Manuel Nickschas [Sun, 14 Sep 2014 14:48:12 +0000 (16:48 +0200)]
Merge pull request #82 from jpnurmi/legacypeer
LegacyPeer: kill the dependency to Quassel::buildInfo()
ramz [Sun, 30 Mar 2014 11:07:25 +0000 (22:07 +1100)]
Retains Channel UnHiddens Status after client Restart.
Steps to Reproduce:
1. Hide channel temporarily
2. Part channel
3. Join channel (it will now be visible again, at the bottom of that network's buffer list)
Optional: talk in the channel, get other people to talk, etc but the buffer must not be blue (new message(s)) when you proceed next
4. Quit the client
5. When you'll restart it, the channel will be temporarily hidden again
Root Cause:
After step 3, we are not saving the bufferviewconfig synchronizable object right.
Fix:
Save the bufferviewconfig object correctly.
Fixes #1218.
Michael Marley [Thu, 4 Sep 2014 00:12:31 +0000 (20:12 -0400)]
Improve the odds of winning the oidentd race when using SSL for IRC
Previously, the .oidentd.conf file was not written until
socketInitialized(), which meant that when using SSL for the IRC
connection, the file would not be written until the SSL negotiation
was complete. This delay meant that, in practice, oidentd
integration almost never worked when using SSL IRC connections.
This patch adds a new socketConnected() signal that fires as soon
as the connection is open, regardless of encryption status, and
connects the call to the oidentd configuration generator to this
signal. This gives oidentd integration the same likelihood of
working with SSL connections as with non-SSL connections.
Michael Marley [Mon, 1 Sep 2014 22:42:22 +0000 (18:42 -0400)]
Use null QVariant instead of "ALL" for select limit on psql
Using "ALL" as it was before was making PostgreSQL error out when
attempting to return backlog for Quasseldroid when Quasseldroid is
in "Unread messages per chat" mode. According to
http://www.postgresql.org/docs/9.3/static/queries-limit.html,
LIMIT NULL is the same thing as having no LIMIT at all. According
to http://qt-project.org/doc/qt-4.8/qsqlquery.html, NULL values
are bound in Qt SQL prepared statements by using a null QVariant.
Daniel Albers [Fri, 29 Aug 2014 11:03:24 +0000 (13:03 +0200)]
Revert "Fix notification icon not appearing in Unity"
This fixes builds currently failing at runtime with this D-Bus error:
arguments to dbus_message_iter_append_basic() were incorrect,
assertion "_dbus_check_is_valid_path (*string_p)" failed in file
dbus-message.c line 2680.
This reverts commit
3939820389d995f56a073fc1d54ab245c6558312.
Unfixes bug #1269
Daniel Albers [Thu, 28 Aug 2014 07:23:25 +0000 (09:23 +0200)]
Merge pull request #87 from denizt/moc
Fix MainToolBar visibility not saving in OSX
Deniz Türkoglu [Wed, 27 Aug 2014 19:08:57 +0000 (21:08 +0200)]
Fix MainToolBar visibility not saving in OSX
The current code, although completely valid, doesn't work due to moc not
seeing built in defines. See;
https://bugreports.qt-project.org/browse/QTBUG-34593
Manuel Nickschas [Thu, 21 Aug 2014 17:27:08 +0000 (19:27 +0200)]
Merge pull request #86 from seezer/c++11
c++11 - from feint to fact
Sebastian Goth [Thu, 21 Aug 2014 16:31:56 +0000 (18:31 +0200)]
c++11 - from feint to fact
By extending CMAKE_CXX_FLAGS twice for gcc, we kept -ansi which
overrides -std=c++11. Cleaned that up a little.
Allan Jude [Tue, 19 Aug 2014 14:11:49 +0000 (10:11 -0400)]
Update README re: manageusers.py
Update README to reflect the deprecation of scripts/manageusers.py in Aug 2010
tecknojock [Thu, 14 Aug 2014 19:53:30 +0000 (14:53 -0500)]
Fixed CTCP Action Spliting
Previously /me would truncate if it was over the 512 character limit,
forming a malformed action as the last \01 character is cut off. Now the
CTCP Action should autosplit the same way that overflow on normal
messages does. Additionally the characters !? were added to the
splittable characters list, bringing the new list to " ,.-!?"
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
m5 [Sun, 3 Aug 2014 00:03:14 +0000 (17:03 -0700)]
fix buffer sorting for hot buffer hotkey
Daniel Albers [Mon, 28 Jul 2014 14:29:37 +0000 (16:29 +0200)]
Travis: Update gcc to 4.7
Fixes Travis builds currently failing with:
cc1plus: error: unrecognized command line option ‘-std=c++11’
Daniel Albers [Mon, 28 Jul 2014 13:55:44 +0000 (15:55 +0200)]
Merge pull request #77 from seezer/bug821
Don't scroll to bottom unexpectedly.
Fixes #821
Sebastian Goth [Wed, 25 Jun 2014 17:16:23 +0000 (19:16 +0200)]
Initialize ChatView scrolled to bottom
Now that we don't scroll on every resize, ensure that we start scrolled
to bottom.
Sebastian Goth [Tue, 17 Jun 2014 17:26:50 +0000 (19:26 +0200)]
Keep scrolled down views through resizeEvents
There's a reason for everything...
ChatView is rendered into the viewport from top to bottom.
If we don't scroll to bottom, the topmost line doesn't change and therefore
the scrollbar will move upwards on shrinking events.
So if the scrollbar was at bottom before the resize, make it stay there.
Sebastian Goth [Tue, 17 Jun 2014 16:07:11 +0000 (18:07 +0200)]
Don't scroll to bottom on resize
We're brave today..
Resizing the chatview (even unintentionally by hovering the topicline)
resulted in scrolling down the current view.
There might have been reasons.. i can't see them..
Fixes #821
Manuel Nickschas [Sat, 26 Jul 2014 12:38:11 +0000 (14:38 +0200)]
Enable C++11
As announced a while ago, we're now allowing C++11 in the codebase.
We now expect one of the following compilers:
* gcc >= 4.7
* Clang >= 3.1
* MSVC >= 18 (Visual Studio Nov 2013 CTP)
Older or different compilers may still work, but break at any time
in the future.
Bas Pape [Mon, 26 May 2014 17:24:01 +0000 (19:24 +0200)]
Do not update activity for hidden message types
Manuel Nickschas [Thu, 24 Jul 2014 21:20:41 +0000 (23:20 +0200)]
Fix support for CMake < 3
Meh. Apparently I can't just set behavior for future policies, so we
need to ensure that we're actually running CMake 3.
Manuel Nickschas [Thu, 24 Jul 2014 21:02:45 +0000 (23:02 +0200)]
Silence deprecation warning from CMake 3.0+
Setting COMPILE_DEFINITIONS_<Config> has been deprecated since CMake 3.0.
However, the "proper" way of doing things now (generator expressions) have
been support only since CMake 2.8.10, and we require 2.8.9 currently.
Thus, we explicitly enable OLD behavior for CMP0043.
Manuel Nickschas [Thu, 24 Jul 2014 20:51:16 +0000 (22:51 +0200)]
Adapt CMake logic for finding lconvert in newer Qt5
Seems like Qt 5.3 has added the long-missing lconvert target in its CMake
support scripts, however it didn't also add the Qt5_LCONVERT_EXECUTABLE
variable to go along with that. Adding the target disabled our workaround,
breaking translation support in Quassel.
We now also cover this case (and made things future-proof in case Qt5 ever
adds the variable upstream).
J-P Nurmi [Thu, 10 Jul 2014 22:11:27 +0000 (00:11 +0200)]
LegacyPeer: kill the dependency to Quassel::buildInfo()
This makes LegacyPeer usable outside Quassel (IRC for Sailfish)
Mattia Basaglia [Thu, 26 Jun 2014 13:23:13 +0000 (15:23 +0200)]
Use correct nick on 001 RPL_WELCOME
Fixes #1020
When the user connects to a bouncer, the bouncer will send back the
welcome message received when it first connected to the network.
This caused Quassel to display a wrong nick if it had changed
since the first connection.
Manuel Nickschas [Sat, 5 Jul 2014 20:14:00 +0000 (22:14 +0200)]
Merge pull request #81 from otih/master
Fix Bug #1293: --syslog option writes to emergency level
otih [Thu, 3 Jul 2014 21:34:44 +0000 (23:34 +0200)]
Fix Bug #1293: --syslog option writes to emergency level
otih [Thu, 3 Jul 2014 20:58:35 +0000 (22:58 +0200)]
Fix Bug #1293: --syslog option writes to emergency level
Daniel Albers [Wed, 2 Jul 2014 15:34:21 +0000 (17:34 +0200)]
Don't treat pure format strings as translatable.
Manuel Nickschas [Wed, 2 Jul 2014 22:43:01 +0000 (00:43 +0200)]
Merge pull request #78 from seezer/qt5contextmenu
Fix contextmenu actions for Qt5
Manuel Nickschas [Wed, 2 Jul 2014 22:24:47 +0000 (00:24 +0200)]
Merge pull request #73 from jpnurmi/signalproxy
signalproxy.h: add missing forward declaration of QIODevice
Manuel Nickschas [Wed, 2 Jul 2014 22:22:57 +0000 (00:22 +0200)]
Merge pull request #72 from jpnurmi/datastream
DataStreamPeer: add missing include <QDataStream>
Sebastian Goth [Tue, 17 Jun 2014 19:33:21 +0000 (21:33 +0200)]
Fix contextmenu actions for Qt5
As documented for QObject::connectNotify
Sebastian Meyer [Fri, 13 Jun 2014 12:52:55 +0000 (14:52 +0200)]
Checking for single char line endings
This fixes problems with IRC servers sending only single char line
endings. Before the last char on each line for thos servers would be
chopped, now Quassel checks how the line ends and chops accordingly.
Fixes #1267
Manuel Nickschas [Fri, 9 May 2014 19:38:37 +0000 (21:38 +0200)]
Fix build on KDE
Oops. This wouldn't have happened with 'auto' :P
Manuel Nickschas [Mon, 5 May 2014 21:38:27 +0000 (23:38 +0200)]
Fix tab completion on exotic platforms
Turns out that using the tab key as an action shortcut in an input widget
does not work reliably on all platforms, at least on OS/2 - sometimes,
the event filter would be called instead of triggering the action.
To make this more robust, we disable the action (just using it as a container
for the custom shortcut) and always catch and handle the key in the event
filter which we need anyway.
Thanks to TeLLie for bisecting this issue and finding the culprit on OS/2.
Manuel Nickschas [Mon, 5 May 2014 18:47:40 +0000 (20:47 +0200)]
Update ChangeLog
Manuel Nickschas [Mon, 5 May 2014 18:37:36 +0000 (20:37 +0200)]
Fix crash with KDE network detection
If the network connection status changed during authentication phase,
Quassel would crash. This is caused by some old bits of code that still
assume that there's always a Peer around, which is no longer the case
after splitting off the AuthHandler.
This commit fixes that issue. Thanks to shadeslayer for the report and
really useful backtrace.
J-P Nurmi [Mon, 28 Apr 2014 21:06:58 +0000 (23:06 +0200)]
signalproxy.h: add missing forward declaration of QIODevice
J-P Nurmi [Mon, 28 Apr 2014 21:05:00 +0000 (23:05 +0200)]
DataStreamPeer: add missing include <QDataStream>
Marcus Eggenberger [Wed, 9 Apr 2014 12:58:49 +0000 (14:58 +0200)]
Build mono client packages on Mac OS X
When deploying on Mac OS X, also build a disk image (dmg) for the mono
client. Also updated cmake files for OS X 10.8.
Manuel Nickschas [Mon, 31 Mar 2014 22:07:45 +0000 (00:07 +0200)]
Make sure our Qt5 is new enough
find_package will happily ignore the requested version if not used
with REQUIRED, plus early Qt5 versions handled versions differently;
thus, let's explicitly check for a supported version and bail out
otherwise.
Manuel Nickschas [Mon, 31 Mar 2014 21:14:55 +0000 (23:14 +0200)]
We don't actually require XmlPatterns for WebKit
No clue why we forced linking against that one, but it doesn't seem
to be required for both Qt4 and Qt5, so let's just remove the dependency.
Manuel Nickschas [Mon, 31 Mar 2014 20:12:28 +0000 (22:12 +0200)]
Make --syslog option work again
The -DHAVE_SYSLOG was missing for compiling main() after the recent
build system changes, so the option wasn't enabled (but caused a
warning).
Manuel Nickschas [Mon, 31 Mar 2014 20:03:54 +0000 (22:03 +0200)]
Bring back COMMON_DEPS
This is actually being used to link the app icon on Windows, so bring
the old variables back for now. Probably want to clean this up later
though, as the naming is confusing.
Manuel Nickschas [Mon, 31 Mar 2014 19:58:27 +0000 (21:58 +0200)]
Fix Qt resource handling
Turns out we have to explicitly initialize Qt resources that are linked
statically, so our last build system change broke things because the
files where nowhere to be found.
This now initializes the resources in main(); additionally, we use absolute
paths for the qrc files now, simply for making things a bit clearer.
Manuel Nickschas [Sun, 30 Mar 2014 20:45:19 +0000 (22:45 +0200)]
Merge pull request #69 from Nevcairiel/qt5msvcbuild
Qt5 MSVC build fixes
Hendrik Leppkes [Sat, 29 Mar 2014 16:18:22 +0000 (17:18 +0100)]
Fix linking against Qt5 on MSVC
QT_QTMAIN_LIBRARY is no longer automatically set on Qt5, as it is part of
a deprecated way to link against Qt. Since Quassel needs to support both
Qt4 and Qt5, setting the variable to the appropriate value again works
around the issue until such a time that the CMake requirement is increased
to make use of policy CMP0020.
Note that this is the same workaround cmake itself uses when building
cmake-gui for Windows.
Hendrik Leppkes [Sat, 29 Mar 2014 16:16:13 +0000 (17:16 +0100)]
Fix building with Qt5 on MSVC by undefining min/max macros
The Windows stdlib.h header defines min/max macros, which interfer with
the Qt5 header files. Luckily, it also provides the define NOMINMAX which
suppresses these macros, and fix building.
Manuel Nickschas [Fri, 28 Mar 2014 23:57:22 +0000 (00:57 +0100)]
Avoid race conditions in the build system
Adding resource files in multiple targets is a bad idea, because CMake
will happily generate them multiple times too, and unfortunately also
in the same location. With Qt5, this actually lead to a misgenerated
oxygen_kde.qrc some times.
Now we add the resources in the module they belong to, which not only
avoids these race conditions, but also saves some time during compile,
because every resource file is only processed once now.
Changing this triggered yet another race condition, where not all translations
would be generated by the time the corresponding .qrc was needed. Adding
a fugly add_dependencies(mod_common po) fixes this; however we may want to
find a cleaner way to properly express deps between the language files and the
qrc.
Manuel Nickschas [Fri, 28 Mar 2014 22:26:53 +0000 (23:26 +0100)]
Properly handle RPC calls that return void
Qt5 loudly warns when trying to construct a QVariant from QMetaType::Void,
thus we should not do that. Instead, check for this case and use an invalid
QVariant instead. Works fine on both Qt4 and Qt5 that way.
Manuel Nickschas [Thu, 27 Mar 2014 22:20:24 +0000 (23:20 +0100)]
Fix rendering of the sender column
Turns out the compositing needed to be tweaked to actually behave
as desired. Senders are now properly rendered including the fade-out
effect on both Qt4 and Qt5.
Manuel Nickschas [Thu, 27 Mar 2014 20:45:13 +0000 (21:45 +0100)]
Merge pull request #68 from jpnurmi/build-date
Protocol::RegisterClient: add buildDate field
Manuel Nickschas [Thu, 27 Mar 2014 20:44:54 +0000 (21:44 +0100)]
Merge pull request #67 from jpnurmi/preset
Move preset networks handling out of Network
Manuel Nickschas [Thu, 27 Mar 2014 20:42:39 +0000 (21:42 +0100)]
Merge pull request #66 from TheOneRing/snore
Fixed detection of snore for Qt4/Qt5
J-P Nurmi [Thu, 27 Mar 2014 19:15:22 +0000 (20:15 +0100)]
Protocol::RegisterClient: add buildDate field
This kills the dependency from DataStreamPeer to Quassel::buildInfo().
J-P Nurmi [Wed, 26 Mar 2014 23:17:41 +0000 (00:17 +0100)]
Move preset networks handling out of Network
This moves highly application specific functionality out of Network
and makes it possible to re-use Quassel's code base for implementing
the Quassel protocol in external applications.
Manuel Nickschas [Wed, 26 Mar 2014 21:03:21 +0000 (22:03 +0100)]
Move the PeerPtr declaration out of types.h
This belongs in peer.h instead, since types.h is included in lots
of places that should not drag the peer deps in. Also, this caused
problems (circular deps?) when building the OSX notification backend.
Patrick von Reth [Wed, 26 Mar 2014 20:16:01 +0000 (21:16 +0100)]
Fixed detection of snore for Qt4/Qt5
Manuel Nickschas [Wed, 26 Mar 2014 19:42:29 +0000 (20:42 +0100)]
Merge pull request #64 from TheOneRing/warnings
silence some warnings
Daniel Albers [Tue, 25 Mar 2014 21:56:13 +0000 (22:56 +0100)]
Update quassel.pot
Daniel Albers [Wed, 19 Feb 2014 04:00:37 +0000 (05:00 +0100)]
Update translations from Transifex
25507 translated messages
Many thanks to:
- fi: Larso <larso@gmx.com>
- tr: volkangezer <volkangezer@gmail.com>
- uk: Yuri Chornoivan <yurchor@ukr.net>, Daniel Albers <daniel@lbe.rs>
- pt_BR: André Marcelo Alvarenga <alvarenga@kde.org>
Manuel Nickschas [Tue, 25 Mar 2014 22:13:21 +0000 (23:13 +0100)]
Use the raster rendering engine by default on OSX
Seems like using the raster engine fixes performance issues on OSX,
and there seem to be no noticeable side-effects compared to native
rendering. So let's use this by default.
If you notice issues, please let us know. Also, you can always enforce
native rendering via command line:
open Quassel\ Client.app -W --args -graphicssystem native
Thanks to rikai for figuring this out!
K. Ernest (iFire) Lee [Mon, 10 Feb 2014 02:29:43 +0000 (18:29 -0800)]
Unnecessary forward declaration of CoreSession.
Manuel Nickschas [Tue, 25 Mar 2014 21:15:13 +0000 (22:15 +0100)]
Make the web preview build on Qt4 again
Prefixing the includes with the Qt module broke things for Qt4. It's
not needed anyway, so just remove the prefix.
K. Ernest (iFire) Lee [Sun, 9 Feb 2014 21:57:30 +0000 (13:57 -0800)]
Webkit changed location and scale() is gone.
Webkit's include location changed and scale() is replaced by transform().
Manuel Nickschas [Tue, 25 Mar 2014 20:51:41 +0000 (21:51 +0100)]
Properly find WebKit for Qt5
It's WebKit, not Webkit.
Manuel Nickschas [Mon, 24 Mar 2014 23:15:47 +0000 (00:15 +0100)]
QTreeView::dataChanged() signature changed in Qt5
Manuel Nickschas [Mon, 24 Mar 2014 23:14:57 +0000 (00:14 +0100)]
Add missing include
Manuel Nickschas [Mon, 24 Mar 2014 23:14:34 +0000 (00:14 +0100)]
QDesktopServices -> QStandardPaths
Manuel Nickschas [Mon, 24 Mar 2014 23:13:42 +0000 (00:13 +0100)]
QSslCertificate::subjectInfo() returns a QStringList in Qt5
Manuel Nickschas [Mon, 24 Mar 2014 23:11:39 +0000 (00:11 +0100)]
Use QSslCertificate::isNull() instead of isValid() in IdentityEditWidget
The intention is to check if the certificate could be loaded, and isNull()
is actually the right thing to do. Saves us from dealing with the no-longer
existing isValid() in Qt5, too.
Manuel Nickschas [Mon, 24 Mar 2014 21:59:37 +0000 (22:59 +0100)]
QSslCertificate::{subject,issuer}Info() returns a QStringList in Qt5
This requires changes in SslInfoDlg. To simplify handling the API differences,
we use small wrappers that join the string lists for Qt5.
Manuel Nickschas [Mon, 24 Mar 2014 21:28:16 +0000 (22:28 +0100)]
QSslCertificate::isValid() no longer exists in Qt5
Qt4's isValid() checked for the date range and if the cert is blacklisted.
In Qt5, these two checks need to be done explicitly.
Bas Pape [Sun, 21 Apr 2013 11:19:44 +0000 (13:19 +0200)]
qVariantCanConvert is deprecated
Bas Pape [Sat, 20 Apr 2013 22:26:21 +0000 (00:26 +0200)]
QHeaderView::setResizeMode is deprecated in Qt 5
Bas Pape [Sat, 20 Apr 2013 22:04:41 +0000 (00:04 +0200)]
Logging changed in Qt 5; use QtMessageHandler
Bas Pape [Sat, 20 Apr 2013 21:45:10 +0000 (23:45 +0200)]
qVariantValue is deprecated in Qt 5.0
Bas Pape [Sat, 20 Apr 2013 21:39:14 +0000 (23:39 +0200)]
setAcceptsHoverEvents is deprecated in Qt 5.
setAcceptsHoverEvents has been around since 4.4
Bas Pape [Sat, 20 Apr 2013 21:33:06 +0000 (23:33 +0200)]
setAlphaChannel is deprecated in Qt 5.
With the default CompositionMode this should work too.
Bas Pape [Sat, 20 Apr 2013 12:48:08 +0000 (14:48 +0200)]
QKeySequence's int operator is deprecated
operator int returns the first key in the sequence, so just use that
one directly.
Bas Pape [Sat, 20 Apr 2013 12:21:06 +0000 (14:21 +0200)]
No more posF in Qt 5; use the new localPos
Bas Pape [Sat, 20 Apr 2013 11:46:29 +0000 (13:46 +0200)]
reset is deprecated since Qt 5.0
Bas Pape [Sat, 20 Apr 2013 11:37:23 +0000 (13:37 +0200)]
Prevent unused-but-set warnings ifndef QT_NO_DEBUG
Bas Pape [Sat, 20 Apr 2013 11:24:34 +0000 (13:24 +0200)]
Qt::escape moved to QString::toHtmlEscaped in Qt 5
Manuel Nickschas [Mon, 24 Mar 2014 20:43:45 +0000 (21:43 +0100)]
Change Q_WS_* to Q_OS_*
The Q_WS_{WIN,MAC} macros have been deprecated since Qt 4.3 or so,
and replaces by the corresponding Q_OS_* macros. Since Qt5 does not
support them anymore at all, let's globally change those to the "new"
version.
Also change Q_OS_WIN32 to just Q_OS_WIN; while the former is supposed to
be defined on all supported Windows platforms including Win64, it's still
a bit confusing.
Manuel Nickschas [Mon, 24 Mar 2014 20:05:54 +0000 (21:05 +0100)]
Signature of QTcpServer::incomingConnection() changed in Qt5
Manuel Nickschas [Mon, 24 Mar 2014 20:00:44 +0000 (21:00 +0100)]
Can't forward-declare meta types in Qt5
K. Ernest (iFire) Lee [Sun, 9 Feb 2014 19:43:56 +0000 (11:43 -0800)]
Replace {from,to}Ascii with {from,to}Latin1
These have been deprecated since Qt 5.0, and rather than use
{from,to}Local8Bit I've decided to match the original functionality.
Manuel Nickschas [Mon, 24 Mar 2014 19:52:12 +0000 (20:52 +0100)]
Remove the useless coreeventmanager.cpp again
Was pulled in by applying harryF's patch, but removed from mainline
earlier.
Harald Fernengel [Thu, 16 Aug 2012 08:07:31 +0000 (10:07 +0200)]
Qt5 fixes
Mostly due to the metatobject changes. Also, for some reason, my
cmake on the mac refused to create a moc file for coreeventmanager
unless there's a *.cpp file
Manuel Nickschas [Mon, 24 Mar 2014 23:18:45 +0000 (00:18 +0100)]
Properly handle DBusMenuQt in the build system
This library ships a CMake config file these days, so we can remove
our own find script (which was no longer used anyway). Also, we need to
treat it as an imported target, there's no dbusmenu-qt_LIBRARIES...
Manuel Nickschas [Mon, 24 Mar 2014 18:24:57 +0000 (19:24 +0100)]
Clean up and fix data/CMakeLists.txt
This was forgotten during the recent changes and still referred to
HAVE_KDE, thus failing now. Taking the opportunity to clean up this
ancient cruft as well (those conditionals...)
Manuel Nickschas [Sun, 23 Mar 2014 22:33:00 +0000 (23:33 +0100)]
Fix handling the QtMain module on Windows
Turns out that the QtMain module can't be used in qt4_use_modules,
and we should just add the library to our targets.
Manuel Nickschas [Sun, 23 Mar 2014 22:02:03 +0000 (23:02 +0100)]
Improve the handling of Qt5 modules in the build system
If a required Qt5 module is not available by the time it is needed,
CMake throws a rather fugly backtrace with a confusing error message.
To avoid this, we now check for all required Qt5 modules by calling
find_package without REQUIRED, and setting the REQUIRED pacakge property.
That way, there will be no errors during configuration, the feature
summary correctly shows the packages as missing, and then bails out
at this point, which results in a much nicer user experience.
For this to work, we also need to show the feature summary before adding
the src/ subdirectory, as this will forcefully require the modules it needs.
Additionally, this commit fixes finding lconvert (and thus translation
support) for Qt5.
Manuel Nickschas [Sun, 23 Mar 2014 20:28:28 +0000 (21:28 +0100)]
Fix SSL detection
As there is no easy way to check for QT_CONFIG with Qt5 and CMake,
and the way we did it for Qt4 is hackish at best, let's do it the
proper way and compile a short code snippet that checks for the
appropriate defines (QT_NO_OPENSSL for Qt4, and QT_NO_SSL for Qt5).
This should be the most robust way of figuring this out, and is rather
easy with CMake.
Manuel Nickschas [Sun, 23 Mar 2014 19:42:55 +0000 (20:42 +0100)]
Update INSTALL file
This hadn't been touched since 2007, and with all the recent build
system changes, I guess it made sense to adapt it to reality.
Manuel Nickschas [Sun, 23 Mar 2014 18:46:58 +0000 (19:46 +0100)]
Remove obsolete CMake options
Now that the new dependency handling is in place, we can remove several
of the CMake options we used to have. If you don't want automagic
dependencies, and want to disable one even though it is installed, you
can use -DCMAKE_DISABLE_FIND_PACKAGE_Foo=TRUE for ignoring the Foo package.
Manuel Nickschas [Sun, 23 Mar 2014 18:44:27 +0000 (19:44 +0100)]
Make SSL and syslog support non-configurable
Both features have no external dependencies; SSL support requires Qt
to be built with SSL support, and syslog support requires the existence
of the syslog.h header. It makes no sense to make this configurable,
as no additional libraries or other dependencies are pulled in.
Instead, just add them to the feature_summary depending on the prerequisites
being in place.
Manuel Nickschas [Sun, 23 Mar 2014 01:39:38 +0000 (02:39 +0100)]
Completely rework the dependency handling in the build system
Traditionally, optional build dependencies are handled in CMake via
offering options and conditionally searching for the needed packages.
This leads to having lots of options, boilerplate for telling the user
which features are enabled or not, unclear handling of automagic
dependencies, and so on.
These days, CMake offers a much nicer way: feature_summary() and friends.
To make use of this, one calls find_package unconditionally and then sets
properties on the package that tell CMake if the dependency is optional
or required, its purpose, a description of the package etc. At the end
of configuration, CMake then displays a nice summary of all the packages
searched for and (not) found, as well as other features.
Quassel now makes use of this modern way of specifying dependencies, replacing
most of the former options (they're still there, but unused now, and will
be removed soon). Note that one can still disable an optional dependency Foo by
passing -DCMAKE_DISABLE_FIND_PACKAGE_Foo=TRUE to CMake.
I also took the opportunity to radically clean up and fix how we handle
dependencies throughout the build system:
* We no longer set an extra CMake variable to indicate that a package was
found, but use the Foo_FOUND variable directly
* Compile definitions and include directories are no longer set globally
in the root CMakeLists.txt, but in the Quassel modules where they're
actually needed; this should speed up compiling ever so slightly and
generally make things much cleaner
* Libraries are now linked directly to the Quassel modules that use them,
rather than to the executable. This prepares us for making the modules
shared libraries at some point, and is anyway The Right Thing™ to do
* Fully support Qt5 in the build system, including finding the proper version
of dependencies where appropriate and available
Note that the changes to the build system are not complete yet, and some things
may be broken now. Note also that while Qt5 is now supported in the build system,
the code itself still requires many changes to actually compile with Qt5,
so don't even try.