperSmb'

so maybe a wrong object is taken to parse the URI?
How can this happen?

Also, I cannot find any way to run gvfsd and see its stderr.
Even running "/usr/lib/gvfsd -r", it just stays there with no message,
not even trying mount.

Please help me to find a reason for this!

Gabriele

For some reasons I don't know, the error trying mount changed:

Error mounting location: Invalid mount spec

and in fact this is what happens when server and share are NULL...

Is there any way to force glib-gobject to tell me who is trying to register
GVfsUriMapperSmb two times?

(In reply to comment #0)
> (gvfs-mount:1256): GLib-GObject-WARNING **: Two different plugins tried to
> register 'GProxyDrive'.

This indicates a problem where duplicit gvfs modules are loaded. Check e.g. /usr/lib64/gio/modules and look for duplicates. I.e. a mixture of prefixes and installed libraries.

FYI: You could try environment variables (e.g. GVFS_DEBUG, GVFS_SMB_DEBUG) to print some useful debug info, see:
https://wiki.gnome.org/Projects/gvfs/doc

(In reply to comment #4)
> (In reply to comment #0)
> > (gvfs-mount:1256): GLib-GObject-WARNING **: Two different plugins tried to
> > register 'GProxyDrive'.
> 
> This indicates a problem where duplicit gvfs modules are loaded. Check e.g.
> /usr/lib64/gio/modules and look for duplicates. I.e. a mixture of prefixes and
> installed libraries.

Thanks Tomas,
I already investigated this.
At the moment, I'm still just using a 32 build, and this is what I have:

gbulfon@xstream:/usr/lib/gio/modules$ ls
libdconfsettings.so             libgsettingsgconfbackend.a      libgvfsdbus.so
libgioremote-volume-monitor.so  libgsettingsgconfbackend.so

gbulfon@xstream:/usr/lib/amd64/gio/modules$ ls
libdconfsettings.so

actually the fact that "-r" doesn't reveal any stderr, let me think that the library is actually instantiating more gvfsd instead of reusing them.
Once I get into the desktop, one gvfsd is already running (probably spawn by lxpanel or cairo-dock, or pcmanfm).
Running the cmd line mount shows those warnings, and then I get more gvfsd 
each time.

(In reply to comment #5)
> FYI: You could try environment variables (e.g. GVFS_DEBUG, GVFS_SMB_DEBUG) to
> print some useful debug info, see:
> https://wiki.gnome.org/Projects/gvfs/doc

Thanks Ondrej,

I tried also this, in many different ways, but no debug is ever output.
I believe the problem is really before smb is ever spwan.

The GVfsUriMapperSmb warning, let me think that the URI is actually paresed badly, and then sent to gvfsd with wrong prams (as for host instead of server, and that prefix containing double numbers!), so the gvfsd-smb is never even
launched.

What I also noticed is:

- I enter the desktop, and I have one gvfsd process, with parent=1
- I run /usr/lib/gvfsd -r, stays there, and the previous process is still there
- I run another /usr/lib/gvfsd -r, and the first one quit saying his name is taken
- When I run successive "dbus-launch gvfs-mount smb:...." I get one new gvfs process with parent=1

Maybe the problem is when gvfsd is spawned?
Dbus is not able to get the running instance and spawn another one?
How can I let the dbus service file run gvfsd with stderr on a file?

(In reply to comment #8)
> - I enter the desktop, and I have one gvfsd process, with parent=1

What desktop? Does it start session bus itself? Check the UID of the process, there may be leftovers from login manager (gdm).

> - I run /usr/lib/gvfsd -r, stays there, and the previous process is still there

That should not happen if both processes are sharing the same session bus. Before doing this, check what's running on the bus, e.g. use d-feet.

> - When I run successive "dbus-launch gvfs-mount smb:...." I get one new gvfs
> process with parent=1

dbus-launch will start new session, isolated from others, available only to parents withing the same shell (the command it starts).

(In reply to comment #7)
> (In reply to comment #5)
> The GVfsUriMapperSmb warning, let me think that the URI is actually paresed
> badly, and then sent to gvfsd with wrong prams (as for host instead of server,
> and that prefix containing double numbers!), so the gvfsd-smb is never even
> launched.

You could try to run the daemon manually and show what happens:
e.g. GVFS_DEBUG=1 /usr/libexec/gvfsd-smb user=gbulfon server=pgbulfon share=Users

(In reply to comment #9)
> (In reply to comment #8)
> > - I enter the desktop, and I have one gvfsd process, with parent=1
> 
> What desktop? Does it start session bus itself? Check the UID of the process,
> there may be leftovers from login manager (gdm).

it's our own build of lxdm login manager, going on with lxde+cairo-dock+pcmanfm,
running on illumos kernel.
I checked that when not yet logged in, no gvfs process is running.
Once entered, one is spawn by some desktop component, and it has my uid,
but parent=1.

> 
> > - I run /usr/lib/gvfsd -r, stays there, and the previous process is still there
> 
> That should not happen if both processes are sharing the same session bus.
> Before doing this, check what's running on the bus, e.g. use d-feet.
> 

I'll check for d-feet

> > - When I run successive "dbus-launch gvfs-mount smb:...." I get one new gvfs
> > process with parent=1
> 
> dbus-launch will start new session, isolated from others, available only to
> parents withing the same shell (the command it starts).

Anyway, I believe we're going wrong direction, because other URIs work correctly (sftp, ftp, file), only the ones warned by glib-gobject end up with "Invalid spec", as if they not even can parse correctly.

Is there any way to check what glib is doing there? Who is registering twice?

(In reply to comment #10)
> (In reply to comment #7)
> > (In reply to comment #5)
> > The GVfsUriMapperSmb warning, let me think that the URI is actually paresed
> > badly, and then sent to gvfsd with wrong prams (as for host instead of server,
> > and that prefix containing double numbers!), so the gvfsd-smb is never even
> > launched.
> 
> You could try to run the daemon manually and show what happens:
> e.g. GVFS_DEBUG=1 /usr/libexec/gvfsd-smb user=gbulfon server=pgbulfon
> share=Users

This shows some SMB debug, trying to mount correctly, but then fails "Not owner" becuase it can't ask for a password.
This is not even reached when I do gvfs-mount, becasue parsing of the URI looks wrong. Look at this output I added in the source on "try_mount" from smb backend,
the mount_spec contains wrong keys:

smb-share:host=pgbulfon,user=gbulfon,prefix=0.000000Users

host, instead of server
prefix instead of share, and with numbers

Looks like I solved it :)

A wrong patch was in glib, adding an additional fixed "/usr/lib/gio/modules"
search path to the first build env GIO_MODULE_DIR.
This was thought for multi arch (like mine), but actually breaks the system:

- 64 binaries will also look in 32 dir, causing many warning logs
- 32 binaries will look twice in 32 dir, causing the problem I had!

I still can't browse the network, with another error, but smb URLS works great!
Solving the browse problem will not be hard ;)


Thanx guys!
Gabriele

I'll mark this as solved.
And it's not a gvfs problem.
Thanks again.

Gabriele.

Anyway, maybe there should be a way for gvfs to help debugging situations like this :)