| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|\ |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
The mount may be already gone or monitoring cleanup is done too late;
in that case handle this case gracefully since the monitoring doesn't
exist anymore.
|
| |
| |
| |
| |
| | |
This is a valid error used to indicate that we should (auto)mount first
and should not be presented to the user.
|
| |
| |
| |
| |
| | |
Found out that if the remote call fails, some variables are left
uninitialized and causes assertion failures while we don't need them.
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
This hopefully covers all public methods.
Seems like this fixes several memleaks as well.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
...so that they are recognizable on a first sight.
|
| |
| |
| |
| | |
and kill the iters.
|
| |
| |
| |
| |
| | |
Reorganize saved references for used objects, getting rid of
some FIXMEs.
|
| |
| |
| |
| |
| | |
Use G_DBUS_PROXY_FLAGS_DO_NOT_CONNECT_SIGNALS and G_DBUS_PROXY_FLAGS_DO_NOT_LOAD_PROPERTIES
flags where applicable.
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
Forgotten from the 722ebaf0ab949ca5b844a387dc684b4e3185db75 commit
|
| | |
|
| |
| |
| |
| |
| |
| | |
It's little messy due to shared code across sync calls with private
main context and async calls that don't have it's mainloop and can't use
private context.
|
| |
| |
| |
| |
| |
| |
| | |
...by turning sync copy calls async and running mainloop for progress
interface skeleton be able to process incoming calls.
Also, new class GVfsJobProgress has been introduced mostly for code sharing.
|
| |
| |
| |
| | |
This allows clients to react on GVfs mounts appearing and disappearing.
|
| | |
|
| |
| |
| |
| | |
Not really a problem in this case, but let's be on a safe side.
|
| |
| |
| |
| | |
...not doing that leads to accessing freed memory
|
| |
| |
| |
| |
| |
| | |
This mimics the old behaviour and works for sync and async.
Requires new glib for g_dbus_connection_get_last_serial().
|
| |
| |
| |
| | |
And conform to usual GNOME naming conventions.
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A fallout of port away from convenient G_DBUS_TYPE_CSTRING type and
_g_dbus_message_append_args() + _g_dbus_message_iter_get_args() functions
that masked NULL strings as non-NULL empty strings ("") for transferring
over d-bus and on the other side reconstructed the original value.
Portions of other code treat non-NULL strings differently even if they
were empty. This fixes e.g. opening http URLs via gvfs-open.
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
The crasher may have been possibly fixed in bug 669211.
It's hard to hit the code path as long as gvfsdaemon calls hard exit().
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
...and fix wrong typecast due to using an object we haven't reff'ed. The
failure we're making silent were coming from a failed d-bus call to mount
tracker.
Also turned out the org.gtk.vfs.MountTracker.unregisterMount() has never
been implemented, daemon mounttracker is watching for clients disconnects
anyway.
|
| | |
|
| |
| |
| |
| | |
This makes http backend working again.
|
| | |
|
| |
| |
| |
| | |
So that nothing depends on gvfsdbusutils.h, which is full of old libdbus code.
|
| | |
|
| |
| |
| |
| | |
Traces of gnome-keyring GetEnvironment() call.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| | |
I hate reverting things but this seems to be required for proper
libdbus threads initialization, further used by gnome-keyring.
This has actually been fixed in gnome-keyring recently (bug 659162)
but gvfs can still be used with gnome-keyring-2.32.
|
| |
| |
| |
| | |
Handle a connection dying and unregister its subscription.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This changes the way registered paths (i.e. interface skeleton exports)
are activated on daemon side. We're keeping a list of connections on which every
registered path is exported so that if connection is closed or the path is
unregistered we always free the right interface skeletons.
We also keep track of active client (peer-to-peer) connections so that if a new
path is registered, it's automatically exported to all active connections. Needed
since client connections are usually shared.
It also has a nice side-effect of reacting to dirty disconnects (e.g. the client
dies).
This whole registration machinery needs unique object paths (for the hash table)
but that's not a problem in the current state.
|