Adam [Tue, 15 Nov 2016 16:48:58 +0000 (11:48 -0500)]
Fix sasl authentication to fail on servers which don't support sasl
Resolves GH-262.
romibi [Sun, 14 Feb 2016 14:06:50 +0000 (15:06 +0100)]
Add some more needed DLL's for Link-Preview support
Manuel Nickschas [Sun, 24 Apr 2016 20:23:44 +0000 (22:23 +0200)]
Bump version for release
Manuel Nickschas [Sun, 24 Apr 2016 20:22:36 +0000 (22:22 +0200)]
Update ChangeLog
Shane Synan [Wed, 30 Mar 2016 23:50:46 +0000 (18:50 -0500)]
Add documentation to isStatusMsg
Michael Marley [Wed, 23 Mar 2016 20:47:02 +0000 (16:47 -0400)]
Handle STATUSMSG messages
Previously PRIVMSGs to channels prefixed with certain characters
(usually "+" or "@") (example "@#quassel") would appear in new
query buffers instead of in the channel buffer. People were using
that problem to broadcast query SPAM in active channels such as
This patch fixes the issue by detecting the condition, first in
ircparser.cpp where it will use the channel name instead of the
sender's nick as the target. ctcpparser.cpp then sees the prefixed
channel name and reacts by truncating the prefix(es) and setting a
new flag so that the client can tell the message was a STATUSMSG.
A look-ahead system is used to ensure that the channel prefix
character is not stripped even when the channel prefix characters
and STATUSMSG prefix characters overlap.
The server-provided STATUSMSG prefixes are used if available,
otherwise "@" and "+" are used.
This patch doesn't make the client display the STATUSMSGs any
differently than regular messages. I omitted this because there
was some discussion on the channel about the best way to display
these messages, but no conclusion was ever reached. This patch
will at least stop the query SPAM and also provides the
infrastructure necessary to change the display format later.
romibi [Sun, 14 Feb 2016 13:12:53 +0000 (14:12 +0100)]
Enable Link-Preview on appveyor builds
Manuel Nickschas [Sun, 24 Apr 2016 19:59:15 +0000 (21:59 +0200)]
Handle invalid handshake data properly in the core
Clients sending invalid handshake data could make the core crash
due to an unchecked pointer. This commit fixes this issue by having
the core close the socket if a peer could not be created.
Thanks to Bas Pape (Tucos) for finding this one!
Sebastian Goth [Sat, 23 Apr 2016 11:39:59 +0000 (13:39 +0200)]
Support session management with Qt5
Qt5 moved away from virtual methods towards signals for commitData() and saveState().
These were simply not called anymore in our code.
In Qt 5.6 a bug was fixed that broke session management at least with KF5.
As we do handle the session management, we can safely disable the fallback.
See QTBUG-49667 and https://codereview.qt-project.org/#/c/148274/ for details.
Daniel Albers [Wed, 25 Mar 2015 04:00:24 +0000 (05:00 +0100)]
Update translations from Transifex
26449 translated messages
Many thanks to:
- cs: Jaroslav Lichtblau <dragonlord@seznam.cz>
- de: Sebastian Meyer <transifex@netzvieh.de>
- es: Javier Llorente <javier@opensuse.org>
- fi: Lasse Liehu <larso@gmx.com>
- fr: Vlavv <vlavv2@gmail.com>
- hi: Raju Devidas Vindane <rajuvindane@gmail.com>
- pa: A S Alam <apreet.alam@gmail.com>
- pt_BR: André Marcelo Alvarenga <alvarenga@kde.org>
- sl: Andrej Mernik <andrejm@ubuntu.si>
- tr: Demiray Muhterem <mdemiray@msn.com>, Volkan Gezer <volkangezer@gmail.com>
- uk: Yuri Chornoivan <yurchor@ukr.net>
Manuel Nickschas [Wed, 10 Feb 2016 18:58:52 +0000 (19:58 +0100)]
Don't prefix the nick with a colon while connecting to a server
For the NICK command during the initial connection handshake, we
always prefixed the parameter with a colon. While this is not an RFC
violation, it is unnecessary as nicks can't contain spaces. Worse, it
confuses some obscure ircds...
Manuel Nickschas [Tue, 9 Feb 2016 21:02:34 +0000 (22:02 +0100)]
Bump version for release
Manuel Nickschas [Tue, 9 Feb 2016 20:59:55 +0000 (21:59 +0100)]
Update ChangeLog
Manuel Nickschas [Mon, 8 Feb 2016 21:37:57 +0000 (22:37 +0100)]
Support new QtInfoMsg message type
Qt 5.5 introduced QtInfoMsg. This commit adds support for this
in our logger, thus removing a warning.
Manuel Nickschas [Mon, 8 Feb 2016 21:21:43 +0000 (22:21 +0100)]
Update AboutData
Manuel Nickschas [Mon, 8 Feb 2016 20:29:14 +0000 (21:29 +0100)]
Don't allow newlines in buffer names
Since it is possible to enter newlines when renaming a buffer via
the GUI, we need to have a sanity check that cuts off the additional
lines.
Fixes #1389.
Manuel Nickschas [Mon, 8 Feb 2016 20:04:03 +0000 (21:04 +0100)]
Fix issues with buffer selection and filtered views
Whenever the current buffer changed, filtered BufferViews need to
check if the previously selected item should be hidden (e.g. due to
activity settings). The existing code for this never really worked
due to the currentChanged() signal emission being wonky (it would
sometimes be eaten for some reason), and fully stopped working in
Qt5 since it connected to a private, internal slot of QSFPM.
This commit removes the old code and replaces it by faking the
emission of a dataChanged() signal from the base model whenever
the current buffer changes.
Manuel Nickschas [Thu, 4 Feb 2016 21:40:49 +0000 (22:40 +0100)]
Fix build with Qt4
Qt4 can't serialize lists of things out of the box, in this case
lists of shortcuts. Rather than declaring a new metatype, we just
go back to storing a single shortcut for Qt4 - it's all we internally
support anyway.
Manuel Nickschas [Wed, 3 Feb 2016 23:10:08 +0000 (00:10 +0100)]
Don't serialize PeerPtr value in RPC connections
PeerPtr is used in RPC signatures for enabling receivers to send replies
to a particular peer rather than broadcast to all connected ones.
To enable this, the SignalProxy transparently replaces the bogus value
received over the network with the actual address of the local Peer
instance.
Because the actual value isn't needed on the wire, it should be
serialized as null. This also prevents the accidental use of a bogus
remote pointer address.
Manuel Nickschas [Wed, 3 Feb 2016 21:50:02 +0000 (22:50 +0100)]
Fix shortcut handling with KDE Frameworks
KDE Frameworks slightly altered the way it handles default shortcuts
for actions; the property name was changed to support lists of
shortcuts. This issue lead to default shortcuts not being active for
Quassel built against KDE Frameworks.
Hannah von Reth [Wed, 3 Feb 2016 09:37:01 +0000 (10:37 +0100)]
Fix possible crash, mostly occurring on notification flood.
Janne Koschinski [Tue, 2 Feb 2016 16:50:32 +0000 (17:50 +0100)]
Fixed a tiny bug in aliasmanager
The `/wait` check is to test if the first command is a `/wait`, which would operate synchronously.
But `/wait` matches also on other commands – so, if, for example, we’d have a command like `/waiting`, it would be matched.
Instead we actually want to match on `/wait ` (with space at the end).
Ayonix [Mon, 18 Jan 2016 21:57:35 +0000 (22:57 +0100)]
Fix the SSL check on systems that don't use OpenSSL
Hannah von Reth [Thu, 3 Dec 2015 14:49:23 +0000 (15:49 +0100)]
Fix build with snorenotify 0.7
Hannah von Reth [Tue, 12 Jan 2016 14:53:57 +0000 (15:53 +0100)]
Fix msvc x64 build and openssl for msc.
Parts of the changes are in the external script.
Patrick von Reth [Tue, 22 Sep 2015 13:31:28 +0000 (15:31 +0200)]
Apply small snore api changed.
Patrick von Reth [Tue, 18 Aug 2015 07:12:46 +0000 (09:12 +0200)]
Update Snorenotify project location.
Patrick von Reth [Thu, 16 Jul 2015 14:15:54 +0000 (16:15 +0200)]
Updated the Snorenotification backend to the new api of Snorenotify.
Use Snorenotify also if Quassel is build with KDE support but only load
secondary backends which allow redirection to a phone etc. are loaded.
esainane [Sat, 12 Dec 2015 05:56:17 +0000 (18:56 +1300)]
signalproxy: Correct signal method name warnings
_id was used to index SignalProxy::SignalRelay::_slots.
signal.signalId is used to index caller->metaObject()->method.
This can be verified by changing eg. displayMsg to use a type that is
compatible with Message but won't work with QMetaType::type, like Message&
Hannah von Reth [Thu, 10 Dec 2015 15:54:26 +0000 (16:54 +0100)]
Only link to ZLIB if it was found.
Patrick von Reth [Tue, 4 Aug 2015 12:12:13 +0000 (14:12 +0200)]
Add support for Appvayor Windows ci.
Add build scripts for appveyor.com ci for mingw and msvc 2013.
The ci automatically creates a Nsis installer and a 7zip package.
Michael Marley [Tue, 1 Dec 2015 00:33:37 +0000 (19:33 -0500)]
Fix migration of hashversion
Previously, hashversion was not migrated. This would cause the
value to default to 0, causing login failures for existing users.
Michael Marley [Sat, 23 May 2015 22:17:04 +0000 (18:17 -0400)]
Use SHA2-512 to store the core certificate digest in the client
Previously, MD5 was used. Like the core password hashing update,
this implements a versioning system so that the digest will be
automagically upgraded without bugging the user. If for some
reason the user downgrades again, the client will still work but
the user will be asked to validate the certificate again.
SHA2-256 is more common for storing certificate digests, but in
this case, SHA2-512 was used. This is because Qt4 does not
implement any hash higher than SHA1 and a SHA2-512 implementation
is already bundled for use with the core password hashing.
Sai Nane [Sat, 28 Nov 2015 23:43:40 +0000 (12:43 +1300)]
Fix ident race condition
This creates an entry for the port as soon as the socket is opened,
without duplicates. This allows an identd service to respond correctly
if a request is received before the SSL handshake has finished and been
processed on our side.
Since socketOpen is not used in the project, it is removed here.
Since the only usage of socketInitialized(...) is with oidentd, its
semantic meaning is slightly changed to occur earlier, like socketOpen,
but only once, like the former socketInitialized(...).
esainane [Sat, 21 Nov 2015 08:38:11 +0000 (21:38 +1300)]
Fix OIdentdConfigGenerator double-printing on ssl
connected() is always connected to socketInitialized().
encrypted() is also connected to socketInitialized() when ssl is enabled.
socketInitialized() fires socketOpen(...) unconditionally, but does not fire socketInitialized(...) if we're using SSL without encryption yet - see line 464.
With socketOpen(...) connected to OIdentdConfigurationGenerator::addSocket, OIdentdConfigurationGenerator writes to .oidentd.conf twice for every connection.
Changing this to have socketInitialized(...) connected instead prevents this duplication.
Weng Xuetian [Thu, 12 Nov 2015 00:13:33 +0000 (16:13 -0800)]
Fix SSL detection
CMAKE_POSITION_INDEPENDENT_CODE will always append fPIE flag as last compiler flag when calling try_compile which makes check_cxx_source_compiles fails. Move the SSL check before set CMAKE_POSITION_INDEPENDENT_CODE and append Qt5Core_EXECUTABLE_COMPILE_FLAGS to CMAKE_CXX_FLAGS solve the problem.
Raul Salinas-Monteagudo [Sat, 31 Oct 2015 10:53:02 +0000 (11:53 +0100)]
quasselcore: --add-user and --change-userpass now exit with non-null in case
of errors.
Konstantin Bläsi [Wed, 21 Oct 2015 01:26:02 +0000 (03:26 +0200)]
Fixes a crash of the core when executing "/op *" in a query.
Jan Alexander Steffens (heftig) [Mon, 19 Oct 2015 19:30:28 +0000 (21:30 +0200)]
Correct capitalization of networkid column name
Fixes http://bugs.quassel-irc.org/issues/1369
Petr Bena [Wed, 23 Sep 2015 14:07:16 +0000 (16:07 +0200)]
Performance tweak: rem. pointless recursive calls
For some reason there was an extra call to IrcChannel::joinIrcUser
from IrcUser::joinIrcChannel. That means that in past we did:
* Call to IrcChannel::joinIrcUsers(...)
* From there we call IrcUser::joinIrcChannel(this)
* And from there we called IrcChannel::joinIrcUser(this)
* Which again called IrcChannel::joinIrcUsers(list that contained only this 1 user)
* Which figured out this user is already in this channel and exits
That is a pointless overhead, so now we call the second function with
extra parameter that tells it not to call joinIrcUser because we know
this user doesn't need to be added there, thus saves few thousands of
pointless function calls and user list contructions, especially while
joining large channels.
Pierre Schweitzer [Thu, 2 Jul 2015 12:31:29 +0000 (14:31 +0200)]
Don't assign a default value to spacePos
Will be ignored later on in CtcpParser::parseSimple.
Pierre Schweitzer [Thu, 2 Jul 2015 12:28:05 +0000 (14:28 +0200)]
Pass argument by reference in CliParserArg constructor
Pierre Schweitzer [Thu, 2 Jul 2015 12:21:55 +0000 (14:21 +0200)]
Use prefix operators when possible for non-primitive types
This affects mostly iterators.
Andrew Brown [Thu, 14 May 2015 08:12:26 +0000 (09:12 +0100)]
Write out warning when a user login fails
Fixes #1356.
Bas Pape [Sat, 28 Dec 2013 12:07:30 +0000 (13:07 +0100)]
Preserve white-space in the input history.
By default QTextHtmlImporter::appendNodeText replaces white-space
characters with a single space and leading white-space is removed. When
white-space is set to pre-wrap in the style of the node, it is preserved
instead.
Given that contents are in a span element, setting white-space to
pre-wrap on span makes the setHtml call do the expected thing.
Bas Pape [Thu, 24 Sep 2015 19:15:49 +0000 (21:15 +0200)]
Store the channel keys in CoreNetwork again
The refactoring from
3146ad0 (and
e0781ee) was not a true refactoring as
it changed behaviour: CoreIrcChannels (as opposed to CoreNetworks) are
ephemeral, which meant disconnecting or parting lost the channel key.
Keys are still kept by the channel's cipher, but synced from/to its network when
the constructor/destructor is called.
Fixes #1248
Bas Pape [Thu, 24 Sep 2015 18:32:13 +0000 (20:32 +0200)]
Disable checkbox if no DockManager is present
This gets rid of the stray debug message introduced by
0bf9227 and adds
some visual feedback on the settingspage instead.
Bas Pape [Thu, 24 Sep 2015 18:29:08 +0000 (20:29 +0200)]
Remove forward declarations in multilineedit.h
TabCompleter isn't used and QKeyEvent's header is included (and required
for inlined functions).
Bas Pape [Thu, 24 Sep 2015 18:11:48 +0000 (20:11 +0200)]
Remove the tabcompletion space at end of text
The "add space after nick when completing mid-sentence" option
introduced by
84381d6 left the space even if it turned out to be at
the end. Remove it again if the text hasn't changed when sending the
text.
Petr Bena [Tue, 22 Sep 2015 08:01:50 +0000 (10:01 +0200)]
Fixed english in ircchannel.cpp
Mostly only comments, it hurt my eyes while reading it :P
Petr Bena [Mon, 21 Sep 2015 12:19:33 +0000 (14:19 +0200)]
Clean up
Removed extra semicolon and resolved some compiler warnings
Armin K [Sat, 2 May 2015 21:04:53 +0000 (23:04 +0200)]
Fix build with Qt-5.5
http://code.qt.io/cgit/qt/qtbase.git/commit/?id=
ebef2ad1360c80ad62de5f4a1c4e7e4051725c1c
Michael Marley [Sat, 13 Jun 2015 21:38:29 +0000 (17:38 -0400)]
Keep _wantReconnect set to true by default instead of false
Previously _wantReconnect was set to false by default (unless the
client was actually disconnected). If the disconnect() signal was
fired without any other signals (socketError(), for example), this
would cause the client to not reconnect. This seems to be
happening rather often with Qt5 client builds because a vast
majority of the time when I restart my core (or the entire core
server), the Qt5 clients do not reconnect. This patch keeps
_wantReconnect set to true whenever the client is connected so that
problem will not occur anymore. Disconnecting manually will set
it back to false so that the client won't try to reconnect after
a manual disconnect.
Jason Joyce [Wed, 13 May 2015 01:53:18 +0000 (21:53 -0400)]
Adding tag file to the ignore list.
Deiz [Fri, 8 May 2015 19:37:52 +0000 (15:37 -0400)]
Respect WITH_WEBKIT when defining HAVE_WEBKIT
Jason Joyce [Fri, 8 May 2015 02:41:10 +0000 (22:41 -0400)]
Adding the build folder to the ignore file.
jjoyce [Thu, 7 May 2015 20:46:27 +0000 (16:46 -0400)]
Adding a gitignore file.
jjoyce [Thu, 7 May 2015 19:03:02 +0000 (15:03 -0400)]
Python PEP8 cleanup of macosx_makebundle.py
jjoyce [Thu, 7 May 2015 18:46:41 +0000 (14:46 -0400)]
Python PEP8 cleanup of macos_DeployApp.py
Michael Marley [Tue, 5 May 2015 01:27:14 +0000 (21:27 -0400)]
Fix SQL errors caused by sender insertion from multiple threads
At least on Qt5, the QSqlQuery.lastQuery() function, when used with
prepared queries, returns the EXECUTE statement and not the actual
query. Qt then attempts to PREPARE this statement, which causes
a syntax error, aborting the transaction and dropping the message that
was to be inserted. A better idea would be to execute the named
prepared query directly, which avoids the problem.
Daniel Schaal [Fri, 1 May 2015 14:34:22 +0000 (16:34 +0200)]
Add cmdline options for SSL certificate/key paths.
Add cmdline options to core to specify the SSL certificate
and SSL key path, --ssl-cert and --ssl-key respectively.
If not set, the usual path CONFIGDIR/quasselCert.pem
will be used for the cert and the key.
If only the path to the certificate is set, it is assumed
to also contain the key.
Alf Gaida [Sat, 25 Apr 2015 17:25:34 +0000 (19:25 +0200)]
Fix read GIT_HEAD and GIT_DESCRIBE from environment
add missed ENV
Michael Marley [Thu, 23 Apr 2015 22:05:31 +0000 (18:05 -0400)]
Call socketDisconnected() from socketStateChanged()
This makes sure that socketDisconnected() is always called when
Quassel disconnects even if it isn't fully connected yet.
Otherwise, reconnect occasionally fails to work properly.
Manuel Nickschas [Thu, 23 Apr 2015 20:47:17 +0000 (22:47 +0200)]
Bump version for release
Manuel Nickschas [Thu, 23 Apr 2015 20:46:48 +0000 (22:46 +0200)]
Update ChangeLog
Manuel Nickschas [Thu, 23 Apr 2015 20:34:07 +0000 (22:34 +0200)]
Move config file migration after QCoreApplication instantiation
Seems like Kdelibs4ConfigMigrator accesses QCoreApplication::arguments()
now, leading to a warning on startup if QCA hasn't been instantiated yet.
Michael Marley [Thu, 23 Apr 2015 12:46:43 +0000 (08:46 -0400)]
Execute initDbSession() on DB reconnects
Previously, the initDbSession() function would only be run on the
initial connect. Since the initDbSession() code in PostgreSQL is
used to fix the CVE-2013-4422 SQL Injection bug, this means that
Quassel was still vulnerable to that CVE if the PostgreSQL server
is restarted or the connection is lost at any point while Quassel
is running.
This bug also causes the Qt5 psql timezone fix to stop working
after a reconnect.
The fix is to disable Qt's automatic reconnecting, check the
connection status ourselves, and reconnect if necessary, executing
the initDbSession() function afterward.
Manuel Nickschas [Tue, 21 Apr 2015 20:41:47 +0000 (22:41 +0200)]
Bump version for release
Manuel Nickschas [Tue, 21 Apr 2015 20:41:11 +0000 (22:41 +0200)]
Update ChangeLog
Michael Marley [Mon, 20 Apr 2015 16:43:53 +0000 (12:43 -0400)]
Set the PostgreSQL session timezone to 'UTC'
With Qt5, the PostgreSQL driver will transparently convert times
to the database's timezone before inserting. Because the default
is 'localtime', this causes the local time to be stored in the DB
instead of the UTC time. This in turn causes the time displayed
in the client to be wrong by the same offset as that timezone's
offset. To fix the issue, just make sure the PostgreSQL is in the
'UTC' timezone, so the passed UTC time will not be converted.
Manuel Nickschas [Wed, 15 Apr 2015 20:58:17 +0000 (22:58 +0200)]
Bump version for release
Manuel Nickschas [Wed, 15 Apr 2015 20:57:53 +0000 (22:57 +0200)]
Update ChangeLog
Manuel Nickschas [Wed, 15 Apr 2015 20:56:03 +0000 (22:56 +0200)]
Output the CMake version during configuration
This will allow us to quickly see if too old (or sometimes even
too new) CMake versions cause problems for building.
Patrick von Reth [Mon, 16 Mar 2015 10:55:41 +0000 (11:55 +0100)]
Use the Qca-qt5 import target if available
Manuel Nickschas [Wed, 15 Apr 2015 18:12:08 +0000 (20:12 +0200)]
Merge pull request #121 from mamarley/unusedcipherfix
Fix compiler warning when compiling without QCA support
Manuel Nickschas [Thu, 9 Apr 2015 11:53:29 +0000 (13:53 +0200)]
Merge pull request #120 from mamarley/dbcreatefix
Update the schema creation DDL for the new password hashing
Michael Marley [Mon, 6 Apr 2015 01:54:36 +0000 (21:54 -0400)]
Fix compiler warning when compiling without QCA support
Michael Marley [Mon, 6 Apr 2015 01:35:09 +0000 (21:35 -0400)]
Update the schema creation DDL for the new password hashing
I'm not sure whether this slipped out in a merge or whether I
forgot to add it in the first place, but it should have been part
of commit
5924533c6d0f9777d38c01ed7e1510a55db2b876. This
completely broke setting up a new Quassel instance from scratch
with 0.12-rc1, but upgrades from previous versions are unaffected.
Daniel Albers [Wed, 25 Mar 2015 12:11:50 +0000 (13:11 +0100)]
Fix typo in PasswordChangeDlg failure message
Daniel Albers [Wed, 26 Mar 2014 04:00:33 +0000 (05:00 +0100)]
Update translations from Transifex
26056 translated messages
Many thanks to:
- cs: Jaroslav Lichtblau <dragonlord@seznam.cz>
- da: Jens True <frazzer@hotmail.com>,
Joe Hansen <joedalton2@yahoo.dk>
- de: contradictioned <contradictioned@googlemail.com>,
Felix Kaechele <felix@fetzig.org>,
Manuel Hoffmann <contradictioned@googlemail.com>,
Tobias Frei <tobias@freiwuppertal.de>
- el: Stefanos Sofroniou <inactive+_stefanos_@transifex.com>
- eo: Tinjo Kohen <tiffypet@yahoo.com>
- es: Adolfo Jayme Barrientos <fitoschido@ubuntu.com>,
Jonathan Farnham <jonathan@haitu-yu.com>
- fi: Lasse Liehu <larso@gmx.com>
- fr: Vlavv <vlavv2@gmail.com>
- gl: Miguel Anxo Bouzada <mbouzada@gmail.com>
- hi: Raju Devidas Vindane <rajuvindane@gmail.com>
- ko: Sungjin Kang <potopro@gmail.com>
- mr: Raju Devidas Vindane <rajuvindane@gmail.com>
- nl: simonphilips <simon.philips@gmail.com>
- oc: Cédric Valmary <cvalmary@yahoo.fr>
- pa: A S Alam <apreet.alam@gmail.com>
- pt_BR: Pedro Araujo <pedroaraujo@colorlesscube.com>
- ro: Arthur Țițeică <arthur.titeica@gmail.com>
- sl: Andrej Mernik <andrejm@ubuntu.si>
- sr: Jovan Jojkić <jovanjojkic@gmail.com>
- tr: Volkan Gezer <volkangezer@gmail.com>
- uk: Yuri Chornoivan <yurchor@ukr.net>
Daniel Albers [Tue, 25 Mar 2014 23:10:11 +0000 (00:10 +0100)]
Update quassel.pot
Lasse Liehu [Tue, 25 Mar 2014 22:53:23 +0000 (00:53 +0200)]
These two settings keys shouldn't be translated
Manuel Nickschas [Wed, 11 Mar 2015 22:21:29 +0000 (23:21 +0100)]
Merge pull request #118 from mamarley/randomfix
Fix FTBFS on certain older libstdc++ versions.
Michael Marley [Wed, 11 Mar 2015 21:54:05 +0000 (17:54 -0400)]
Fix FTBFS on certain older libstdc++ versions.
Manuel Nickschas [Wed, 11 Mar 2015 19:47:23 +0000 (20:47 +0100)]
Bump version for release
Manuel Nickschas [Wed, 11 Mar 2015 19:46:47 +0000 (20:46 +0100)]
Update ChangeLog
Manuel Nickschas [Wed, 11 Mar 2015 19:43:25 +0000 (20:43 +0100)]
Merge pull request #116 from mamarley/splitfix2
Improve the message-splitting algorithm for PRIVMSG and CTCP
Chris H [Sun, 14 Sep 2014 17:30:09 +0000 (13:30 -0400)]
Move websearch from botton of context menu to underneath copy selection.
It seems the 'Copy Selection' button got moved between when the patch
was made and when it finally got applied. Got missed during the merge.
Manuel Nickschas [Wed, 11 Mar 2015 19:15:24 +0000 (20:15 +0100)]
Merge pull request #117 from mamarley/oldcoresslconfig
Add back the SSL protocol selection dialog box for old cores.
Manuel Nickschas [Wed, 11 Mar 2015 19:13:12 +0000 (20:13 +0100)]
Merge pull request #111 from mamarley/nosslloopback
Don't require SSL for loopback connections with --require-ssl
Manuel Nickschas [Wed, 11 Mar 2015 19:10:41 +0000 (20:10 +0100)]
Merge pull request #110 from mamarley/passhash
Introduce new password hash versioning system and new salted SHA2-512 hash
Manuel Nickschas [Wed, 11 Mar 2015 18:49:27 +0000 (19:49 +0100)]
Provide a proper dialog for changing the core password
The password change dialog now asks for your old password, and forces
you to confirm the new password by typing it twice, as is customary.
To make this works, we need to add bidirectional communication between
client and core, forcing some more changes to the way the first iteration
of the feature was done.
Note that this commit breaks compatibility with the previous implementation.
If you run a beta1 core or client, you need to upgrade both to make things
work again. Since no real release happened in between, I don't deem it
necessary to provide backwards compat or yet another feature guard.
Manuel Nickschas [Wed, 11 Mar 2015 18:43:21 +0000 (19:43 +0100)]
Make PeerPtr work for RPC calls
While most of the functionality was already in place, it turns out
that we need to enable loading/saving PeerPtr from/to a QVariant in
order for SignalProxy to work properly.
I'm honestly not sure why it seemed to work without this when implementing
the TransferManager, but it's obvious it's required.
Manuel Nickschas [Mon, 9 Mar 2015 20:53:00 +0000 (21:53 +0100)]
Add missing dependency to KWidgetsAddons
This is pulled in by some other frameworks modules anyway, but we should
list it explicitly.
Michael Marley [Fri, 27 Feb 2015 17:31:17 +0000 (12:31 -0500)]
Add back the SSL protocol selection dialog box for old cores.
Cores before 0.10 will default to SSLv3 if the user doesn't make a
selection. If a >=0.10 client is used with a <0.10 core to connect
to a server that has SSLv3 disabled, it is impossible to connect
to that server without upgrading the core or using an old client
to change the SSL protocol settings.
This also changes the SSLv2 and SSLv3 options to indicate their
insecurity and therefore discourage their use.
Cores from 0.10 and up use SSL autonegotiation and to not need the
protocol setting.
This partially reverts commit
e53fc69a91553b57932ba599b39999d550114588.
Michael Marley [Sat, 14 Feb 2015 01:12:18 +0000 (20:12 -0500)]
Don't require SSL for loopback connections with --require-ssl
Encrypting a loopback connection is a pointless waste of computing
power. This patch will allow people to use loopback connections
from things like quassel-webserver without encryption while still
requiring encryption for remote connections.
Michael Marley [Tue, 3 Mar 2015 01:31:02 +0000 (20:31 -0500)]
Add support for SHA2-512 hash on Qt4 builds
Qt4 doesn't support anything higher than SHA1 in
QCryptographicHash, so it is necessary to include a fallback. This
bundles sha512.c from PolarSSL/mbed TLS. It is only used for Qt4
builds and does not require a dependency on external libraries that
might be hard to compile on Windows® or Mac OSX®. It can be
removed once Qt4 is no longer supported.
Michael Marley [Sat, 21 Feb 2015 12:33:57 +0000 (07:33 -0500)]
Improve the message-splitting algorithm for PRIVMSG and CTCP
This introduces a new message splitting algorithm based on
QTextBoundaryFinder. It works by first starting with the entire
message to be sent, encoding it, and checking to see if it is over
the maximum message length. If it is, it uses QTBF to find the
word boundary most immediately preceding the maximum length. If no
suitable boundary can be found, it falls back to searching for
grapheme boundaries. It repeats this process until the entire
message has been sent.
Unlike what it replaces, the new splitting code is not recursive
and cannot cause stack overflows. Additionally, if it is unable
to split a string, it will give up gracefully and not crash the
core or cause a thread to run away.
This patch fixes two bugs. The first is garbage characters caused
by accidentally splitting the string in the middle of a multibyte
character. Since the new code splits at a character level instead
of a byte level, this will no longer be an issue. The second is
the core crash caused by sending an overlength CTCP query ("/me")
containing only multibyte characters. This bug was caused by the
old CTCP splitter using the byte index from lastParamOverrun() as
a character index for a QString.
Manuel Nickschas [Wed, 18 Feb 2015 21:09:54 +0000 (22:09 +0100)]
Bump version for release
Manuel Nickschas [Wed, 18 Feb 2015 21:07:59 +0000 (22:07 +0100)]
Update INSTALL