Manuel Nickschas [Thu, 14 Jun 2018 19:47:26 +0000 (21:47 +0200)]
qtui: Hide "fallback" and "Override" if no system theme is configured
On platforms that don't support icon themes, it does not make sense
to show related options in appearance settings.
Hide the override option in that case, and change wording from
"Fallback icon theme" to just "Icon theme".
Manuel Nickschas [Thu, 14 Jun 2018 19:13:41 +0000 (21:13 +0200)]
qtui: Don't show config for DockManager if not available
Instead of just disabling the checkbox for DockManager notifications,
hide the config widget if the D-Bus service is not available.
Otherwise it's really confusing to see a disabled option for a
notification backend that was basically retired 6 years ago.
Manuel Nickschas [Thu, 14 Jun 2018 19:12:17 +0000 (21:12 +0200)]
qtui: Show core connection dialog only after initialization
Showing a blocking dialog in init() is a bad idea, because then
initialization doesn't actually finish until the dialog is closed.
For example, this delays the setup of the tray icon.
Instead, go through the event loop before showing the initial
connection dialog.
Manuel Nickschas [Thu, 14 Jun 2018 17:58:21 +0000 (19:58 +0200)]
icons: Fix icons not showing for format color buttons
Adding a menu to a button seems to clear an icon set previously
until the button is clicked. Fix this by reordering things.
Manuel Nickschas [Thu, 14 Jun 2018 17:33:38 +0000 (19:33 +0200)]
qtui: Fix tooltip and change wording in Appearance settings
Let the tooltip for "Fallback icon theme" wrap. Change the wording
from "Invert colors" to "Invert brightness", which is what the option
really does.
Manuel Nickschas [Wed, 13 Jun 2018 22:03:23 +0000 (00:03 +0200)]
qtui: Properly update tray icons
Refresh tray icons in case the icon theme is changed, or when
switching between normal and inverted tray icons.
Manuel Nickschas [Wed, 13 Jun 2018 21:47:52 +0000 (23:47 +0200)]
qtui: Refactor the system tray implementations
Replace the mess of virtual methods and init() and other weird
shenanigans by a cleaner architecture. Handle mode/state changes
in the base class and let the implementations explicitly act on
changes. This should make it much clearer how the implementations
behave if mode or state change.
Simplify the D-Bus parts of StatusNotifierItem quite a bit by
keeping the SNI watcher always alive; no need to kill it when
unregistering. Add error handling to the final D-Bus registration
call and fall back to the legacy icon if something went wrong.
Manuel Nickschas [Wed, 13 Jun 2018 20:57:37 +0000 (22:57 +0200)]
icons: Add hicolor fallback for Qt < 5.5
Turns out that Qt only added support for split icon themes in 5.5,
so with older Qt versions our icon theme support doesn't work
properly.
Install hicolor icons for Qt < 5.5, as we already do for Qt 4.
Add a hicolor icon resource, too, and make use of it in the system
tray implementations, which is arguably the most important place
to have proper icon support.
This adds around 130 KiB to the binary size, but should save a lot
of headaches.
Manuel Nickschas [Tue, 12 Jun 2018 20:56:59 +0000 (22:56 +0200)]
qtui: Switch to tray-specific icons
Change icon names for the system tray to use the tray-specific
icons. Add support for inverted colors, too.
Manuel Nickschas [Tue, 12 Jun 2018 20:00:01 +0000 (22:00 +0200)]
qtui: Use icon names in tray implementations
Instead of keeping QIcon instances in the base class, have a central
place to define the icon names for each state, and use that in the
implementations.
Manuel Nickschas [Mon, 11 Jun 2018 22:06:41 +0000 (00:06 +0200)]
icons: Update application icon
Use the ones generated by justJanne, rather than the stock version.
Looks identical.
Manuel Nickschas [Mon, 11 Jun 2018 20:52:16 +0000 (22:52 +0200)]
qtui: Clean up SystemTray
Remove inlines, use override, default-initialize attributes,
reorder methods.
Manuel Nickschas [Mon, 11 Jun 2018 19:07:37 +0000 (21:07 +0200)]
icons: Fix icon theme support for StatusNotifierItem
If SNI is used, the tray icons are passed along to the visualizer
by name rather than as a pixmap (the SNI protocol would support
pixmaps too, however that is a) discouraged and b) not supported
e.g. in Unity).
The SNI interface supports a more or less undocumented attribute
IconThemePath, which may contain an additional path for the
visualizer to look into for specific icons. At least Plasma and
LXQt seem to expect a hicolor theme in the given directory.
Have StatusNotifierItem create a skeleton hicolor theme containing
the tray icons in a temporary directory, and pass that along via
IconThemePath. The tray icons to be saved are dynamically determined
using the current icon theme/fallback.
All available icon sizes are saved as pixmaps; sadly, with the dynamic
approach, scalable icons cannot be supported. Alternatively, one could
hard-code the relevant .svg files from the fallback themes, and copy
those, but that would just be annoying - and kill themeability.
Manuel Nickschas [Wed, 6 Jun 2018 22:49:20 +0000 (00:49 +0200)]
interfaces: Fix types in SNI interfaces, remove old annotations
Not sure why, but the pixmap types in the SNI interface were all
wrong (didn't seem to cause any issues, though, for some reason...).
Fix this.
Remove old com.trolltech.QtDBus.QtTypeName annotations, which are no
longer needed in any version of Qt we still support.
Daniel Albers [Tue, 5 Aug 2014 10:36:51 +0000 (12:36 +0200)]
Fix notification icon not appearing in Unity
Unity's indicator expects certain properties to be exported by an
application, of which Quassel is lacking some. This adds IconThemePath,
Menu, ItemIsMenu to org.kde.StatusNotifierItem.
Thanks a lot to shiznix for providing this patch.
Fixes #1269
(cherry picked from commit
3939820389d995f56a073fc1d54ab245c6558312)
Manuel Nickschas [Mon, 11 Jun 2018 18:37:05 +0000 (20:37 +0200)]
icons: Install Quassel-specific icons in hicolor for Qt4
Qt4 sadly doesn't support split icon themes yet, so our injection
and mechanism does not work.
Install the Quassel-specific icons into hicolor, so they should be
found at least in a proper system installation. Use the Oxygen
theme, because its layout matches the hicolor one (simplifying
the install rule), and it was default for Qt4 anyway.
Manuel Nickschas [Tue, 5 Jun 2018 21:20:58 +0000 (23:20 +0200)]
cmake: Clean up icon-related build options
Remove -DWITH_BREEZE, -DWITH_BREEZE_DARK and -DWITH_OXYGEN options
from CMake. Instead, provide two new options that should make things
a bit clearer:
-DWITH_BUNDLED_ICONS now controls if bundled icon themes should be
installed/embedded. Defaults to ON to ensure that required
icons are available unless explicitly disabled (e.g. by a
package maintainer, who should then make sure that a dependency
to Breeze/Oxygen exists).
Disabling this option saves less than 2 MB of disk space.
-DWITH_OXYGEN_ICONS controls whether the Oxygen icon theme should
still be supported. This affects both the Oxygen-themed
Quassel-specific icons and the bundled icon theme. Defaults to
OFF when building against Qt5, ON for Qt4.
Adapt code accordingly, and update INSTALL.
Manuel Nickschas [Mon, 4 Jun 2018 20:13:12 +0000 (22:13 +0200)]
qtui: Fix icons in the AboutDlg
With the icon reorganization, no aliases are created inside resource
files anymore. Explicitly put the Breeze and Oxygen Quassel pixmaps
for the AboutDlg into pics/ and add it to the :pics/ resource.
Remove pics/quassel.png which was shipped, but not actually used
due to it missing from the resource file.
Manuel Nickschas [Sun, 3 Jun 2018 23:16:01 +0000 (01:16 +0200)]
qtui: Set proper icon for "About Quassel" menu option
With the icons reorganization, we no longer have fallbacks available
on a resource path. Use QIcon::fromTheme() instead.
Manuel Nickschas [Sun, 3 Jun 2018 23:14:43 +0000 (01:14 +0200)]
qtui: Set default window icon independent of connection state
Manuel Nickschas [Mon, 4 Jun 2018 22:07:01 +0000 (00:07 +0200)]
qtui: Fix icon loading and improve icon theme support
Quassel relies on quite a number of icons that are part of KDE/Plasma
icon themes such as Breeze and Oxygen, but are not specified by
freedesktop.org's Icon Naming Specification. As such, icon themes
other than the afforementioned most likely don't contain all the
icons Quassel needs.
Additionally, Quassel ships several additional icons that are not part
of any theme, but are made to fit in with Breeze/Oxygen.
In order to ensure that all required icons are available, we now
assume that at least one of Breeze, Breeze Dark or Oxygen themes is
available as a fallback, either as a system-wide installation, or by
having Quassel itself deploy the required subset in its data dir
(related build options will be adapted in a follow-up commit).
Allow the user to configure which fallback theme is to be used,
and if the system icon theme should be completely overridden (as
opposed to just filling in missing icons). If the system theme
is Breeze/Oxygen, just inject the corresponding Quassel-specific
icons. If the configured system theme doesn't match the selected
fallback (and should not be overridden), create a proxy icon theme
that inherits from first the system theme and then the selected
fallback. That way, missing icons are found through the inheritance
chain.
Note that as of Qt 5.11.0, icons from the hicolor default theme are
preferred over the ones from inherited themes, which violates the
Icon Theme Specification (cf. QTBUG-68603). This affects Quassel's
application icon, which will thus always use the Breeze variant
if a non-supported system icon theme is used. Other icons will be
picked from the selected fallback theme, as expected.
Manuel Nickschas [Tue, 15 May 2018 22:36:44 +0000 (00:36 +0200)]
qa: Use override in QtUi
Manuel Nickschas [Tue, 15 May 2018 20:13:00 +0000 (22:13 +0200)]
icons: Use more common names for irc-channel-* icons
irc-channel-joined and irc-channel-parted are not shipped with Oxygen.
Use irc-channel-active and irc-channel-inactive instead, which are
supported by both Oxygen and Breeze (and are identical to the previous
ones in Breeze).
Manuel Nickschas [Mon, 4 Jun 2018 20:38:01 +0000 (22:38 +0200)]
icons: Fix message-quassel-tray-inverted icon
Use the correct base shape for this icon.
Manuel Nickschas [Tue, 15 May 2018 19:40:47 +0000 (21:40 +0200)]
icons: Update bundled Breeze and Breeze Dark icon themes
Update to latest upstream commit.
Manuel Nickschas [Tue, 15 May 2018 19:14:09 +0000 (21:14 +0200)]
icons: Adapt and improve icon import script
No longer require running it from inside a specific directory.
Remove theme-specific blacklists, as they are not needed anymore.
Modernize and improve the code a bit.
Manuel Nickschas [Mon, 14 May 2018 21:59:32 +0000 (23:59 +0200)]
icons: Remove obsolete hicolor icons
The hicolor directory should only contain the application icon
for external use. Remove other icons (which are no longer used
by Quassel anyway).
Move irc-channel-joined and irc-channel-parted into the quassel-
specific oxygen icon set, as they're not part of the upstream theme.
Manuel Nickschas [Mon, 14 May 2018 21:44:21 +0000 (23:44 +0200)]
icons: Reorganize icon directories
Move bundled icon themes into 3rdparty/icons for clarity. Move
Quassel-specific icons from icons/extra-icons into icons proper.
Rename breezedark to breeze-dark to match the upstream directory
name. Update corresponding qrc files accordingly.
Note that this breaks icon loading until the code is adapted to
properly support icon themes and fallback mechanisms. This will
be fixed in a follow-up commit. Since then Quassel itself will
no longer use the hicolor icons (which are only intended to contain
the application icons for use by desktop environments), there is
no need to bundle them anymore.
The import_theme.pl script also still requires adaptations that will
be delivered in a later commit.
romibi [Sat, 17 Mar 2018 17:06:29 +0000 (18:06 +0100)]
Add a README to the icons folder
romibi [Thu, 1 Feb 2018 22:47:30 +0000 (23:47 +0100)]
Update qrc's for new Icons and update Import Script
romibi [Thu, 1 Feb 2018 22:42:36 +0000 (23:42 +0100)]
Update Quassel-Icons based on justjannes icon-repo
See https://github.com/justjanne/quassel-icons commit
18a8974
inkl. new slim Tray-Icons
inkl. Inverted Tray Icons (Light to Dark and vice versa)
Updated Quassel App Icon
Put Extra Icons in a separate folder in Resources/Data
Updated quassel.ico and .icns
romibi [Thu, 1 Feb 2018 22:50:44 +0000 (23:50 +0100)]
Ship light and dark Breeze Icon theme by default
romibi [Mon, 5 Feb 2018 17:51:54 +0000 (18:51 +0100)]
Move systray animation settings to notification settings page
romibi [Mon, 5 Feb 2018 17:40:37 +0000 (18:40 +0100)]
Split systray notification backend
into SystrayAnimationNotificationBackend for tray icon animation
and SystrayNotificationBackend for tray message popup
Manuel Nickschas [Mon, 4 Jun 2018 21:44:32 +0000 (23:44 +0200)]
qtui: Properly fix the SettingsDlg layout
We've suffered from the page tree in the settings dialog being
squashed horizontally for a long time. A previous commit (
65c463f0)
tried to fix this by calculating a minimum size and explicitly
resizing the dialog, but this didn't properly work at least on HIDPI
displays.
Instead, fix layouting properly by setting the correct size policy
for the tree widget, and removing both the afforementioned workaround
and a previous hack that tried to set the minimum width manually
(and failed due to a still-hidden widget reporting wrong size hints).
Shane Synan [Wed, 6 Jun 2018 19:02:40 +0000 (14:02 -0500)]
mono: Move Local Highlights -> Legacy Highlights
Rename "Local Highlights" as "Legacy Highlights", and rename
"Remote Highlights" as plain "Highlights". Update the info-bar text
and other references as appropriate.
This reduces confusion over which highlights to configure in
Monolithic mode.
Manuel Nickschas [Wed, 30 May 2018 23:16:47 +0000 (01:16 +0200)]
mono: Show a popup during database migration
Database upgrades may take a long while, during which the monolightic
client can not be used. Show a modal popup during migration to both
inform the user about what's going on, and preventing them from
interacting with the client until the internal core is up and running.
Show an "Initializing..." message in the status bar while the core
is starting up, so there's some indication for the user besides the
modal popup.
Manuel Nickschas [Wed, 30 May 2018 22:34:13 +0000 (00:34 +0200)]
mono: Run the internal core in a separate thread
Prevent the internal core from blocking the UI of the monolithic
client by running it in a separate thread. This is particularly
important during startup, where e.g. a database upgrade may block
the core for a significant time, preventing the main window from
being rendered, or even showing up.
Manuel Nickschas [Wed, 30 May 2018 22:25:00 +0000 (00:25 +0200)]
core: Only load translations if running stand-alone
For the monolithic client, the UI part takes care of loading the
translations configured by the user; the core should not override
them.
Only if the core is running stand-alone, it needs to load
translations itself (using the system locale, since there is no
core-side configuration for this).
Manuel Nickschas [Wed, 30 May 2018 20:20:17 +0000 (22:20 +0200)]
core: Properly parent QObject-derived attributes
In order to be able to move the Core instance to another thread,
all QObject-derived attributes must be properly parented, so they
move with their parent object.
Manuel Nickschas [Wed, 30 May 2018 20:18:55 +0000 (22:18 +0200)]
core: Change signature of Core::syncTimer()
Returning a pointer is more consistent with the rest of the code
base and established patterns than returning a non-const reference.
Manuel Nickschas [Wed, 30 May 2018 20:01:50 +0000 (22:01 +0200)]
common: Make InternalPeer (more) thread-safe
The two internal peers used by a mono client live in different
threads. Previously, they each stored a pointer to their counterpart
to be able to send protocol messages wrapped in an event.
While this has never been observed in the wild, this may cause subtle
lifetime issues if one of the sides goes away before the other.
While currently, as the two peers are at least created in the same
thread (and thus, setting the mutual pointer should not cause issues),
this will change in the near future, since the mono core is slated
to move into its own thread.
Rather than trying to deal with this ourselves, let Qt do the work.
Instead of storing a pointer and sending events, use signal/slot
connections to transfer the messages between the peers. Qt will
ensure that this happens in a thread-safe manner, and that connections
are killed if one of the peers is destroyed.
Manuel Nickschas [Tue, 29 May 2018 22:05:07 +0000 (00:05 +0200)]
core: Use QCoreApplication::exit() instead of just exit()
Ensure that the application is cleanly terminated (by stopping
the main event loop instead of just exiting) when core
initialization fails. This allows for proper cleanup, shutting
down threads etc.
Manuel Nickschas [Tue, 29 May 2018 21:11:07 +0000 (23:11 +0200)]
core: Simplify core initialization logic
Move all core initialization into Core::init(), including things
that were previously done in CoreApplication and MonoApplication.
Make several functions private and non-static that no longer need
to be called from the outside.
Allow to construct and destroy the core directly, rather than
relying on Core::instance() and Core::destroy(). Explicitly
create and destroy the instance in the application to better control
lifetime.
Ensure core state is saved and storage synced on destruction, instead
of relying on explicit calls (which were partially missing for mono).
Manuel Nickschas [Tue, 29 May 2018 21:18:05 +0000 (23:18 +0200)]
common: Ensure the custom config path is stored as an absolute path
This prevents creation of a bogus and empty directory inside the
custom config directory if the path given by --configdir was
relative.
Manuel Nickschas [Tue, 29 May 2018 21:15:57 +0000 (23:15 +0200)]
common: Set umask in main() instead of Core::init()
Ensure that a restricted umask is set for all binaries, before
creating any files.
This fixes the fact that client-side configuration files, including
the configuration directory itself, were created with world-readable
permissions if the system's umask was set accordingly.
Manuel Nickschas [Tue, 29 May 2018 19:26:24 +0000 (21:26 +0200)]
Remove ancient settings migration code
Nobody should be running Quassel 0.3 anymore in this decade, so
remove ancient and messy config location migration code.
Shane Synan [Fri, 1 Jun 2018 03:26:43 +0000 (22:26 -0500)]
client: Remap "Reload Stylesheet" to Ctrl+Shift+R
Move "Reload Stylesheet" to Ctrl+Shift+R instead of Ctrl+R from
QKeySequence::Refresh. "Set Marker Line" maps to Ctrl+R by default,
resulting in conflict.
Reloading the stylesheet shouldn't be done often enough to warrant
the easier-to-press key combination.
Shane Synan [Fri, 1 Jun 2018 03:24:07 +0000 (22:24 -0500)]
client: Add keyboard shortcuts for formatting
Add keyboard shortcuts for formatting options as follows:
* Apply foreground color: Ctrl-Shift-G
* Apply background color: Ctrl-Shift-B
* Clear formatting: Ctrl-Shift-C
* Toggle bold: Ctrl-B (from QKeySequence::Bold)
* Toggle italic: Ctrl-I (from QKeySequence::Italic)
* Toggle underline: Ctrl-U (from QKeySequence::Underline)
Shane Synan [Fri, 1 Jun 2018 03:13:07 +0000 (22:13 -0500)]
client: Add input tooltips, clear format, cleanup
Add tooltips to all of the input widget items, improving
accessibility for otherwise icon-only buttons and menus.
Add button to clear formatting, making it easy to immediately remove
all those garish foreground and background colors you added, and the
italic/bold/underline, too.
Cleanup formatting options into distinct functions, paving the way
for external signals changing formatting.
Shane Synan [Fri, 1 Jun 2018 00:20:00 +0000 (19:20 -0500)]
client: Clarify QKeySequence::Quit matter in docs
QKeySequence::Quit does not set "Ctrl-Q" on Windows, while many
Windows programs use that to provide Quit, including Quassel IRC.
Modify the comment to explain the situation.
See https://doc.qt.io/qt-5/qkeysequence.html
Shane Synan [Thu, 31 May 2018 23:21:33 +0000 (18:21 -0500)]
core: Save schema version for intermediate steps
When running upgrade queries, update the schema version for each
intermediate step. This ensures that any interrupted upgrades have a
greater chance of resuming correctly after core restart.
Almost all databases make single queries atomic (fully works or fully
fails, no partial), and with many of the longest migrations being a
single query, this makes upgrade interruptions much more likely to
leave the database in a valid intermediate schema version.
As a side effect, downgrading from a partial upgrade will now require
setting schema version manually, instead of it possibly happening to
work depending on what changed. An unsupported operation failing in
a known, recoverable way seems better than failing in an unknown,
possibly recoverable way.
Update logging to be more explicit about what fails.
Tested by kill -9'ng Quasselcore during migration, then re-running
core afterwards.
In the future, for databases that support it (e.g. almost only
PostgreSQL), we may want to wrap upgrades in a transaction. This
will need careful testing of potential additional space requirements
and any database modifications that might not be allowed in a
transaction.
sqlite> select * from coreinfo;
schemaversion|26
$ (sleep 0.5 && pkill -9 quasselcore ) & \
./run-profile.sh master core local
sqlite> select * from coreinfo;
schemaversion|30
$ ./run-profile.sh master core local
sqlite> select * from coreinfo;
schemaversion|31
Shane Synan [Wed, 30 May 2018 00:27:37 +0000 (19:27 -0500)]
common: Add '$i:identd', '*' for empty, tooltips
Add "$i:identd" paramater variable to aliases, representing the ident
if verified, or "*" if unknown or unverified (prefixed with '~').
This allows for a general banning rule to target all idents for a
nickname that doesn't provide verified idents (e.g. general clients),
while limiting bans to specific idents for a nickname that provides a
verified ident (e.g. shared bouncer/core).
For example...
/mode channel +b *!$1:identd:$1:hostname
...would target "*!*@bad.example.com" for someone connecting without
an identd server, and "*!userid@shared.example.com" for someone
connecting with an identd server.
For all nickname-based parameters, substitute in "*" when empty
instead of leaving it blank. This means using the variables for
nicknames that haven't been WHO'd will still provide results.
This may be dangerous if using the above example /ban alias on a
nickname that doesn't have hostname information available, e.g. if
you joined a channel after they did.
Update alias tooltips with 'identd', and totally revamp them using
the table layout for better readability and (hopefully?) easier
translation.
Update the comments in AliasManager, too, so it's a bit more
readable.
Thanks Exterminador for the suggestion/reporting, and @justJanne for
further clarifications!
Shane Synan [Tue, 29 May 2018 18:36:31 +0000 (13:36 -0500)]
core: Include timezone for CTCP TIME replies
Explicitly specify the default Qt::TextDate date/time format to allow
modifying it.
On Qt 5, include the timezone data. This doesn't reduce privacy as
timezone data could be figured out before by manually comparing with
UTC, this just makes it easier for those who want that information
shared.
See https://doc.qt.io/qt-5/qdatetime.html#toString
And https://doc.qt.io/qt-5/qt.html#DateFormat-enum
Shane Synan [Tue, 29 May 2018 02:17:10 +0000 (21:17 -0500)]
common: Sort user prefix channelmodes on add/set
Sort prefix channelmodes whenever adding/removing/setting per-channel
IrcUser modes. This keeps the highest mode at the front, necessary
for backlog storage to allow only showing the highest mode.
It's not possible to get away without sorting as modes can be added
in any order.
modesToPrefixes/prefixesToModes operate in order, so sorting prefixes
results in sorted modes for free. IrcUser modes don't need sorted as
they're not hierarchical and not stored anywhere.
Shane Synan [Tue, 29 May 2018 16:51:35 +0000 (11:51 -0500)]
Rename strict-ident, apply to non-identd response
Rename "--oidentd-strict" to "--strict-ident", better representing
what it now does - makes all ident replies strict.
Apply strict ident matching to normal (non-identd-enabled) ident,
too. This allows strict ident to be enforced on IRC networks that do
not check identd responses.
(Yes, some IRC networks behave this way, expecting your shared user
bouncer to enforce ident. ZNC allows this, so Quassel probably
should too. No need to run an ident server for those networks.)
Clean up functions, making them a little more generic. This also
sets the stage for exposing the strict ident mode to the client later
on.
Michael Marley [Wed, 9 May 2018 03:17:21 +0000 (23:17 -0400)]
Fix crash caused by repeated kicks of an already-kicked client
As it turns out, the QHash[] operator implicitly adds nullptrs to
the map when called with a nonexistent key. If a user clicked the
button to kick another client multiple times, this would cause a
nullptr to be inserted into the peerMap, then causing a crash later
on at least when another client connected. To solve the problem,
the value() method is used instead, which has no such side-effect.
A big thanks to @digitalcircuit for discovering this behavior.
Shane Synan [Mon, 14 May 2018 06:00:19 +0000 (01:00 -0500)]
client: Fix core sessions unsupported warning
Move core session supported check to CoreInfoDlg, and show a
warning bar if sessions aren't supported. A "Details" button
offers additional information on needing to upgrade.
Older core versions didn't provide any core session information, so
the individual session widgets would have never shown up.
Thanks to @justjanne for reporting this.
Shane Synan [Wed, 9 May 2018 01:28:29 +0000 (20:28 -0500)]
client: Fix Core Info session widget spacing
Fix spacing for Core Info session widget by keeping only one stretch
layout item and inserting before it, rather than manually adding
stretches after every client list update.
See https://doc.qt.io/qt-5/qboxlayout.html#insertWidget
Janne Koschinski [Fri, 2 Mar 2018 03:04:50 +0000 (21:04 -0600)]
Automatically synchronize CoreInfo when on connect and disconnect
This refactors CoreInfo to be a single common class rather than
separate classes for the core and client, adjusts it to properly
use the SyncableObject infrastructure, and adds support for
dynamically updating the list on both the client and core when a
client connects or disconnects, keeping the list up-to-date at all
times. Forwards and backwards compatibility is maintained, though
the updated client is required in order for the automatic updating
to work.
Signed-off-by: Janne Koschinski <janne@kuschku.de>
Co-authored-by: Janne Koschinski <janne@kuschku.de>
Co-authored-by: Michael Marley <michael@michaelmarley.com>
Shane Synan [Wed, 28 Mar 2018 21:42:02 +0000 (16:42 -0500)]
client: Show when End Session unavailable, polish
Show when the End Session option is unavailable, adding a tooltip to
either point out an outdated core, or a client that doesn't support
it.
Before RemoteDisconnect was added to the core, it didn't share any
client features with other connected clients, so there's no way to
tell if a client supports a feature or not.
See
78b37c1a0610d9a4fb26ba0dec7337d6f7960041
"Transmit feature flag for remote disconnecting with peer info"
Disable and rename the End Session button when clicked, giving
immediate feedback when Quassel's used with a high-latency connection.
Shane Synan [Mon, 14 May 2018 04:53:57 +0000 (23:53 -0500)]
client: Migrate old highlight Channels to RegEx
Add new settings minor version 8 to migrate existing local highlight
rules to the new format. Specifically, any highlight rule with a
specified "Channel" filter is adapted into a regular-expression rule.
> Before
Name: a word
RegEx: disabled
Channel: (ChannelA|AnotherPlace)
> After
Name: (^|\W)a word(\W|$)
RegEx: enabled
Channel: (ChannelA|AnotherPlace)
This way RegEx channel rules will work by default now that the RegEx
flag controls wildcard vs. reg-ex for channel names, too. And this
specific migration is backwards-compatible with previous client
versions, so downgrading won't break highlight rules.
(Functionality changed in other migration steps might break on
downgrade, however.)
Shane Synan [Mon, 28 May 2018 20:33:05 +0000 (15:33 -0500)]
common: Apply RegEx, CS column to sender, channel
Modify scopeMatch to optionally treat the rule as a regular
expression, and to treat it as case-sensitive or not.
Modify highlight rule processing to...
* Use scopeMatch for "Sender", allowing easy multi-sender rules
* Treat both "Sender" and "Channel" as regular expressions if "RegEx"
is checked, allowing for full regular expression power.
(Breaks existing rules with RegEx matching, but in a different way)
This means it is no longer possible to use RegEx matching for the
"Sender" and "Channel" yet not on the phrase itself. However, it
should be easy enough to work around this by using the same phrase
regular expression Quassel does.
This should simplify common highlight rules without limiting those
who know regular expressions. Win-win, hopefully?
Update tooltips to reflect these changes.
Shane Synan [Fri, 2 Mar 2018 00:51:32 +0000 (18:51 -0600)]
client: Show feedback for Local Highlights import
Prompt before importing local highlight rules to guard against
accidental clicks.
Show a success message at the end indicating how many highlight rules
were imported.
Show an information message if no local highlight rules exist to
import.
Shane Synan [Fri, 2 Mar 2018 00:25:49 +0000 (18:25 -0600)]
client: Explain Local Highlights vs. Remote
Add footer to "Local Highlights" page to note that local highlights
apply to the current device only, including a "Details..." button.
Add a "Details..." dialog with a brief explanation and points out
where to configure highlights for all devices.
For Monolithic builds, indicate that "Remote Highlights" replaces
"Local Highlights" to help avoid confusion. There's no remote core,
so local highlights have no reason to exist other than to provide a
path for upgrades.
Shane Synan [Fri, 2 Mar 2018 16:20:12 +0000 (10:20 -0600)]
client: Show when Remote Highlights unsupported
Add a header to "Remote Highlights" page to warn when the Quassel
core is too old to support core-side highlights, including a
"Details..." to explain the purpose and what version is required.
Core connection is still required to make the UI selectable,
following the convention set by other panels.
This should help avoid people asking about "Remote Highlights" when
their core is too old.
Wrap the rest of the settings page in a widget to simplify enabling
and disabling all of it.
Shane Synan [Thu, 1 Mar 2018 12:48:47 +0000 (06:48 -0600)]
client: Add Highlight Rules tooltips, UI polish
Add tooltips to every element of Highlight/Highlight Ignore tables,
and to the Local Highlight table.
Update the "Channel" tooltip to describe the non-regular-expression
approach used now.
Rename "Import" to "Import Local", add tooltip to better describe
purpose.
Move "Highlight Ignore Rules" table into QGroupBox to place at even
height with "Highlight Rules" box, making it less jarring to switch
between the two.
Shane Synan [Thu, 1 Mar 2018 10:06:22 +0000 (04:06 -0600)]
client: Unify Local/Remote Highlights table order
Put "Enabled" at front of Local Highlights table column order,
matching the ordering and name of Remote Highlights.
Shane Synan [Thu, 1 Mar 2018 08:29:29 +0000 (02:29 -0600)]
client: Fix Local Highlights import to remote
Make "importRules()" a private slot, not just a private function.
This allows the button signal to connect to this function, making
imports work.
Shane Synan [Thu, 1 Mar 2018 23:59:50 +0000 (17:59 -0600)]
Really use Remote Highlights for nick, not Local
Really disable local nick highlights by default. This will affect
those who upgrade and haven't changed the default highlight settings.
No settings migration provided since we intend to migrate to
core-side highlights entirely in the future.
Really enable remote nick highlights by default, setting a default
value for when highlights are not yet set up. Fixes behavior of no
nickname highlighting by default.
Shane Synan [Thu, 1 Mar 2018 08:06:42 +0000 (02:06 -0600)]
common: Put scopeMatch in util, use in highlights
Move ignore rule scopeMatch() to util to share usage with highlight
rule scope matching.
Use scopeMatch() in highlight rule channel name processing.
Advantages:
> Simpler to manage common cases
> Shares the same rules as ignore lists, already-learned behavior
Disadvantages:
> No way to escape ';' in rules
> No longer able to use full regular expressions
> Breaks existing channel name rules (also applies to local!)
(Another approach is to keep only local highlight rules inconsistent)
Shane Synan [Thu, 1 Mar 2018 05:54:10 +0000 (23:54 -0600)]
common: Add inverted scope match rules to ignores
Add support for inverted scope matching subrules to ignore entries.
A match now only happens when the string does NOT match ANY inverted
rules and matches AT LEAST one normal rule, unless no normal rules
exist (implicit wildcard match).
Example:
> #quassel*; #foobar; !#quasseldroid
Matches "#foobar" and anything "#quassel...", except for
"#quasseldroid"
Example:
> !#quassel*; !#foobar
Matches anything except for "#foobar" or anything "#quassel..."
Shane Synan [Wed, 28 Feb 2018 01:34:10 +0000 (19:34 -0600)]
client: Disable case-sensitive check for no nick
Disable "Case sensitive" highlight checkbox when highlight nicks is
set to "No nicks", as it has no effect on outcome.
genius3000 [Wed, 23 Nov 2016 09:04:46 +0000 (02:04 -0700)]
Add a Configure menu item to Networks in the buffer list
Add 'Configure' above 'Disconnect' in the right-click menu from a
network in the buffer list. Clicking 'Configure' will open the
Configure Networks dialog to the selected network.
Fixes #607
Manuel Nickschas [Sun, 3 Jun 2018 22:04:47 +0000 (00:04 +0200)]
common: Notify if key was removed from settings
Settings have a notification feature so consumers can subscribe to
value changes. This notification should also be emitted in case
a settings key is removed altogether.
Manuel Nickschas [Wed, 23 May 2018 22:24:28 +0000 (00:24 +0200)]
common: Improve logger configuration
Initialize logging earlier, so output during core initialization
gets routed properly.
Enable the logging options also for quasselclient; there's no good
reason to limit this to the core and monolithic client. However, as
a side-effect, the default log level is now Info also for the client,
so "-l Debug" must be specified on the command line to get debug
output.
Manuel Nickschas [Wed, 23 May 2018 21:08:54 +0000 (23:08 +0200)]
common: Use override instead of virtual
Avoids a warning for inconsistent overrides.
This commit finishes up the giant rollup PR provided by
@digitalcircuit to consolidate a number of conflicting and
interdepending PRs. Thanks a lot for all the work!
Closes GH-357.
Manuel Nickschas [Wed, 23 May 2018 21:01:41 +0000 (23:01 +0200)]
common: Remove the lastAwayMessage Q_PROPERTY declaration
Since this property is only synced by pre-0.13 cores, and not used
by any released clients, it doesn't need to be marked as a Q_PROPERTY.
Also avoids a warning for the property not having a getter.
Shane Synan [Sun, 13 May 2018 20:46:11 +0000 (15:46 -0500)]
core: Migrate SQLite schema to MSec, warn of slow
Migrate SQLite schema to read and store time in milliseconds,
offering greater precision matching that of PostgreSQL.
On upgrade, run "UPDATE backlog SET time = time * 1000" to adjust
existing times. This is a full backlog update, potentially taking
minutes to hours.
Based on testing, this was the fastest "proper" way to migrate an
existing database. A pessimistic scenario, 328 MB database on a
Raspberry Pi 2 with Ubuntu 16.04 running from a Class 10 SD card
takes...
> In-place update (chosen method): 361.991s
> Create new table, multiply int: 905.749s
> Create new table, numeric: 983.501s
A normal HDD is much faster, e.g. 97.686s for the same database on a
5400 RPM drive, Seagate ST96812AS.
(Thanks @romibi for testing!)
Other methods may be faster, but may be more prone to mistakes down
the road.
Update the schema upgrade message to warn of delays for major
upgrades, and provide feedback when upgrade succeeds.
> Installed Schema (version %1) is not up to date. Upgrading to
> version %2... This may take a while for major upgrades.
> Installed Schema successfully upgraded to version %1.
CAUTION: Monolithic does not show any migration GUI. Monolithic
installs with large databases on slower computers may appear to not
be responsive. Include a mention of upgrade time in the release
notes.
Michael Marley [Sat, 12 May 2018 20:20:58 +0000 (16:20 -0400)]
IrcUser: Rename lastAwayMessage to lastAwayMessageTime
This will make it easier for clients to implement both at the same
time. Also, add a backwards-compatibility function to allow old
cores to set the lastAwayMessageTime. There is still no forwards-
compatibility (support for old clients on new cores) since no
existing clients do anything with lastAwayMessage.
Michael Marley [Thu, 10 May 2018 18:32:31 +0000 (14:32 -0400)]
IrcUser: Use QDateTime for lastAwayMessage, not int
This converts the lastAwayTime to a 64-bit time. Backwards-
compatibility is also included in Network to convert back and forth
between the old 32-bit type for legacy peers. No backwards-
compatibility is provided for the SYNC call because the value is
never actually used in any existing clients. New clients that want
to implement this functionality would need to implement the
LongTime feature.
Shane Synan [Fri, 11 May 2018 00:20:12 +0000 (19:20 -0500)]
Finish 64-bit time conversion, modify protocol
Convert the remaining uses of 32-bit quint-based
"[to|from|set]Time_t()" to the modern 64-bit qint64-based
"[to|from|set]MSecsSinceEpoch()", handling the Y2038 problem while
also gaining higher precision in some places.
For places where seconds are needed, e.g. converting incoming time
from the IRC protocol, make use of "[to|set|from]SecsSinceEpoch()"
instead. As this was only added in Qt 5.8, use the "QT_VERSION"
macro for backwards compatibility, manually dividing by 1000 for the
"MSecs" function equivalents.
Beware, all new code should use "[to|from|set]MSecsSinceEpoch()" and
NOT use "[to|from|set]Time_t()", or you'll re-introduce Y2038 bugs!
Modify protocol to rename "LongMessageTime" to "LongTime", and
send/receive all times as "qint64" instead of "quint64" or "quint".
"LongMessageTime" was never in Quassel git master, so renaming it
should have no impact.
Modify Event timing to be 64-bit compatible, too.
"IrcUser::lastAwayMessage()" will be fixed in a follow-up commit.
SQLite millisecond precision will be fixed in a follow-up commit.
See https://doc.qt.io/qt-5/qdatetime.html
And https://doc.qt.io/qt-5/datastreamformat.html
And https://doc.qt.io/qt-5/qdatetime-obsolete.html
Shane Synan [Tue, 8 May 2018 23:24:38 +0000 (18:24 -0500)]
Change Message::Flags/Type cast to Int, not UInt
Use toInt() for casting `Message::Flags` and `Message::Type`, instead
of toUint().
MSVC has trouble with uint-based enums, so Quassel switched to int
enums. However, not all of the enum casts were converted, resulting
in needless unsigned-to-signed casting.
Thanks to @justJanne for reporting this.
Shane Synan [Sun, 13 May 2018 22:07:07 +0000 (17:07 -0500)]
core: Fix 64-bit ID SQLite -> Postgres migration
Change skipSteps and stepSize() from 32-bit int to qint64, avoiding
integer wraparound during backlog reading for migration.
This only gets triggered when a message with a 64-bit ID exists in
the SQLite backlog before migration. One can force this by sending
a message with the text 'MARKER_MSG', then running this on the SQLite
database before migration.
INSERT INTO backlog (messageid, time, bufferid, type, flags,
senderid, senderprefixes, message)
VALUES (
5000000000,
5000000000000,
(SELECT bufferid FROM backlog WHERE message = 'MARKER_MSG'),
1, 1,
(SELECT senderid FROM backlog WHERE message = 'MARKER_MSG'),
"", "Test 64-bit message");
> Before, debug output
SQLi::readMo - skipSteps 42948, start
2147400383, stop
2147450383
SQLi::readMo - skipSteps 42949, start
2147450383, stop -
2147466913
SQLi::readMo - skipSteps 42950, start -
2147466913, stop -
2147416913
...
SQLi::readMo - skipSteps 85898, start -66913, stop -16913
SQLi::readMo - skipSteps 85899, start -16913, stop 33087
SQLi::readMo - MsgId: 1, message: Connecting to chat.freenode.net:...
PSql::writeMo - MsgId: 1, message: Connecting to chat.freenode.net...
> After, debug output
SQLi::readMo - skipSteps 99999, start
4999950383, stop
5000000383
SQLi::readMo - MsgId:
5000000000, message: Test 64-bit message
PSql::writeMo - MsgId:
5000000000, message: Test 64-bit message
Michael Marley [Fri, 9 Mar 2018 17:17:05 +0000 (12:17 -0500)]
Use 64-bit IDs for messages
Increase the length of message IDs from 32 to 64 bits, allowing for
many more total messages be stored. Because messages are synced to
the client, this requires a protocol change. The protocol change is
guarded behind a new feature flag that will retain backwards
compatibility with existing clients, at least until the IDs go above
the maximum for a 32-bit number. At this point, legacy clients will
behave incorrectly or perhaps crash, but this is an acceptable risk
because in the current implementation, this event would cause the
core to behave incorrectly or crash. The client is also still
limited to 32 bits worth of backlog per buffer displayed at once
(due to Qt limitations), but it would be highly unlikely to hit
this limit because the client becomes unusably slow with a small
fraction of that amount of backlog loaded in a single buffer.
For SQLite, no schema change was necessary as SQLite makes no
differentiation between the bitnesses of integers.
For PostgreSQL, this patch also changes senderId to 64-bit, but
this has no effect outside the PostgreSQL database driver.
This patch only changes the message ID to 64-bit. All other IDs
are left at 32 bits because it is either exceedingly unlikely or
technically infeasible to have more than 32 bits worth of entries
in any of the other tables. If, at some point, it is decided that
other IDs should also be 64-bit, the way in which the feature flag
affects (de)serialization will need to be changed. This is because
the flag LongMessageId currently acts upon all IDs inheriting from
SignedId64 and it would be necessary to independently control the
(de)serialization of each ID if other IDs became 64-bit in order to
continue to maintain backwards compatibility. One possible method
of doing this would be to subclass the SignedId64 class for each
ID that inherits from it to allow for separately controlling
(de)serialization for each of the types. I have not implemented
this, however, because of the unlikelihood of any of the other IDs
ever being made 64-bit.
Closes GH-343.
Janne Koschinski [Tue, 8 May 2018 21:49:11 +0000 (16:49 -0500)]
core: Allow configuring DB/auth from environment
Add support for environment variables to configure database and
authenticator backends. This improves using Quassel in a container
-based system as no configuration file needs to be written.
Toggle use of environment variables via a new CLI switch,
'--config-from-environment', which will disregard the core settings
file (if it exists) and use environment variables instead.
New environment variables:
> Backend selection
DB_BACKEND = ["SQLite", "PostgreSQL"]
AUTH_AUTHENTICATOR = ["Database", "LDAP"]
> PostgreSQL connection properties
DB_PGSQL_USERNAME
DB_PGSQL_PASSWORD
DB_PGSQL_HOSTNAME
DB_PGSQL_PORT
DB_PGSQL_DATABASE
> LDAP connection properties
AUTH_LDAP_HOSTNAME
AUTH_LDAP_PORT
AUTH_LDAP_BIND_DN
AUTH_LDAP_BIND_PASSWORD
AUTH_LDAP_BASE_DN
AUTH_LDAP_FILTER
AUTH_LDAP_UID_ATTRIBUTE
Closes GH-341.
Janne Koschinski [Sat, 3 Mar 2018 07:45:31 +0000 (08:45 +0100)]
core: Store session state in database, migrate old
Store core session state in database rather than configuration file.
This improves containerization as the configuration file doesn't need
to be preserved, and also makes database restoration handle restoring
active sessions.
Migrate existing legacy sessions to database storage, ensuring that
the core state fallback is handled correctly, too.
Part of GH-341.
Janne Koschinski [Tue, 8 May 2018 20:50:41 +0000 (15:50 -0500)]
Implement type-filtered backlog fetching
Add new backlog fetching methods that only return messages of the
specified types. This allows clients to only fetch the types of
messages that they need, e.g. if join/part/quit is hidden.
This is not yet used in Quassel itself, but third-party clients
including QuasselDroid use this for better performance on limited
bandwidth networks.
Signed-off-by: Janne Koschinski <janne@kuschku.de>
Closes GH-339.
Shane Synan [Fri, 11 May 2018 06:37:55 +0000 (01:37 -0500)]
core: Fix PostgreSQL realname/avatarurl handling
In insert_message, when selecting a senderid coalesce "realname" and
"avatarurl". This fixes inserting redundant senders due to NULL
values, e.g. messages from the server during connect.
Test case: start up core, client. Connect, check PostgreSQL database
"sender" table. Disconnect, reconnect. There shouldn't be new NULL
senders in the table (e.g. sender set, but NULL realname).
NOTE: The realname/avatarurl columns can be NULL values. Due to
this, we need to coalesce them to '' in order to use the same queries
rather than "column = some value" and "column IS NULL". Both column
and the input parameter need coalesced in case one or the other is
NULL. As there's minimal functional difference in protocol handling
between '' and NULL, we consider them the same.
(This could be squashed down if preferred, I just wanted to document
what was needed to fix this for PostgreSQL, and to simplify figuring
out what to change should a better solution arise.)
Shane Synan [Fri, 11 May 2018 05:36:21 +0000 (00:36 -0500)]
core: Fix SQLite realname/avatarurl handling
In insert_message, when selecting a senderid also match on "realname"
and "avatarurl". This fixes cases where sometimes senderid would
refer to a sender without a realname after a realname was added, etc.
Test case: start up core, client. Send several messages quickly after
joining. Shut down core, check SQLite database "sender" table.
Before this fix, once a sender entry is put in, sometimes
permutations of realname/avatarurl would NOT create a new sender
entry, clearly visible in QuasselDroid's display of of the realname.
After this fix, changing realname/avatarurl causes new sender entries
to be added every single time there's a change, as expected, and they
display correctly in QuasselDroid.
PostgreSqlStorage was immune to this, but for once, not due to being
a better database, but due to how the queries are constructed. For
PostgreSQL, senders were always fetched separately, properly checking
avatarurl and realname.
NOTE: The realname/avatarurl columns can be NULL values. Due to
this, we need to coalesce them to '' in order to use the same queries
rather than "column = some value" and "column IS NULL". Both column
and the input parameter need coalesced in case one or the other is
NULL. As there's minimal functional difference in protocol handling
between '' and NULL, we consider them the same.
(This could be squashed down if preferred, I just wanted to document
what was needed to fix this and why only SQLite was changed.)
Janne Koschinski [Tue, 8 May 2018 20:17:38 +0000 (15:17 -0500)]
Implement sender realname/avatarurl storage
- Store sender realname and avatarurl in database
- Guard protocol changes behind feature flag 'RichMessages'
- Avatar URLs are not actually stored yet, as we do not implement
METADATA yet.
This is not yet used in Quassel itself, but third-party clients
including QuasselDroid use this to provide more context in messages
and allow for realname-based fallback avatar fetching to work in the
chat message backlog.
Incorporates suggestions from @mamarley.
Closes GH-351.
Janne Koschinski [Tue, 10 Apr 2018 23:58:09 +0000 (01:58 +0200)]
protocol: Implement 64-bit message timestamps
Add serialization and feature flag for 64-bit message timestamp.
Avoids a Y2038 problem and offers higher message time accuracy.
Closes GH-340.
Janne Koschinski [Tue, 8 May 2018 18:36:25 +0000 (13:36 -0500)]
Store highlight status per buffer coreside
- Add highlightCount to buffersyncer, same as with buffer activities
- If this value is increased, it is synced with clients, and they add
it to the buffer status, showing it in orange
- This allows you to see highlights even if they are not loaded in
initial backlog, but only core-side highlights.
Closes GH-333.
Michael Marley [Tue, 8 May 2018 18:49:33 +0000 (13:49 -0500)]
Persist Blowfish keys in the database
Add a new text field called "cipher" and store the hex-encoded
cipher key there whenever a new key is set or exchanged. Also,
when each network is initialized, load the ciphers out of the
database and initialize the in-memory hashmap. Then, the existing
behavior of each CoreIrcNetwork automatically using these keys
upon construction occurs.
Additionally, this makes PM buffer ciphers persistent both across
destruction/construction and across core restarts.
Note that the existing "key" field in the database is confusingly
named. It does not contain any sort of cryptographic key but
instead holds channel passwords.
Closes #1473
Closes GH-332.
Manuel Nickschas [Thu, 17 May 2018 19:37:46 +0000 (21:37 +0200)]
core: Avoid confusing log message when legacy client connects
Older clients don't report the features they support, which resulted
in a log message listing all features as missing if such a client
connected.
Handle this special case by checking if any client feature is enabled,
and if not, providing a different log message to avoid user
confusion.
Janne Koschinski [Tue, 17 Apr 2018 02:08:10 +0000 (04:08 +0200)]
Fixes the Highlight Rule Manager
- Previously, sync calls included the new HighlightNickType
- This causes deserialization trouble in older clients
- Now int is used, and the data is static_cast'd back and forth
Manuel Nickschas [Mon, 7 May 2018 20:20:10 +0000 (22:20 +0200)]
Some cleanups
* Use properly camel-cased method names
* Use override instead of virtual to avoid warnings
* Simplify some code
Closes GH-348.
Michael Marley [Wed, 4 Apr 2018 12:25:14 +0000 (08:25 -0400)]
Cache strict sysident mappings
To avoid unnecessary database hits, we store a mapping of all permitted
strict sysidents in memory. This mapping is reloaded on SIGHUP.
A miss will make a call to the database.
Co-authored-by: Sai Nane <esainane+github@gmail.com>
Co-authored-by: Michael Marley <michael@michaelmarley.com>
Michael Marley [Tue, 3 Apr 2018 18:21:11 +0000 (14:21 -0400)]
Add strict-oidentd mode
This mode is disabled by default. When specified, this restricts
authoratative ident responses to be the authuser account name.
This is a cut-down version of,
https://github.com/quassel/quassel/pull/164 which was originally
authored by @esainane. @AlD requested that the database schema
change be removed, but the request fell dormant and was eventually
closed. This version removes the schema change as requested,
always using the user's core username as the ident when enabled.
Co-authored-by: Sai Nane <esainane+github@gmail.com>
Co-authored-by: Michael Marley <michael@michaelmarley.com>
Michael Marley [Sat, 31 Mar 2018 18:06:16 +0000 (14:06 -0400)]
Add support for Elliptic Curve keys for CertFP
This functionality is controlled by a feature flag so that clients
supporting EC keys won't try to set an EC key on a core that
doesn't. As with the previous EC patch, this requires Qt 5.5 to
work but uses the QT_VERSION macro to remain compileable with Qt4.
This patch also fixes the QSsl::KeyAlgorithm indices to be correct
for both Qt4 and Qt5. In Qt4, RSA was 0 and DSA was 1, but in Qt5
0 became Opaque, 1 because RSA, and two became DSA. EC is 3 and
wasn't added until 5.5. The new macro handles all three versions
correctly. (See https://doc.qt.io/qt-5/qssl.html#KeyAlgorithm-enum
and https://doc.qt.io/archives/qt-4.8/qssl.html#KeyAlgorithm-enum.)
Thanks to @justJanne for pointing out this discrepancy.
Closes GH-347.