| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
| |
This commit moves detection of the location
of the wtmp file to configure and fixes that
detection to work on Solaris and NetBSD.
https://bugs.freedesktop.org/show_bug.cgi?id=90882
|
|
|
|
|
|
|
|
| |
This debug message unnecessarily implies that the accounts daemon is
not running. There are other reasons why loading a user could fail;
e.g. if a username simply does not exist.
https://bugs.freedesktop.org/show_bug.cgi?id=93655
|
|
|
|
|
|
| |
No need to dup and free a bunch of string literals.
https://bugs.freedesktop.org/show_bug.cgi?id=93654
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
So far the only way of setting the password hint is by setting also the
password.
But there are several cases where setting the password isn't done through
AccountsService, but still we want to set the password hint. For instance GDM
setups the pasword for new used directly from PAM.
Thus, let's add a new function to set the hint without changing the password.
https://bugs.freedesktop.org/show_bug.cgi?id=89588
|
| |
|
|
|
|
| |
https://bugs.freedesktop.org/show_bug.cgi?id=89472
|
|
|
|
|
|
|
|
|
| |
In commit 2c51bd83b65076cf98bd1baf48d9add31435d118,
DBusGProxy *bus_proxy was converted to GDBusProxy but the daemon
hasn't needed or used the bus_proxy since the conversion to GDBUS.
This patch simply removes it.
https://bugs.freedesktop.org/show_bug.cgi?id=89318
|
|
|
|
|
| |
<PaulePanter> halfline: Could you please add the language to the error message in
`src/libaccountsservice/act-user.c`?
|
| |
|
| |
|
|
|
|
|
| |
Not all X11 sessions know their display ahead of time,
so don't require that it be known to proceed.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
accountsservice exports an object on the system bus
for every user it tracks. The path of the object is
derived from the user's uid.
The format string used for computing the object path
supposes the UID is signed. UIDs are unsigned quantities,
though, so very large UIDs get misrepresented as negative
values in the resulting object path. The negative sign
"-" is invalid in an object path. Under these circumstances,
accountsservice will crash.
This commit corrects the problem, by using a format string
that is appropriate for unsigned values.
https://bugs.freedesktop.org/show_bug.cgi?id=85688
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The various NSS calls don't give accurate results in some configurations
until midway through boot up. This is because SSSD or winbind (or
whatever) needs to initialize.
In order to prevent accounts service from using NSS prematurely, we need
to add an ordering constraint between the nss-user-lookup.target and
accountsservice.
This commit accomplishes this by adding the appropriate Wants= and
After= directives to the accountsservice systemd unit file.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
if accountsservice is activated by the ListCachedUsers call,
then the call may get processed before the list of users is loaded.
This commit closes the race by synchronously loading the list of
users at start up.
|
|
|
|
|
|
|
| |
It is needed for deleting enterprise accounts in gnome-control-center.
See https://bugzilla.gnome.org/show_bug.cgi?id=727871 for details.
https://bugs.freedesktop.org/show_bug.cgi?id=84091
|
|
|
|
|
|
|
|
| |
If add_user gets called on user object that's already being tracked
by object path, we can end up replacing and freeing the user object
but leaving the old object path memory in the hash table.
This commit makes sure we replace the key when replacing the object.
|
|
|
|
|
|
|
| |
When srcdir != builddir gtkdocize needs to be run in the right
directory, or it will complain about missing configure.ac
https://bugs.freedesktop.org/show_bug.cgi?id=84047
|
|
|
|
| |
https://bugs.freedesktop.org/show_bug.cgi?id=84047
|
| |
|
|
|
|
| |
Another dose of g_debug to help find a crasher
|
|
|
|
| |
To help find a crasher
|
| |
|
|
|
|
|
|
| |
AccountsService occassionally sends out a new-user message twice in
a row for a particular user. There's probably a bug there, but for
now be more vigilant about filtering out the spurious message.
|
| |
|
|
|
|
|
| |
I'm trying to get a grip on a crasher, so this commit introduces
a few more debug messages.
|
|
|
|
| |
It's a little strange to call a cancellable object a call.
|
|
|
|
|
|
| |
This means we don't have to rely on the cancellable getting cleaned
up from the caller when failing (which might not happen depending on the
reason it finishes prematurely).
|
|
|
|
|
| |
This will let us clear the cancellable in the finished handler
unconditionally, removing some subtleness in the code.
|
|
|
|
| |
Just a small clean up.
|
|
|
|
| |
This just adds some debug spew that might help with debugging an issue.
|
|
|
|
|
|
|
|
| |
Change the polkit policy so accountsservice allows use of the
DBus API by admin users that are not logged in via a monitor+keyboard.
This includes users logged in via ssh or Cockpit.
https://bugs.freedesktop.org/show_bug.cgi?id=78279
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
After transporting the generator to a separate file and creating a stub,
I renamed it. Unfortunately, I forgot to rename the stub. This breaks
non-utmpx systems.
Fix that.
https://bugs.freedesktop.org/show_bug.cgi?id=76410
|
| |
|
| |
|
|
|
|
|
|
|
| |
We hardcode the value of the file that we need to watch for wtmp changes
on FreeBSD since it is not exposed in any header file.
https://bugs.freedesktop.org/show_bug.cgi?id=75938
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Instead of using this private macro, use WTMPX_FILENAME.
Also, deal explicitly with the implications of this define being
missing: throw an #error.
This means that some systems that are currently broken (FreeBSD for
example) will continue to be broken, but will throw an #error about it.
Fixes coming in the following commits.
https://bugs.freedesktop.org/show_bug.cgi?id=75938
|
|
|
|
|
|
|
| |
The logic for how to begin the iteration of the wtmp file is
platform-specific. Split out.
https://bugs.freedesktop.org/show_bug.cgi?id=75938
|
|
|
|
|
|
|
|
|
| |
Split the code for opening and iterating the wtmp entries into a
separate 'wtmp-helper' file.
For now, this is a direct transplant of the existing code to a new file
with some minimal renaming and a function to get the name of the file to
monitor for changes (which is still setup from daemon.c).
|
|
|
|
|
|
|
|
|
|
|
| |
We use fgetpwent directly on /etc/passwd in order to ensure we only get
a list of local users (and not ones from the network directory service).
Unfortunately, this function is not commonly found on non-GNU systems.
Provide our own implementation of fgetpwent if the operating system does
not provide it.
https://bugs.freedesktop.org/show_bug.cgi?id=41747
|
|
|
|
|
|
|
|
|
|
|
| |
The save_extra_data function serializes a key file
assocated with the user to disk.
It fails to free the in memory buffer, however.
This commit fixes that.
see downstream bug https://bugzilla.redhat.com/show_bug.cgi?id=1003033
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
commit c5905497733bebf9936a7b028a11ca87caf2d71f changed one of
the two calls to userdel in the code to use -f. This change
was made to ensure users in the logind "closing" state can be
deleted (userdel runs pgrep to make sure a user isn't logged in
unless -f is passed).
Unfortunately, that commit missed the other userdel invocation,
and so now there is inconsistent behavior.
This commit changes makes sure both invocations pass -f.
https://bugs.freedesktop.org/show_bug.cgi?id=65752
|
|
|
|
|
|
|
|
|
|
|
|
| |
relying on login.defs is fragile, and the
user heuristics are fragile.
This commit requires an explicit uid minimum
get configured, and heuristics now only get
applied to the specific problematic range
they were added to address.
https://bugs.freedesktop.org/show_bug.cgi?id=71801
|
|
|
|
|
|
| |
This reverts commit f86c93014e698d81d43fe1ebaf805fa794e5a984.
This was meant to be a downstream patch.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The check we have in place against deleting the root user can
be tricked by exploiting the fact that we are checking a gint64,
and then later cast it to a uid_t. This can be seen with the
following test, which will delete your root account:
qdbus --system org.freedesktop.Accounts /org/freedesktop/Accounts \
org.freedesktop.Accounts.DeleteUser -9223372036854775808 true
Found with the dfuzzer tool,
https://github.com/matusmarhefka/dfuzzer
|
|
|
|
| |
It requires newer glib than we're shipping
|