Clients which implement a client-side modality might cause
livelocks by reverting the focus to the transient (after the
WM tried to put it on the leader as the transient's modality
is unknown)
So while cycling we revert the focus whenever it moves somewhere
where we don't expect it.
When done, we also focus the window that should have the focus anyway
to allow the client to redistribute the focus (as we prevented it
during cycling)
Hall of Shame: Softmaker Freeoffice uses (only) client side modality.
Allow setting relative value for x and y or width and height separately in
the apps configuration file. This makes these settings compatible with ones
available in the keys file.
Previous buggy behavior:
If someone has specified, e.g. "[Dimensions] {50% 100}" it was parsed as
"{50% 100%}" not as "{50% 100px}" which was inconsistent with the "keys"
configuration file.
From now on it is possible to write something like this:
[app]
[Position] (RIGHT) {50% 0}
[Dimensions] {300 100%}
[end]
Signed-off-by: Arkadiusz Bokowy <arkadiusz.bokowy@gmail.com>
A bug sneaked into my implementation of Boyer-Moore-Horspool. This lead
to not finding certain patterns. Given the text 'abcdde' and the pattern
'dd', the faulty implementation would not find 'dd':
1. 'ab' does not match, skip 2 (length of pattern)
2. 'cd' does not match, skip 2 (length of pattern) <- the bug.
3. 'de' does not match, end of string
The bug in step 2 is to not use 'd' to detect how far to skip but to
use 'c' (which is not in the skip-table) and thus 2 bytes are skipped).
Different styles makes the menu width different.
When the original menu width is bigger than the newly selected style's
width, the rendering produces pretty strange effects:
The old style's frame not cleared, so it was rendered and visible next
to the new style edge.
With this change, the menu width will be as wide as the widest menu
item.
Style switching still not perfect, because the height of a menu item is
from the "first" selected menu, also font color are not updated.
We define the value ICONV_NULL = -1, but when we attempt to set the
s_iconv_convs array to all NULL values, we zero the array instead of setting
its entries to -1.
This patch properly initializes and wipes s_iconv_convs.
Instead of creating the titlebar buttons with a size of 10x10 pixels
and rely on resizing later on we now pick the correct dimensions
right on.
This fixes also bug #1125 ("Detaching a window from a tab-group renders
app-icon to 1/2"); the problem also occurred on restart.
I took the chance to refactor a little bit.
Fluxbox now supports three MenuSearch modes:
* NoWhere - essentially "disabling" the menu search.
* Somewhere - the search string matches somewhere.
* ItemStart - the search string matches at the start of a menu item.
The default value is "ItemStart", just in the good old times. As long as
this feature is not configurable via the menu it would irritate users
with distinct muscle memory who type without thinking OR checking the
visual feedback: they would trigger items they did not intent to trigger
after years of the old behavior.
Once this feature get's an entry in the config menu the default value
might change.
Under some circumstances a WinButton might be tiny (1x1 pixel). Substracting
values > 1 yields either 0 or something reaaaaly big (UINT_MAX). This will
make fluxbox either crash or make it act weird. We don't want this.
This commit fixes a subtle regression introduced by myself in
f64727ca: WinClient wc = FbMenu::window()->winclient() creates a
copy and this copy gets cleaned up at the end of the scope. What
I really wanted was a reference.
Among the first steps to produce better i18n support is to test the created
translations adhoc without running "make install". To achieve this, fluxbox
now honors several environment variables:
- NLSPATH: fluxbox won't create the absolute path to the catalog and thus
catopen() is free to use NLSPATH as described in the manpage. Example
given: "/tmp/%N" will pick "/tmp/fluxbox.cat". %N refers to FLUXBOX_CATFILE.
- FLUXBOX_CATFILE: By setting FLUXBOX_CATFILE the users can make fluxbox to
use a different name for the catalog file. Default: "fluxbox.cat"
- FLUXBOX_CATDIR: Per default fluxbox tries to find FLUXBOX_CATFILE at several
places. Setting this environment variable allows to point fluxbox to a
different search path for the catalog files.
Then, fluxbox tries catopen() first without changing the deduced catalog file
name. After that it applies some heuristics to get a good catalog file name.
When the windowmenu file is empty in ~/.fluxbox/init the code falls
back to the default. In ab2eef4 I forgot to set the rc_path which
caused this bug to happen.
gcc-4.2.1 on OpenBSD-5.6 hinted that strcpy() is not the safest function on
earth. While seeing the code I wondered why it we first create copies
of the names at all (let alone using memset() and then strcpy() after it).
In f64727ca I removed one 'else' too much. As a result all [maximized] lines
were mapped to MAX_NONE. Fixed.
The rest of the commit is just some cosmetic to reduce indentation and shorten
function names, easier to read.
This commit implements a tweak to the typeahead feature already existent in
fluxbox: If the user opens up a menu and starts typing, fluxbox tries to
detect matching menu items and makes them available for quick selection.
The typed pattern is now search also in the middle of the text.
I opted to strip down the code quite a bit and remove complexity by throwing
out FbTk::TypeAhead and FbTk::SearchResult because I do not see the need for a
general solution when the only use case for such a feature is in fluxbox'
menus. FbTk::ITypeAheadable shrunk down to 2 functions; the whole file might
be combined with the code that implements FbTk::Menu::TypeSearch.
FbTk::Menu::setIndex() and related code is also gone: the position of each
menu item is defined by it's position in the items container. This reduces the
mount of book keeping fluxbox has to do. Fewer moving parts is a good thing.
It's possible that users start to complaint because they expect their
typed pattern to match only at the beginning of the text OR that some
demand other tweaks. We will see.
This commit also fixes a regression introduced by 8387742c. The bug made
the menu vanish.
'width' might imply something 'aligned horizontally'. The first parameter
of maxTextLength() is given in pixels. To avoid confusion, the name of
the parameter is changed.
The comment before the binary search reflects better why and what we
need to do.
A vertically rotated text should be checked against the height of the
button, not the width of it.
Fixes#1119 ("Toolbar cannot display full text of workspace/clock/window
title.")
Sometimes fluxbox is used as a component of a desktop environment (
e.g., the "Lumina Desktop"). It's easier and more stable for this kind
of "users" to disable the toolbar or the slit alltogether. Why not
configure these features out? Because then they would need 2 installations
of fluxbox: One without the slit and the toolbar (and perhaps other
components) and one with them.
This implements feature request #314, at least partially.
This commit also looks bigger than it actually is: I refactored and cleaned
up classes Fluxbox and BScreen.
There was a subtle flaw in the way fluxbox detects to which BScreen a
given Window belongs: We have to compare the RootWindow of the given
Window against the RootWindow of each BScreen.
That underlying flaw made _NET_REQUEST_FRAME_EXTENTS fail: the code path
needs a valid BScreen for the given window, otherwise we return early.
Closes#1121.
std::remove_if() removes an item from a container and returns a
past-the-end iterator ... which equals m_icon_list.end(). As a result
the check
if (erase_it != m_icon_list.end()) {
iconList().erase(erase_it);
iconListSig().emit(*this);
}
will never succeed and thus the iconListSig() will never be emitted.
There was a problem deep within how the menus were connected and when and
what gets deleted. It was clearly related to a menu which was kind of
global. In order to better understand the code flow I eliminated the
ExtraMenu code: it was used only to get the Remember-Menu into the
Window-Menu. Instead of having a singleton of the Remember-Menu and fight
against the shaky interconnections we just create a new one on demand
and delete when the menu gets deleted. Looks like this fixes the problem.
The menu code needs more love anyway.
Closes#1118
This reverts 58e09b7190 after having regressions
and bug reports in regards to autorepeated key events are not working anymore.
So, the problem with the lost keypress after the workspace change is still
open but at least the behavior for other folks is as it was. Since 58e09b7190
seems to be corelated to "auto repeat" keys it might indicate the problem
which 58e09b7190 "fixed". We will see.
Fixes#1115.
Note: I expect only modern compilers will hit the source of fluxbox. It
seems futile to guard stdlib headers while demanding <algorithm> etc. This
should trim down the noise in the source quite a bit.
Part of this massive looking but simple rather simple change is a the way we
check for active Xinerama: XineramaIsActive() was used before. The docs for
XineramaQueryScreens() state that
XineramaQueryScreens() returns NULL and sets number to
0 if Xinerama is not active.
Sold.
Every time either the Slit menu or the Toolbar menu were added to the
root menu, the whole root menu was (re)created from scratch. Now we
create and remove only the menus needed.
Side effect: the position of these menus is now at the end of the root
menu and not somewhere in between. If users complaint, I' ll think about
it.
This commit is a preparation step for some menu cleanup ahead. To make it
easier to understand which types of MenuItems are added where and when, I
converted the overloaded FbTk::Menu::insert() functions into explicit ones
(Menu::insertSubmenu(), Menu::insertItem(), etc. ). This makes it possible
to just grep for 'insertSubmenu()'.
Side effect this commit: it trims down the very verbose creation of menu items
in regards to how the labels are created.
Minor: FbTk::Menu::shown and FbTk::Menu::s_focused are moved out of the class.
The font might have an impact to the result of width() and height(). In order
to render the new pixmap correctly fluxbox now first sets the font and the
width of the border and after that it recreates the pixmap.
When fluxbox comes up some of it's drawables span a 1x1 area. Subtracting from
such small numbers bigger ones always lead to massive problems when 'unsigned
int' are involved:
'button_width' is an unsigned int of '1' (this might be caused by another
issue or on purpose, anyway), subtracting -10 or any other number should
result in something < 0 when in reality an integer underflow happen: max_width
is now MAX_INT-something big. This makes fluxbox crash under certain
circumstances.
This fixes bug #1116.
* Correctly build data files when operate outside of the $(srcdir)
* Actually install data files
* Use pkg-config to detect X11, works better on MacOSX. We used pkg-config
already anyway, double checking for X11 does not yield better results.
* Replace EXEEXT in some files while before installing them
xxx_LDFLAGS place the libraries like '-lX11' or '-lXft' at the beginning of
the linker command. Some systems were not able to handle the situation and
this lead to a lot of unresolved symbols. Reading the documentation of
automake:
... you can use LDADD to do so. This variable is used to specify
additional objects or libraries to link with; it is inappropriate for
specifying specific linker flags, you should use AM_LDFLAGS for this
purpose.
In addition we link against -lm in order to satisfy symbols refering to 'cos'
and 'sin'.
POSIX states that 'd_name' in 'struct dirent' is char[], so it cannot be NULL.
This will result in the compiler complainting about an expression which always
evaluates to true ... for this compiler (clang). But in some implementations
'd_name' is a 'char*' that's why it's better to keep the check for possible
NULL.
OSDWindow::resize(const FbTk::BiDiString&) shadowed FbWindow::resize(x, y). To
fix this I renamed the function to OSDWindow::resizeForText() to make the
intention clear.
While running code analysis tools this issue popped up. As it is written it is
clearly a NOP, but the reason for why it is written that way remains unknown.
I suspect that it was a textfragment introduced by some repeat-function of the
used editor (think '.' in VIM). The code was introduced by commit a932a7a801
and looked like this in the original form:
if(!XQueryTree(FbTk::App::instance()->display(), window().window(),
&root_return, &parent_return,
&children_return, &nchildren_return))
parent_return=parent_return;//return;
Look at the outcommented return statement. I decided to return from that
function instead of the NOP.
this commit implements feature-request #317: "Add support for GTK dockapps.":
"Back in 2010, WindowMaker implemented a system where windows with WM_CLASS
res_class = DockApp would be treated as if they had initial_state =
WithdrawnState, since GTK refuses to allow this."
kudos to Arkadiusz Bokowy (this commit is a slightly changed version
of a patch sent to the devel-ml): when retrieving the '.res_name' of
a XClassHint we should check '.res_name' and not '.res_class'.
the other changes only reduce the code.
handing over the dimensions of a WinClient client must not contain properties
of the FbWinFrame, otherwise they get added twice in
FbWinFrame::moveResizeForClient() and thus result in a resizes when no resize
is wanted.
other changes: it's easier for me to detect the nth bit when the value looks
like (1 << 8) instead of 0x0100 (for the 8th bit). that is why i changed
0x0100, 0x0200 etc. in the nearby code.
the -1 in the FbTk::FbTimer::remainingNext() function was pointless in the
first place anyway: reducing the timeout by just 1 microsecond improves
nothing (in this case). if the timer triggers exactly at a full unit (second)
then it's correct to wait for the full next unit.
the lag / skipping of the clock was not caused by faulty timer code
on fluxbox's side but by the behavior and inner workings of time().
since this is fixed now (913244789f) we can now rollback ec7fe513c8
and detect strftime-formats which need intervals of seconds or minutes.
minor: the small change to FbTk::Timer::setTimeout() reduces one
start() / stop() cycle for a running timer.
from time to time (...) time() might be off to gettimeofday() by a
second. the reason for this is that time() is usually implemented
by just returning the field 'second' of the struct that represents
the clock inside the kernel. gettimeofday() on the other hand also
takes the 'fraction' field (mostly 'nanoseconds') into account and
thus is closer to the current time than time().
the result of using time() was a perceived 'lag', sometimes the
clocktool even skipped a second. by using FbTk::FbTime()::system()
instead fixes the issue.
xmodmap (and other tools) trigger MappingNotify events. a single xmodmap
expression such as "keycode comma = comma semicolon" might trigger 4 or 5
MappingNotify events. loading the keys file on each of them is quite
unefficient.
fluxbox now uses a (250ms) timer which is reset upon further events.
there is problem that x/y ended with unsigned int value due to
width()/height() and negative result of division ended up being big
it causes Focus to move window due to screen boundary checks
fixes annoying behaviour of window moving few pixels with
Mod4 KP_8 :MacroCmd {ResizeTo 100% 50%} {MoveTo 0 0 Top} {Raise} {Focus}
This commit fixes primarily a race condition that occurs when xinit(1) shuts
down: by not acting properly fluxbox gets caught in an infinite loop. It
caused bug #1100.
xinit(1) sends a SIGHUP signal to all processes. fluxbox tries to shutdown
itself properly by shutting down workspaces and screens. While doing that, the
Xserver might be gone already. Additionally, fluxbox used to restart() itself
on SIGHUP, which is clearly not the right thing to do when xinit(1) is about
to end the session.
So, fluxbox does this:
* handling SIGHUP now shuts down fluxbox without clearing workspaces and
screens.
* A 2 second alarm() is triggered in Fluxbox::shutdown() as a last resort
* XSetIOErrorHandler() is used to recognize the disconnect from the xserver.
* SIGUSR1 is for restarting fluxbox, SIGUSR2 for reloading the config
* FbTk/SignalHandler.cc/hh is gone; this unused abstraction served currently
no real purpose. Signal handling is now done in main.cc
* Unrelated to the issue itself src/main.cc was trimmed down quite a bit and
the code (responsible for handling the command line interface) was moved to
src/cli*
* Fix integer wrap around for some situations (resulting in a very wide
window):
w = max_width - x_offs - (*closest)->widthOffset()
If the given window is on the right most display and thus 'x_offs' is
bigger than 'max_width' (half of the display width), the resulting 'w' wraps
around and becomes very large.
* Place a single window via ArrangeWindowsStack* as well
* Some minor code styling / reordering along the way
Do not Ungrab the keyboard for a KeyPress event, this causes us to lose the
KeyRelease event which is needed to ungrab the keyboard after the event
happened.
std::bind2nd() has problems with the existing call because it uses the
argument type passed to std::equal_to() in std::binder2nd()'s constructor,
but WinClient* does not have the const specifier that `client' does.
The call works fine with libstdc++ because of the way it implements
std::bind2nd(), but fails with libc++ and possibly with other STL
implementations as well.
usually $HOME is set when fluxbox runs. in some rare scenarios (eg., fuzzying
binaries to detect bugs) one could launch fluxbox by using 'env -i' and thus
eliminating $HOME from the environment. to prevent crashes fluxbox uses now
'getpwuid()' when $HOME is not set to detect the home folder.
Detecting very long window titles is done via FbTk::TextUtils::doAlignment().
Instead of removing one char from the title at a time to see if it fits into a
given 'max_width', we now use a binary-search like approach to get faster to
the right value. This massively improves the speed for windows with
(arbitrary) long window titles (see bug #1090, javascript
document.title = new Array(4999).join(".");
leads to massive waiting for fluxbox to detect that this window has a very
long title).
In addition to that Xft returns 'wrapped' shorts ('integer overflows') for
long texts: XGlpyhInfo.xOff is declared as signed short, it's able to hold
~32k pixels. A monospace font with font-size 10 produces an integer
overflow after 3276 chars / glyphs, thus rendering the check
if (text_width < max_width) { /* ... */ }
pointless and leading rendering the whole title. By calculating some kind of
upper limit for a pseudo-wide glyph ("WW") and strictly cutting off the input
string at that limit prevents this issue.
This commit alters XRANDR (X Resize And Rotate) extension dependency,
which is expected to have at least version 1.4. Earlier old versions of
xrandr were supported, at least to some extent.
Do not try to be too smart which compilations need config.h, as most of
them will simply because of the config.h has information about system
capabilities.
* a reasonable initial placement is important for later movements to
different heads and correct head detection (required by apps file)
* it did not work well in case when (0,0) was not near any head
* if you have e.g a 1920x1200 monitor and a small 800x600 monitor to the
right of the bigger one and a small window at the right side of the
big monitor (but still the whole window area at the big monitor) then
the original head detection would claim the window is on the small
monitor because (800/2+windowWidth/2 < 1920/2+windowWidth/2) is true
for small windows; but that is obviously wrong, the window is entirely
on the big monitor
* these incorrect head detections did lead to incorrect window
placements as they were required in the apps file
Commit 79fe2fca1d checks for pending
motion events and drops out of the FluxboxWindow::motionNotifyEvent() function
early if so. When the user does not use the opaque window movement method an
outline will be drawn to the screen. That outline was not cleaned correctly
with commit 79..
