Previously it depended on addFolder() / removeFolder() calls to adjust
watchers when new folders were added or removed. There also needed to be
complex move handling.
Now, any folder creation/move-in notifications automatically trigger
watcher additions and folder deletion/move-out triggers removal.
This makes sure that the conflict list can be updated if a user resolves
a conflict by deleting the local conflict file. Previously one had to
wait for the next sync run.
If a folder was renamed A -> B, the folder watcher for the inode
would be unaware and still report changes for A/foo. Now directory
renames in the watched folders are tracked and paths are updated
accordingly.
In case of past collisions during the 3.0 times... well one will resync
from scratch unfortunately. But if that happened there are likely other
problems which occurred.
Also this might fix some of the bugs with people loosing settings from
the database. Indeed the -wal and -shm concatenations were wrong. Using
append was in fact changing the folderDefinition member which (I guess)
would potentially lead to funny ".db-wal-shm-wal-shm" names.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We had signals just to call those backs in ownCloudGui, they were
otherwise unused. So let's move them inside of Systray since it's
specific to it anyway.
Also fix the dangerous call to sender(). We can call this function
without going through a signal/slot connection and also it's never
connected to an AccountState.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Because of PR #2580 the settings dialog doesn't always exist. We need to
check for it first before placing calls to it.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This was right now on the check connectivity beat which is too much when
you got many users. Be more conservative there and only update the list
of apps when the account gets connected.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
QtKeychain provides Qt5KeychainConfig.cmake and friends nowadays, so no
need to have a less reliable and outdated find module on our end.
Also this shows that we were including keychain.h in the wrong way and
were not using the link target, so both got fixed as well.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
When we skip a watcher event on mac, we log the file path but not the
event flags which came with it. Let's add it, it should help figure out
what's going on with #2578.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This is quick enough that no flickering should appear in practice.
We end up doing this because for some reason on Windows (I dug up deep
into the Windows QPA without nailing it down) not showing that systray
window at least once prevents the app object to return from exec() when
the session ends.
It's as if that window would be in some limbo state (neither opened nor
closed) which would prevent quitting. Clearly what we're doing here is a
workaround...
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
No need to go to the file manager first to then have the user go through
the context menu, just popup the dialog directly.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We better do this before this business logic grows in the QML side and
gets out of control. We'll need finer grained information due to the
conflict handling anyway.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This was completely disabled code and the connect was gone... So let's
introduce the connect back and have it trigger the main window while
selecting the right user. This way the new activity list is displayed
and since the conflicts are shown at the top this is hopefully obvious
again.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This should be safe in the case of conflicts in folders on which the
user can write. For other cases we still use the older actions.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
The socket api move and delete commands are not strictly about conflicts
since they also deal with files which couldn't be uploaded for some
other reason. Still the new ConflictSolver could be used in those cases.
This opens the door at reusing that logic in other places.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This could only work at the root of the sync folder where the record for
the parent folder would be invalid. Otherwise the negation would be
wrong... assuming you can add a file only if the permission is not
there.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
For conflicts generally as well as new files in read-only directories
the context menu will now present delete and move options.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
setWindowFlags triggered changeEvent, thus causing a crash in customizeStyle.
This fix should be kept even if we decide against delayed init in the future.
Signed-off-by: Michael Schuster <michael@schuster.ms>
ownCloudGui::slotShowSettings already got what it takes to create it only when we try to show it for the first time.
This however has some implications:
Pros:
- Only created when needed, while testing saved ca. 20 MB of RAM and got freed again after closing the dialog.
- Since we defaulted to the new Tray UI from 3.0, this is an added bonus for users don't opening the settings.
Cons:
- Resources like the avatar image have to be refetched everytime the dialog is recreated.
This may be desired as well, because it ensures displaying no outdated info (e.g. on connection issues).
Signed-off-by: Michael Schuster <michael@schuster.ms>
Use a similar trick of a semi-transparent rectangle on top when the
mouse area is hovered. This way it will always work whatever is the
background color.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Otherwise it would spin forever while we know we're not doing any work
anymore since we got a message from the server.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
All the other ShareLinkWidgets process that signal (which allows to
display error messages for instance) but not that one for some reason.
That being said it might need to deal with an enforced password
situation.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
It turns out the shareDeleted() signal is connected to a function
cleaning up the ShareLinkWidget holding the last shared pointer to the
Share object. Since we use member variables for calling updateFolder()
this would lead to using deleted objects.
Just swap the call and the signal to have everything back in order.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Make sure we got a fixed size for that label. Also ensure that the
pixmap we display there is properly scaled to fit while maintaining the
aspect ratio.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Combining translated strings like this makes them hard to translate since the order of words is different between languages.
Use proper placeholder strings instead.
Signed-off-by: Nicolas Fella <nicolas.fella@gmx.de>
By using properties and property bindings the QML code gets more declarative rather than imperative, which is considered better.
This patch:
- Introduces a currentUserId property in UserModel that replaces the equivalent Q_INVOKABLE call
- Introduces an avatar property in User that contains the avatar's image provider url without any fallback
- Introduces new image provider urls for fallback images
- Moves the fallback image selection to QML since we want different fallbacks according to where it is used
- Wires up the necessary signals to propagate a changing avatar
Signed-off-by: Nicolas Fella <nicolas.fella@gmx.de>
This way the server displays a less scary string while granting access
from the browser. Also this same string will be used in the "Devices and
sessions" section of the server settings.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This will harvest everything we might need for debugging purposes:
* config file
* sync journal dbs
* log files
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
I'm not a huge fan of using private APIs but QZip is really the API with
the least hassles for our debug archive need. No external dependency and
we know it is generally available and stable despite the lack of
stability promise.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Since we changed the default in the config file and since log dir had
precedence on log file, the --logfile command line option wasn't doing
anything anymore.
We make sure it has an effect again overriding --logdir or the logDir
config entry.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This is the same approach used on the server side. Turns out I quite
like it, this avoids popping up a dialog to the user and since she won't
know the password she'll have to set a new one anyway or disable it.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This is necessary for downloads coming from Github for instance. They
are systematically redirected and we'd just fail the download.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We're not interested in any found element in these cases, just to check
that none of the elements matches.
Signed-off-by: Harald Eilertsen <haraldei@anduin.net>
Most instances have been converted to range based for, but std::find_if
has been used where it made sense.
Signed-off-by: Harald Eilertsen <haraldei@anduin.net>
APPLICATION_SERVER_URL can be either empty or be specified. This commit adds the new CMake option APPLICATION_SERVER_URL_ENFORCE to decide whether to enforce the
URL's unmodified use (like before, default: ON) or to allow modification by the user (new).
By default APPLICATION_SERVER_URL_ENFORCE is set to ON, to no break with the previous implementation's expectations.
If APPLICATION_SERVER_URL is empty, APPLICATION_SERVER_URL_ENFORCE will be ignored by the Account Wizard.
The previous behaviour confused me a bit with branded builds. When the URL was (usually) specified but not forced, it was simply discarded, forcing the user to
manually supply it.
Signed-off-by: Michael Schuster <michael@schuster.ms>
Otherwise we get lots of "No description available" lines in the
activity list which is basically noise. Also trains the user to ignore
the secondary line.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Indee the MenuItem might not be linked to its Menu at creation time
which will make the binding fail and give a warning. Delay for the menu
availability.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Now that we depend on Qt 5.12 anyway, the count property is available
just fine on the Menu item.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Under Wayland QCursor::pos() is unlikely to give us anything meaningful,
so fallback to the primary screen information.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Users get rightfully confused with the "You changed ..." messages in the
activity list for syncs. Indeed, some of those changes might be coming
from the server in which case we don't really know who did the change.
So now we use the old "Synced ..." messages for changes pulled from the
server and we have a more precise "You changed ..." (renamed, deleted,
created) when the changes were initiated locally (since there we know
the user reading the message did it).
Also changed how the messages are constructed so that they can be
properly translated.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
I wrongly (and stupidly) assumed the int in QBitArray was treated as a
hash as well but it is an index so it must be positive. Also to make
things even worse I misread on which expression clang-tidy was
complaining regarding the implementation specific narrowing
conversion... This is happening after the modulo operator and not
before. We're in a safe range of values at that point, so it's fine to
let the narrowing happen.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This flagged mostly unused parameters. Didn't enable the
misc-non-private-member-variables-in-classes check as we got a lot of
those. Hopefully we'll get to fix them at some point but that feels too
early and too much work for now.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
The visibility of the share button for the activity list is tied to the
path role in the ActivityListModel, so make sure we don't return a path
for E2EE files and folders.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Those files and folders are not shareable and the "Edit" or "Open in
browser" actions will lead to showing an error in the web GUI. No need
to lead users there so just disable them.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Turns out that showing the button straight from the page ctor is a bit
too early. At that point the account might not be connected yet and thus
we wouldn't have proper information. Currently we were displaying that
button all the time, now we wait for the account to be connected to
decide to show it or not.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We lost the ability to display the E2EE mnemonic during the GUI
redesign and the info message wasn't displayed again on restart. So now
we display it every time, it still can be dismissed and the button text
is different in such a case to make the intent clearer.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We update properly the CLSID of folders to integrate them with the
Explorer pane when they are added or removed from the GUI. That said,
there might be cases of users adding folders to sync straight from the
GUI. In such a case the CLSID might be missing on config load so now we
assign one on the fly when that happens.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
The context menu should allow access to the Settings even when no accounts are configured.
Users may specify proxy / startup / update and other settings at any time.
This slipped through in #2164.
Signed-off-by: Michael Schuster <michael@schuster.ms>
- Fix for #2173.
- Removed the workaround for the left round corner which was also
fixed with the OpacityMask.
Signed-off-by: Camila <hello@camila.codes>
The context menu offers to open the main dialog and the settings
even if no accounts are configured. In this case, the main dialog
is useless and the settings are probably confusing. Hence, this
commit replaces these actions in the context menu by an action
to open the wizard (which also opens on left click, so this is
the most natural thing to do).
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Previously we were using QCursor::pos() in the Linux case, which is a
fair heuristic except it will always be relative to the primary screen
and not the current screen. This explains why we had to adjust with the
virtual geometry.
In the Windows and Mac case we got the position out of QSystemTrayIcon
and that one was already relative to the current screen.
So now we use QCursor::pos(currentScreen()) which ensures we give the
coordinates relatively to the current screen also in the Linux case.
Since all platforms are now having that point in the same coordinate
system we don't need to mess around with the virtual geometry to
compensate.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
- The default displayed was not in sync with the server because
the client was setting permissions when creating a share while it
should get the default permissions from the server to display it to the
user first.
Signed-off-by: Camila <hello@camila.codes>
The value gets already initialized by default member initialization,
so there is no need to set it again in the constructor.
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
The menus may become too large if they have too many items
(the account menu has this in case of many accounts, the apps
menu in case of many apps). This commit limits the maximum
height such that the menus do not hide the header.
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Turns out that the ListView embedded in the Menu (reachable via
contentItem) would create a binding loop if we're using its
childrenRect. But really we're interested in the total width for the
instantiated delegate *inside* the ListView. That's why we go one level
deeper and get the childrenRect of the inner contentItem instead.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
When specified in the config file, the Proxy password will be migrated
to the keychain, for backward compatibility and to allow admins to
overwrite an existing password by rolling out updated config files.
Once migrated to the keychain, the password will be removed from the
config file.
Signed-off-by: Michael Schuster <michael@schuster.ms>
Originally this was in the WebFlowCredentials class. Since we've abstracted everything
from there already, let's also move this in case some other code may use
KeychainChunk::ReadJob prior to WebFlowCredentials.
Signed-off-by: Michael Schuster <michael@schuster.ms>
Ultimately we would avoid the code duplication by creating our own
proper QQC2 style instead of directly tuning everything. That's a battle
for another time (more elements would need to be adjusted for that, I
don't want to dilute this branch too much).
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
- The code handling the api response was not up to date with the latest
api changes.
- Unlimited quota display: use qint64 to avoid data loss
- Change total with quota: total holds the quota total and not the storage total.
Signed-off-by: Camila <hello@camila.codes>
Update systray behavior and context menu:
- left click brings up the new QtQuick based dialogs on all latforms
- right click brings up the new QtQuick based dialog on Mac OS only
- right click brings up a context menu on all other platforms than Mac OS
- "Quit Nextcloud" => "Exit Nextcloud"
- Add "Open main dialog" option.
Signed-off-by: Camila <hello@camila.codes>
During the wizard we currently don't have much information about the
encrypted folders. In particular we can only display their mangled names
which is far from ideal for the user to make an informed choice.
That's why in the wizard we now forbid creation of subfolders in e2ee
folders and we also don't display subfolders of e2ee folders.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Since the e2e oracle works only in term of absolute remote paths and
that our model x._path was relative, we need to properly convert.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Otherwise it was a bit confusing and annoying for filter rules:
e.g. "nextcloud.sync.*" vs "sync.*".
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This way this whole logic isn't stuck into the settings dialog anymore.
Also cleaned up the unused "decrypt folder" logic.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This is a much better place than the GUI, this way we ensure the
propagator is always operating of up to date information. Previously if
the propagator kicked in without user interaction from startup (not
showing the settings dialog) it would have no E2E information available
whatsoever... unsurprisingly it would thus take wrong information at
every turn.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Since we want to move to a place where the encryption of subfolders is
always enforced it makes no sense to leave it in control of the user.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Using a reference capture is a disaster waiting to happen here, if for
some reason we'd move from exec() to popup() for the menu below we'd be
getting garbage in the lambda call.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
The Qml position setting code is moved to C++, which allowed to get rid
of the currentScreenIndex() method (which seemed to be just a detour to
pass the screen from C++ to Qml).
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Commit 07bede8 (PR #1892) introduced a new helper method currentUser()
that didn't check for _users.count() thus causing to throw an
"index out of range" exception when no accounts are configured.
This commit uses the opportunity to add more sanity checks to UserModel.
Signed-off-by: Michael Schuster <michael@schuster.ms>
Commit a12205f322 (PR #1891) introduced
a circular ownership: qmlRegisterSingletonType<Systray>(...) makes the
QQmlEngine own the resulting singleton Systray instance, however, the
QQmlEngine _trayEngine itself is owned by the Systray instance. This
circular ownership results in a crash when the destructor of Systray
calls the destructor of _trayEngine which attempts to call the destructor
of Systray.
This commit solves this problem by making ownCloudGui, which is the
parent of Systray, the parent of the _trayEngine.
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
In case the default color value will ever change, this
commit makes sure that the colors of the menu border and
the menu separator coincide.
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Build failed on Windows, missing libsync export for printQNetworkProxy after
moving ClientProxy class from gui to libsync.
Signed-off-by: Michael Schuster <michael@schuster.ms>
Previously we were showing up the settings dialog but it doesn't quite
make sense anymore now that we got two dialogs.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Note this won't work on all platforms. KDE Plasma and GNOME Shell (with
systray extension) assume that right click is necessarily for a context
menu exposed via D-Bus, there's not nice way to make the right click
popup the main dialog on those platforms.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We were calling accountState() on a "folder" member which could be
nullptr. In fact this would happen any time one right click on a file
outside of a sync dir under Windows, this thus led to a crash.
Since the capabilities variable was unused anyway, we just removed it.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
We then get more items clickable, in particular the bottom one proposing
to open the activity application. Now we can click it and it opens the
activity application straight away.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
For the "Activity" type activities, displayPath contains the file name
which is also in the subject. This was redundant, so now we don't
display an info line at all for that type of activities.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Context properties are deprecated and will be removed in Qt 6.
QML singletons are the better option here, especially given that
UserModel and UserAppsModel already are singletons.
Signed-off-by: Nicolas Fella <nicolas.fella@gmx.de>
Deduplicate the Button code by moving it into a separate file.
Also use the same background code for the apps button since as far as I can tell it didn't behave any different
Signed-off-by: Nicolas Fella <nicolas.fella@gmx.de>
Seems like Qt sometimes doesn't like the QML window's screen property to be set
to a C++ (QVariant) pointer value, so we use the index: Qt.application.screens[]
See Qt docs: https://doc.qt.io/qt-5/qml-qtquick-window-window.html#screen-prop
This fix returns the matching window's index from the QGuiApplication::screens()
list to the QML side, instead of the window pointer.
Steps to reproduce the crash with the previous code:
- Open the Tray menu and close it a few times,
or scroll randomly up and down in its activity list.
Tested with Qt 5.12.5
Signed-off-by: Michael Schuster <michael@schuster.ms>
This leads to simplifying the computation code quite a bit as well.
Indeed we're separating concern between what is window size dependent or
not and that shows.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
This was leading to the same logic being duplicated several times. It's
fine to return the QScreen* on the QML side directly but wrapped into a
QVariant.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
The API is just more convenient there, the rect and point types on the
QML side are just pale shadow of their C++ counterparts.
Also improved a bit the constness of the Systray class.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
The top left corner of that screenRect is (0,0) or based on
(virtualX,virtualY) depending on the platform. This leads to easier
reduction in code duplication.
Signed-off-by: Kevin Ottens <kevin.ottens@nextcloud.com>
Make the codebase consistent, we already have a lot of implicit pointer comparisons.
Exception: Stay explicit on return's, example:
return _db != nullptr;
Signed-off-by: Michael Schuster <michael@schuster.ms>
