This patch avoids to create Cached Icons for all applications. Only
the applications docked should create it, as Zoltan said:
-------8<-------
Also app icon caching was broken around the same time. The app icon cache
in CachedPixmaps was meant to store icons retrieved from X clients so the
dock or clip can display those when the client is not running like after
startup. The cache should contain only such icons and the path should never
appear in WMWindowAttributes because the cache is an internal thing used to
look up icons not otherwise available. If you look at your WMWindowAttributes
now it is full of entries referring to the cache that should not be there and
if you look at the cache dir you'll find a lot of icons from all apps you've
ever started while there should be only the few docked ones that use client
side icons. Also the cache is never cleaned up only new icons are added to it.
-------8<-------
Signed-off-by: Rodolfo García Peñas (kix) <kix@kix.es>
As Josip Deanovic reported:
-------8<-------
In previous versions e.g. 0.80.2 up until 0.95.3 when an application
attributes are set with "NoAppIcon = Yes;" ("No application icon" option
in attributes window), it was possible to launch multiple instances of
the application from wmdock using double-click.
After doing a git bisect per your suggestion I have found and reported
this:
bc0700e016 is the first bad commit
commit bc0700e016
Author: Rodolfo GarcÃa Peñas (kix) <kix@kix.es>
Date: Mon Jun 18 11:15:19 2012 +0200
Create WAppIcon always
When the application is created, the WAppIcon now is created always,
but it is only painted if the flag is not set.
The icon initialization to NULL can be done now at
app_icon_create_from_docks
because it is always called.
:040000 040000 7c58877ad533d52affb3 M src
-------8<-------
This patch reverts this change (not the patch). Now the function
create_appicon_from_dock checks if the flag no_appicon is set,
and then, do not execute the code related to the appicon.
Because the connection between the icon and the window is broken
(icon->owner is null) we need check if the icon->owner exists
when we try to re-create the icon in the Window Attributes window.
Signed-off-by: Rodolfo García Peñas (kix) <kix@kix.es>
This patch introduces two new configuration values, SnapEdgeDetect and
SnapCornerDetect, which users can set to change the distance from an edge
or corner at which window snapping will begin. The defaults are 1 and 10,
respectively.
Suggested-by: Josip Deanovic <djosip+news@linuxpages.net>
As reported by Josip, the code in Window Maker to detect the use of the
legacy keyword "MiniwindowApercuBalloons" and "ApercuSize" was broken,
which means they were always seen as used even when not present.
This patch fixes the detection to only use them if they were effectively
used.
Reported-by: Josip Deanovic <djosip+news@linuxpages.net>
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
In the crash handling dialog there was a reference to an old email address.
This patch replaces it with the official one, taken from the 'configure.ac'
definition so it will not need manual update anymore.
The mail address is inserted in the string with a '%s' in order to not have
it in the translation files (po) to ease the maintainer's task in case of
future change.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Since C99 we have floating point functions available for single precision,
so as it is what we need we detect them (configure) and use them when
appropriate. The goal is to avoid unnecessary float->double + double->float
conversion.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
To preserve the accuracy of the operation, the C standard request that the
mathematical operation is performed using double precision, but in many
case this is not necessary so this patch fixes a few constants to avoid
that conversion.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The standard specify that the pseudo-selection which is used to
cooperatively replace window managers should actually contain the version
of the ICCCM standard which is supported. As this could be handy for
compatibility checks in the future, let Window Maker comply.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As it is really unlikely that in normal use case someone would need this
feature, it is now conditional code, which is not enabled by default; the
configure scripts now propose a '--enable-wmreplace' option to enable the
corresponding code, as people making package for distributions may want to
enable the feature to provide users the ability to give a try of all the
window managers.
Suggested-by: Carlos R. Mafra <crmafra@gmail.com>
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Use the same logic used by xfwm4, metacity et al to replace an existing
window manager on the screen and allow other window managers to replace
us, as defined by the ICCCM 2.0:
http://www.x.org/releases/X11R7.6/doc/xorg-docs/specs/ICCCM/icccm.html
Communication with the Window Manager by Means of Selections
By convention those window managers try to become the selection owner of
the WM_Sn atom where n is the screen number. If the atom is owned by
another window manager and the --replace argument was not given to wmaker
we fail to start. If the argument was given we try to become the new
owner and wait for the existing window manger to exit.
After a successful startup we watch for SelectionClear events on the
atom and initiate a shutdown if one arrives, as that implies that
another window manager was started with --replace.
As reported by Nerijus Baliunas and Paul Jakma, the GNOME application,
which use the GTK toolkit, are asking to have no window decoration. This
can be solved by editing the window's attributes in Window Maker, but this
can be tedious when there are many GNOME application used.
This patch adds a configuration option: Window Maker tries to detect for
GTK-based windows and in this case ignore the decoration hints that were
provided by the application.
Suggested-by: Paul Jakma <paul@jakma.org>
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As reported by Nerijus Baliunas, there was a problem when unchecking an
attribute in the Window Inspector and saving it. The original code was
meant to save an attribute that is being checked by user, but not one that
is explicitly unchecked, which means than although it looked ok when using
the "Apply" button, it was not remembered when restarting the application.
In continuation to the clean-up started in the previous patches, this one
is updating the Window Inspector to display 2 check-boxes, one read-only on
the left, displaying the state requested by the application, and a second
one which makes use of the new Tri-State button in WINGs to let the user
specify if he wants to force-on, force-off, or leave as-is the attribute.
The saving to the property list is then updated to take into account this
new 3-state when saving to the file, so relaunching the application will
remember correctly the user choice.
Reported-by: Nerijus Baliunas <nerijus@users.sourceforge.net>
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As there is already an attribute to allow changing this behaviour, and this
is being using for the Net WM hints, there is no reason to not support it
for MWM Hints also; contrary to the initial guess suggested by the comment
it costs actually nothing to properly support that.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The structure containing the information on windows contains 2 sets of
attributes, client_flags which contains those asked by the application
through Hints (like MWM Hints and others) and user_flags which was defined
to allow the user to override them.
Unfortunately many places of the code was using the wrong structure to save
the attributes to (for example by using the WSETUFLAG macro) which was
merely ok as the user_flags have priority, but when we want to provide a
clean consistent behaviour to users, we need to get things in the right
place.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The Window Inspector is used to let user change a list of advanced options
for all the windows of an application. This list was defined through many
hard-coded things; by defining an array with everything at the beginning of
the file it is easier to maintain (the code is simpler because it is more
generic) and to make it evolve.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The Window Inspector is used to let user change a list of advanced options
for a window. This list was defined through many hard-coded things; by
defining an array with everything at the beginning of the file it is easier
to maintain (the code is simpler because it is more generic) and to make it
evolve in the future.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The Window Inspector is used to let user change the list of attributes for
a window. This list of attributes was defined through many hard-coded
things; by defining an array with everything at the beginning of the file
it is easier to maintain (the code is simpler because it is more generic)
and to make it evolve.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by the "checkpatch.pl" script, one line was too long in respect
of the coding style.
This line contains the calculation of an offset when storing a value in an
array, as this offset is a constant during all the loop, this patch is
calculating the offset only once before the loop and then uses this result,
which should make the code faster (although gcc may already optimise this
kind of things), makes it compliant with coding style, and takes the
opportunity to explain the reasons behind this offset.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
By defining the const array with everything at the beginning of the file,
it is easier to maintain and to make evolve.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
By defining the const array with everything at the beginning of the file,
it is easier to maintain and to make evolve (all stuff at the same place).
Merged the function makeMakeShortcutMenu into makeOptionsMenu because it
allows to use a simple callback in the structure like for the other
submenus and because its complexity did not justify a dedicated function.
Took opportunity to give a more appropriate name to the index
MC_DUMMY_MOVETO which did not mean what the entry is about.
As a side effect, this should close Coverity #50251 which was about saving
the return value of function into a variable but not using it, which is not
really good for maintainability.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
There was the possibility to include a global sub-menu to the root menu
using a dedicated file, but this does not look like it is really useful as
this root menu is already defined by a user file, and the root menu is
already specific enough that it does not need anything more 'global'.
Furthermore, the variable enabling this feature (GLOBAL_SUBMENU_FILE) is
not defined anywhere (neither from "configure", nor in "wconfig.h", ...) and
is also not documented at all.
Considering it was introduced a very long time ago (1999!) and was not
touched anymore, this patch removes the dead code associated with it.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
It is not really efficient to call it twice because the strings will not
change, and by using the appropriate trick it can make the code smaller,
with less redundancy, so less prone to bugs and easier to maintain.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The function does not need to be passed a array of pointer to windows
because it does not change content of the pointers. It is more efficient to
directly use an array of windows as the parameter, and simpler also.
As 'SlideWindow' is also concerned, it can be simplified and turned into
an inlinable function so the compiler can optimise it.
Took opportunity to de-CamelCase the name of the functions to comply with
the project's coding style.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Changed the code to return as soon as the result is known because it makes
the code simpler to understand, which is good for maintainability.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
It is generally considered bad practice to place an assignation inside the
expression for an "if" statement because it is often a source of bug,
because of possible typos and because it makes reviewing code more
complicated.
This patch fixes as much cases as possible to make the code easier to read.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The code for animating stuff needs some parameters in a few functions,
which are not used when animations are disabled at compile time, so the
compiler issues some warnings about them.
This patch adds the appropriate statements to disable declaration that
are not needed in this case, and tell the compiler it is ok to not use
the parameters not needed, thus fixing the warnings.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The old behaviour was to expect the user to go modify manually a source
file which is not a great idea because that's typically the kind of things
in charge of the configure script.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The usual way to define a macro in is to name macro with 'USE_xxx' when
they are used to enable a feature 'xxx'.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by the new check script, the preference "show_clip_title" is
defined as a Boolean value in the preference loading function, which uses
the callback function "getBool", which expect a variable of type "char",
but the variable was not defined with that type.
This patch changes the type for the appropriate one to avoid possible
problems.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Because the C compiler cannot check the consistency between the type of the
variable being pointed to, and the type expected by the call-back function
that is used to parse and store the value from the configuration file,
there is a risk of mismatch that can cause Window Maker to misbehave due to
data corruption, which depends strongly on the configuration values and the
architecture on which Window Maker is running.
This patch introduces a script that checks the consistency where possible, to
raise the issues when performing "make check".
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The original code assumed that the sizeof returns a long int type, which is
true on 64 bits platform but not on 32 bits platforms, as pointed by gcc.
The new code does an explicit conversions to 'int' (which should be enough
for that case) so the format specifier in printf will always be ok,
whatever platform Window Maker is compiled for.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As reported by Martin Dietze, when switching workspace (mainly with
keyboard shortcut) while the focus was set on a sticky note from the
GNUstep application "Affiche.app" would cause Window Maker to crash.
This crash was due to the application creating its menu with an Omnipresent
state; Window Maker tried to be smart by smartly setting the focus to this
window after the workspace change, unfortunately when removing the window
of the note from the screen the application would decide to remove also the
omnipresent menu (because it is not necessary anymore).
Because we kept a pointer to an outdated WWindow, we would silently corrupt
memory, which would later cause a crash in an unrelated place (fortunately
in this case it happened quite soon).
This patch adds a check to make sure the window we want to focus is still
a valid one, and if it is not we just ignore it to fall back on another
mechanism already in place for picking the window to focus.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
There are some risks associated with the way arguments are used in macros,
and using a function also allows check on the type of arguments and leaves
more room to the compiler for making the best optimisation choice; it also
allows writing easier to read code (and thus, to maintain).
As a side effect, this should also help Coverity in avoiding false positive
bug reports (like #109605 and #109607).
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, the pointer in wwin->frame have already been
dereferenced many times in the function, so it is useless to include a
check later; removing it makes the code smaller thus easier to maintain.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, there was a memory leak because the buffer used to
create the wmsetbg command is allocated twice.
Because it is not really necessary to call 'wmalloc' for such a small case
(it is not efficient and participates to memory fragmentation), this patch
replaces the dynamic memory allocation by a buffer on the stack, which also
solves to de-allocation issue.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, there is a safety check on the number of workspace
which aborts the function, but the storage memory have already been
allocated so it would leak this buffer.
The case where the number of workspace is 0 is probably not supposed to
happen (there should always be at least 1 workspace, the current one), but
it is better to keep safety checks, so this patch is moving the check at
the beginning so no leak will occur.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, when increasing the size of the array allocated to
store the pointers to menus in a cascaded menu, the incorrect value was
used in argument to the sizeof which lead to over-allocating memory.
This patch replaces the name of the structure (which should have been the
pointer type) by the variable actually being used, fixing the size issue
and making maintainability easier by tracking the type of the variable
which is less prone to bugs in case of change.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As reported by Coverity, the return value for XGetWindowAttributes is
usually checked in many places, because it is a good practice to ensure
that it did work, but it was not checked when called in the function
'wClientCheckProperty'.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The usual way to define a macro in is to name macro with 'USE_xxx' when
they are used to enable a feature 'xxx'.
As the feature concerns the Dock and not DnD in general (WINGs support is
not disabled), make it a bit more clear in the macro name and document the
configure flag in the Installation Manual.
Took opportunity to compile the corresponding file only when the feature is
not disabled.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The old behaviour was to expect the user to go modify manually a source
file which is not a great idea because that's typically the kind of things
in charge of the configure script.
As a side effect, we can now use an automake conditional to avoid compiling
the source file in charge of the feature when not used, instead of trying
to compile an empty-looking file.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The usual way to define a macro in is to name macro with 'USE_xxx' when
they are used to enable a feature 'xxx'.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
There were 2 options that were not documented at all, which are now listed
in the 'wmaker --help';
The man page have been updated according to the 'check-cmdline-opts' script
feedback, including:
- adding 2 options that were missing (global_def_path, no-drawer);
- removed 2 options that do not exist anymore (no-cpp, create-stdcmap);
- sorting the options, to make them easier to search for.
In addition, took opportunity to:
- remove the double description for 'visual-id';
- improve a number of descriptions to make them clearer;
- fix minor stuff (typo, spaces at end of lines).
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As reported by Juan Giordana, the autocomplete/history feature which is
available to Run dialogs brought up by the root menu is not available to Run
dialogs brought up by a keyboard shortcut. This patch enables this
behavior.
Autoconf have been providing the option '--localedir' for a long time now,
so this patch removes the deprecated '--with-nls' option and makes use of
the standard '--localedir' instead.
Took opportunity to define the path in the 'config-paths.h' in the same way
the other paths are defined to be consistent, which also simplify the
compilation commands.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
