On a non-xinerama (user-side) configuration it is straight up wrong, as
in such case each screen's coordinates start at 0:0. Which makes the
apps on the smaller screen assume incorrect screen size.
And on a xinerama config it is incorrect as well as it results in access
past allocated data in rects and screen arrays.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
In case there are static TQRegExp instances the regexp engine cash might
be destroyed before them; hence we need some special handling of
construction/destruction of the cache objects.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
This fixes crashes cause by dangling pointer returned from
TQTextCodec::codecForLocale(). That might happen when tqWarning() and
such are called during final cleanup.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
This assures that TQRegExp who checks if there is an active thread
through this interface won't try to access its already destroyed
TQThreadLocalStorage, which might happen during destruction of static
TQRegExp objects.
Bug: https://mirror.git.trinitydesktop.org/gitea/TDE/tqt3/issues/142
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
There were two flags for read-only with different capitalization:
readonly and readOnly. This was probably a mistake.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
mightStartDrag might have been used uninitialized when moving mouse with
pressed mouse button into window.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
Improve TQFont cleanup making sure that all instances of TQFont are
destroyed before TQApplication (or specifically before disconnect from
X11). This gets reed of several valgrind complains about leaks deep
inside fontconfig.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
- rename fontsetRefCount -> fontsetCacheRefCount
- move fontsetCacheRefCount to constructor as setHolderWidget() could be
called more than once
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
As derivative from it e.g. TQPrinterUnixPrivate are getting deleted (e.g
in unix TQPrinter implementation) by a reference to base class.
Also added missing include to the header.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
valgrind was complaining about access to uninitialized data because
unused elements in client_message.data.l array weren't set o any value.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
Before that the allocations of TQThreadStorage objects from the main
thread were never destroyed and memory associated with them were never
freed. The second one isn't a huge problem as at that point program is
terminating anyway (but it still makes valgrind complain). The first one
is the bigger issue as destructors might contain some essential external
cleanups like removing temporary files.
Also make `TQApplication::guiThread()` return `0` when the thread is
destroyed (may happen on the program exiting during destruction of
statics).
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
Also rearrange code a bit so it would be obvious that the function
doesn't return a dangling reference.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
TQXIMInputContext::setHolderWidget() function may be (and actually is)
called more than once. This results in multiple instances of the same
object being added to ximContextList. But the destructor removes only
one instance, which effectively results in leak of several bytes
whenever a window is opened.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
As for now here are two sets of inputmethod options:
- -enable-inputmethod/disable-inputmethod - which seems to supposed to
control whether build the 'inputmethod' module or not
- -inputmethod/-no-inputmethod - which seems to supposed to
enable/disable inputmethod support without changing the ABI.
Before the patch both -disable-inputmethod and -no-inputmethod were just
breaking the build: -no-inputmethod were disabling some code with
support for the module, but didn't disabled the module build itself nor
build of plugins. -disable-inputmethod were disabling build of plugins
and module, but didn't disabled code depending upon it.
It seems the inputmethod support were still WIP when the last release of
Qt3 came to be, hence the mess.
This patch fixes the build if both -disable-inputmethod AND
-no-inputmethod are supplied. Disabling only one is not enough due to
tqmake/configure have problems handling two different options of the
same name. Later the -inputmethod/-no-inputmethod should be probably
removed entirely.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
The analogue of XkbKeycodeToKeysym() used to be XKeycodeToKeysym(), but
it was deprecated in favour of XGetKeyboardMapping() method.
Signed-off-by: Alexander Golubev <fatzer2@gmail.com>
This is the first part of the replacement process.
Usage of Q_WS_* has been replaced with the equivalent TQ_WS_*.
Definition of Q_WS_* has been mirrored into TQ_WS_* defines, to allow
TDE code to continue building till replacement is carried over to all
other modules.
Once that is completed, the original Q_WS_* defines will
be removed.
Signed-off-by: Michele Calgaro <michele.calgaro@yahoo.it>
A new README with markdown format, updated information on TQt, contributor section and translations status.
The old README has been preserved as README.Qt3 for historical reasons.
Signed-off-by: Mavridis Philippe <mavridisf@gmail.com>
Alexander Golubev
Hugo Carvalho
Michele Calgaro
Slávek Banko
Mavridis Philippe
Andrei Stepanov