Tails

#2 Updated by sajolida 2013-09-06 10:04:52

That’s definitely related to GNOME bug 604015, since the Trash on a separate USB stick is handled correctly: its content is merged into the Trash presented by Nautilus from the main home folder for example.

https://bugzilla.gnome.org/show_bug.cgi?id=604015

Reading this report, I think the bad behavior we experience with the persistent storage is relate to how it is mounted. It says:

« The trash backend currently only tracks trashed files on “user interesting” mounts — those that g_unix_mount_is_system_internal() returns FALSE for. This is approximately equal to “stuff mounted in /media/”. »

I tried to unmount it from /home/amnesia/Persistent and bind it onto /media/Persistent but it doesn’t work either. It would have been too easy of course.

Anyone feels like exploring g_unix_mount_is_system_internal() and finding a way of having it return FALSE for our Persistent folder?

#3 Updated by intrigeri 2013-09-06 10:18:42

> Anyone feels like exploring g_unix_mount_is_system_internal() and finding a way of having it return FALSE for our Persistent folder?

Even if we can workaround this now, once we get better boot medium
protection in Wheezy (Bug #5518), then it’s likely that the
workaround breaks.

There might be some conflict between what we want for security (that
is, see the boot medium as a system one, which bilibop possibly does,
or not yet?) and what we want for better UX :(

#4 Updated by alant 2013-09-14 07:16:44

> « The trash backend currently only tracks trashed files on “user interesting” mounts — those that g_unix_mount_is_system_internal() returns FALSE for.

Verified, see [https://git.gnome.org/browse/gvfs/tree/daemon/trashlib/trashwatcher.c]

> Anyone feels like exploring g_unix_mount_is_system_internal() and finding a way of having it return FALSE for our Persistent folder?

I had i look at [https://developer.gnome.org/gio/2.29/gio-Unix-Mounts.html#g-unix-mount-is-system-internal] and [https://git.gnome.org/browse/glib/tree/gio/gunixmounts.c], especially guess_system_internal and g_unix_is_mount_path_system_internal.

I fail to see why it wouldn’t be FALSE for our case.

#5 Updated by alant 2013-09-15 04:27:49

I found that `/live/persistence/*_unlocked` is not considered as internal by GIO `g_unix_mount_is_system_internal`.

However GIO (responsible for putting the file to trash) and GVFS (responsible for looking for trashes) disagrees on the location of the trash in our specific case:

- when one trashes a file from Persistent, GIO’s `g_file_trash` calls `g_local_file_trash` that lookup each parent directory until it finds a device and thus find `/home/amensia/Persistent` and creates a `.Trash-1000` there;
- GVFS trashlib get mounts from GIO’s `g_unix_mounts_get`, that explicitely ignore bind mounts and thus returns `/live/persistence/_unlocked` but not `/home/amensia/Persistent`. It thus look for `.Trash-1000` in `/live/persistence/_unlocked`.

#9 Updated by intrigeri 2017-10-02 15:53:49

A user on XMPP was affected by this problem again today. Good candidate for UX improvements we may want to prioritize higher, perhaps?