Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
|
|
The returned object path for non-human users can not really be used
since there is no guarantee for how long it will be valid. The next
run of reload_users will unregister non-human users, and that can
happen at any time.
|
|
|
|
|
|
Also clean coverage files on 'make clean'
https://bugs.freedesktop.org/show_bug.cgi?id=63868
|
|
We currently emit the "changed" signal when most aspects of a user
change, but we neglect to do it when the account type changes.
This commit fixes that.
https://bugs.freedesktop.org/show_bug.cgi?id=65846
|
|
This way we pick up changes to wheel membership
https://bugs.freedesktop.org/show_bug.cgi?id=65846
|
|
This deduplicates a bit of code.
https://bugs.freedesktop.org/show_bug.cgi?id=65846
|
|
Right now we set the password mode of a user based on the password
hash in shadow. SSSD (and other remote login facilities
presumably) don't send the password hash over the wire to the client,
but of course those users do have a password set.
This commit changes the code to notice when there's no shadow entry
for a user and treat that case conservatively (assume user has a password).
Related to: https://bugzilla.gnome.org/show_bug.cgi?id=708997
https://bugs.freedesktop.org/show_bug.cgi?id=70005
|
|
Bring back the simple login.defs-based check for if a user is human or
not and enable it by default.
Add a build option --enable-user-heuristics to get the old behaviour
back again.
Split out all human vs. system user divination into a new file,
user-classify.c in order to clean up daemon.c a bit.
|
|
So that other projects can depend on recently committed changes
|
|
There's no reason to, so don't bother.
https://bugs.freedesktop.org/show_bug.cgi?id=67920
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=67700
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=67700
|
|
...and a polkit action file to go with.
https://bugs.freedesktop.org/show_bug.cgi?id=63733
|
|
First pass at what a patch might look like. Requires the new GDBus
async property handling changes.
https://bugs.freedesktop.org/show_bug.cgi?id=63733
|
|
When updating a User object from keyfile, keep reference on the keyfile
object passed in and modify the Daemon not to destroy it.
When saving a User, instead of creating a new keyfile, reuse the one we
stored on the object.
We still record the properties from the User object into the keyfile in
the previous way (and include a modification to clear out the 'User'
group before doing this).
The intention here is to allow other groups stored in the keyfile to be
preserved. These other groups are likely to correspond to extension
interfaces (which will be introduced in future patches).
An alternative approach would have been to only preserve groups for
extensions that we currently have loaded but this was abandoned as being
excessively brittle since the result of a temporarily missing extension
file would be the destruction of data.
https://bugs.freedesktop.org/show_bug.cgi?id=63733
|
|
Right now if a nonexistent user is requested from the accounts service,
we don't give the caller any sort of notication that the user doesn't
exist. This cascades into d-bus timeouts and ultimately a broken login
screen.
This commit adds a new "nonexistent" user property to the user object
and sets it to TRUE when a requested user fails to load. It also
makes sure the user objects is-loaded property gets set.
Based on a patch by Lionel Landwerlin <llandwerlin@gmail.com>
https://bugs.freedesktop.org/show_bug.cgi?id=66325
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=65649
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=65160
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=65160
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=65753
|
|
Otherwise, we randomly stop reacting to session changes when
it happens to return with FALSE by chance.
|
|
|
|
It's caused various problems with distcheck, vala projects don't
actually need it (they can just use gobject-introspection instead),
and the Vala FAQ recommends shipping .vapi files in the consumer
source, not provider source:
Is having a local copy of a .vapi in my project
good practice?
Yes, it certainly is. There are a number of
reasons why you want to do this.
1. Sometimes fixing bindings may lead to breakage in
projects using such bindings. So having a local
copy means that your project will not break
whenever a bindings is updated system-wide.
2. If your project works with such bindings,
there's no need to upgrade the bindings just for
the sake of being up-to-date. It works, that's
what matters.
3. It's you that decide when to upgrade the
bindings, not the system. Thus it's you that
decide when to break your own project in order to
have more up-to-date bindings. Your project will
keep compiling fine, even if system-wide bindings
change.
|
|
|
|
|
|
|
|
This is like the previous commit, but for disabled
users instead of deleted users.
|
|
Right now it's possible to delete a user who is configured to
automatically log in.
Doing so causes the system not to boot.
This commit turns off autologin before deleting the user.
https://bugs.freedesktop.org/show_bug.cgi?id=54506
|
|
|
|
In particular, don't assume that there is an active session on seat0
and that it is us.
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=65172
|
|
Instead, offer act_user_is_logged_in_anywhere and
act_user_get_num_sessions_anywhere to inquire about them.
https://bugs.freedesktop.org/show_bug.cgi?id=65160
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=65159
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=63159
|
|
To reload users history.
https://bugzilla.gnome.org/show_bug.cgi?id=697040
https://bugs.freedesktop.org/show_bug.cgi?id=63159
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=63159
|
|
The function isn't called only for passwd monitor, but for all of them.
https://bugs.freedesktop.org/show_bug.cgi?id=63159
|
|
Right now if a greeter session happens to start in the background, then
libaccountservice will suddenly recognize gdm as a valid user (since
we automatically recognize users for sessions that come in).
This commit makes that detection logic only consider user sessions, not
greeter sessions.
https://bugzilla.gnome.org/show_bug.cgi?id=701514
|
|
Right now the check for "is system user" is less than solid.
It depends on checking things like whether or not a password is set,
and whether or not a shell is set. Unfortunately, sometimes a password
is not set for a non-system account, and sometimes a shell is set for
a system account, so the heuristics aren't perfect.
In one particular case, we currently get things wrong: right after a
user is created before they have a password.
This commit attempts to shore up the heuristics by never treating cached
users as system users.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
We use some less than perfect heuristics for determining whether or
not a user is a "system account".
This commit defers to what's in the key file when the key file
explicitly says.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
This client already handles filtering system users
from the cached user list, so there's no point in doing it
daemon side.
Furthermore, it's harmful since then clients never get informed
when a user account type changes from a system account to a user
account.
Let the client do the filtering.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
We currently don't show system users (e.g. 'mysql') in the
control-center user list. Occasionally a system user gets promoted to
a normal user account. When that happens, we fail to notice.
This commit fixes that.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
This commit changes account service to write whether or not the
account is a system account in the key file about the user.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
The cache_user function currently just does:
comment = g_strdup_printf ("# Cached file for %s\n\n", user_name);
g_file_set_contents (filename, comment, -1, &error);
This commit changes it to write out a full key file.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|
|
AccountsService can be used for creating user accounts, but shouldn't
be used for creating system accounts. Therefore, all accounts created
with AccountsService can safely be considered user accounts.
This commit changes the daemon to immediately, implicitly cache all user
accounts created via the accounts service. This ensures they'll never
get errnoneously tagged as system accounts.
https://bugs.freedesktop.org/show_bug.cgi?id=64769
|