Focusing child from a workspace should now skip over the floating con and
go directly to its child. Focusing parent from that grandchild should leave
the workspace focused again.
As the workspace might be reached via recursion (e.g. moving from the edge
of a fullscreen split container), it's necessary to check for a fullscreen
container whenever a workspace is reached.
If a window with _NET_STARTUP_ID set is moved to another workspace, it
will delete any associated startup sequence immediately. This will also
occur if a window has a leader with _NET_STARTUP_ID set, if the leader
has no container (never been mapped).
A startup sequence may also be deleted if it's matched by
startup_workspace_for_window() and its 30-second timeout has elapsed.
This avoids a case where a fullscreen container could be moved onto a
workspace that already had its own fullscreen container, leading to
two fullscreen containers on top of each other.
Previously, while the loop would not be executed anymore, the callback
(for actually repositioning the window after a MotionNotify) would still
get called once. This commit avoids that, thus fixing the crash.
fixes#819
Fix a crash that occured when moving a window to the scratchpad that is
seemingly floating to the user, but actually a descendant of a floating
parent con (and itself non-floating). If that is the case, move the
floating parent container to scratchpad instead of the window.
fixes#740
This commit breaks floating window keyboard focus order
(t/135-floating-focus.t) when you have > 2 floating windows. Since
keyboard focus is more important than saving one click to raise floating
windows, I revert the commit.
Note that we cannot implement this without keeping a third list (beneath
floating_windows and focus) for the z coordinate of a floating window.
This seems not worth it.
This reverts commit 064be457e5.
This makes our signal handler useless and leads to infinite SIGSEGV
loops because the ev callback handler gets called only from within the
event loop, and control doesn’t necessary get to the event loop…
This reverts commit 514265b529.
If there is a single floating con on a separate workspace that is not
focused, and this con becomes urgent, switching back to that workspace
may result in a crash of i3. This is because while setting the urgency
of parent containers, 'parent' may become NULL in case of floating
containers. This commit checks the validity of parent.
fixes#821
We need to verify that setting the event mask works, and we need to
include StructureNotify to get unmap events at any point in time.
Thanks darkraven for the pointer.
fixes#718
Previously, when you had an urgent container in a stack on some
invisible workspace (say urxvt) and you switched to it, the urgency hint
was not properly cleared.
Previously, i3 would send width=0, height=0 to windows which were put on
workspaces created by an assignment (that is, invisible workspaces,
which do not get rendered normally).
fixes#653
The removed code was add by commit 61b8a62 to fix#564. That bug is cause
by rendering the mplayer window again as a floating window (even if it
has been rendered before as a fullscreen window, at line 202). So simply
checking for fullscreen window is enough to solve this problem. Treating
floating/tiling fullscreen window differently is not needed.
A good visualization of the new algorithm is this:
+--------+
| |
+--------+=| S1 |========================
| | | |
| S0 | +--------+
| | +--------+
+--------+=========| |================
| S2 | +--------+
| | | |
+--------+ | S3 |
| |
+--------+
When focus is on S0, 'focus output right' will first match S1 (the
closest output which overlaps in the highlighted area), then S2, but not
S3 (since S3 does not overlap into the highlighted area).
fixes#669fixes#771
While it’s certainly better and clearer to specify it, we should do the
right thing when the unit is missing, just like CSS for example
(margin: 0; is okay, margin: 0px; too).
If there is a client with an urgency hint on another workspace and
switching to this workspace would cause the urgency to be reset (by
moving the focusing to the client), delay the reset by some time. This
gives the user the chance to see it.
This commit adds the possibility to configure the urgency delay timer
duration using the 'force_display_urgency_hint' directive. Also,
documentation and a testcase was added to allow for automated checks of
the intended behavior.
fixes#482
this implements both the "move container to workspace back_and_forth" command
and movements to the same workspace when auto_back_and_forth is set.
it includes documentation and test suite additions by michael.
it also simplifies the workspace_show_by_name function (making use of
workspace_get accepting NULL pointers).
Introducing a new event to subscribe called mode. It's fired up
when i3 changes binding mode (like switching from default to resize).
IPC guide adjusted also.
Functions such as fprintf() might be unsafe to use in a signal handler,
see http://stackoverflow.com/questions/3941271/#answer-3941563
By using ev_signal, libev will use a tiny signal handler which just
passes on the information and then calls (outside of the signal handler)
our callback function which can use fprintf() and other unsafe
functions.
fixes#803
Before commit 4976fa3350, setting the
layout of workspaces to something else than the default would just mess
up the parent container of the workspace (the content container).
After that commit, it would create an unnecessary split container when
you change the layout _before_ opening any containers. To avoid this, we
now store the layout (similar to how the 'workspace_layout'
configuration directive works) and apply it when the first container is
attached to the workspace.
Fixes#796
If the focused application is in fullscreen mode and 'scratchpad show'
is executed, toggle the fullscreen mode before showing the scratchpad
application. Otherwise, the scratchpad app would not be visible anyways
but focus would be stolen.
fixes#613
The problem was that scratchpad_move() didn’t check whether the source
workspace was focused. Therefore, 'move scratchpad' only worked reliably
interactively , but not when used with criteria.
For the following binding:
# Simulate ctrl+v upon pressing $mod+x
bindsym --release $mod+x exec --no-startup-id xdotool key --clearmodifiers ctrl+v
you can now use either:
1. press $mod, press x, release x, release $mod
2. press $mod, press x, release $mod, release x
fixes#485
The implementation is naive because the user has to generate exactly the
event he specified. That is, if you use this binding:
bindsym --release $mod+x exec import /tmp/latest-screenshot.png
Then it will only be triggered if you hit $mod, hit x, release x,
release $mod. It will not be triggered if you hit $mod, hit x, release
$mod, release x. The reason is that the KeyRelease event in the latter
case will not have the modifier in its flags, so it doesn’t match the
configured binding.
In tabbed mode, the available width (say 1280) is divided by the amount
of child containers (say 3). Before this commit, we just truncated the
result and would end up with 426 + 426 + 426 = 1278 pixels that we
render to. Now we render a bit too much, but that’ll at least not give
us graphics corruption on any side :).
fixes#791
When the workspace layout (formerly orientation) was forced to change
due to a move command, the split container we created was not marked as
split = true, which caused tree_flatten() to errnously kill the contents
of it and thus one window ended up unmanaged.
Also, the logic in tree_flatten() was inverted due to commit de94f6da.
fixes#790
This changes the fact that Firefox would not be launched on the correct
workspace because it marked the startup sequence as completed *before*
actually mapping all of its windows.
To test this, go to workspace 3 and run this command in a terminal:
i3-msg 'exec iceweasel; workspace 4'
That will make i3 start iceweasel (and create a proper startup
notification context for it), then immediately switch to workspace 4
(before iceweasel could possibly start).
The iceweasel window(s) should appear on workspace 3.
Basically, this is the same fix as commit 914ca6cf :-/. Once again, we
called exit() instead of _exit(), but this time it lead to a kill(0,
SIGTERM), effectively killing all processes in the i3 process group,
including i3 itself. The cause for the kill(0) is that nagbar_pid is set
to 0 by fork(), signaling we’re in the child process. The cleanup
handler only checks for nagbar_pid being -1 as a special value, however.
Previously, in case 'layout stacked' (for example) had been called
interactively, con_set_layout would be called with focused->parent,
while with for_window, it’d be called on the actual matching container.
This difference in behavior was the cause for the inability to use
'for_window [class="XTerm"] layout tabbed', which now works \o/, but
more on that below.
The change also allows us to handle the case of the user selecting a
CT_WORKSPACE container properly, that is, by using the special case and
creating a new split container on the workspace which gets all the
contents, but a new layout.
Now, before you are enthusiastic about the change and try to use
for_window magic in your config file, keep in mind: The 'layout' command
acts on the parent split container. That is, when using a line such as
this one:
for_window [class="XTerm"] layout tabbed
…and opening an XTerm when on a workspace with one single other window,
the whole workspace will be set tabbed (just as previously when you
opened an XTerm and sent 'layout tabbed' manually).
Therefore, to open XTerm in its own tabbed split container, you need to
split before:
for_window [class="XTerm"] split v, layout tabbed
The comma here is important! It says that the second command should not
be treated as an entirely unrelated command, but it should also relate
the matching window (while it does work with a ';', that is prone to
race-conditions and should be avoided).
fixes#358
When you have a tabbed container which has had more than one container
but currently has precisely one container, there was a bit of flickering
when switching workspaces occasionally. This commit fixes it by properly
setting the height of the deco_rect (and thus the X11 window) to not
make the old window contents show up for a minimum period of time.
fixes#777
This fixes a compatibility issue with gnome-terminal and xfce’s
terminal, where fullscreening would lead to moving the window and not
displaying the contents properly.
fixes#788
When resizing floating windows, changing the height was not correctly
handled. This commit fixes that and adds testcases for shrinking and
growing the width and height of floating windows.
To automagically do the right thing when rotating monitors with regards
to splith/splitv layout (depending on width/height of the monitor), we
change the orientation of existing workspaces and the first child.
If that first child happens to be a stacked/tabbed con, we cannot change
the layout unconditionally (previously, the orientation was not in the
layout, so we never noticed this problem).
fixes#768
The latter is actually wrong. For example, when running i3
--moreversion, it will print $(pwd)/i3 instead of $(which i3). In my
previous tests, this coincidentally was the same.
This changes the SHM log format, it doesn’t use 0-bytes to separate
entries anymore. Instead of using lots of printf() calls in i3-dump-log,
we now do precisely one big write().
So, to be clear: i3-dump-log and i3 both need to be upgraded.
Mismatching versions will lead to garbage output (no crashes of i3, just
garbage output).
The -f flag uses an inter-process pthread_cond_t in the shared memory
header to broadcast the arrival of new messages to all i3-dump-log
processes. This internally uses futexes and thus doesn’t even mean a
kernel call in most cases. inter-process pthread_cond_ts require NPTL
(the Native Posix Thread Library, introduce in Linux 2.6).
From the source:
When starting i3 initially (and after each change to the connected
outputs), this function fixes the resolution of the __i3
pseudo-output. When that resolution is not set to a function which
shares a common divisor with every active output’s resolution,
floating point calculation errors will lead to the scratchpad window
moving when shown repeatedly.
fixes#632
Since we advertise _NET_ACTIVE_WINDOW support (but only set the
corresponding atom currently), it makes sense to also support the
ClientMessage. Apps such as Gajim use it to set focus to the roster
window when clicking on the tray icon for example.
fixes#767
Compilers store the path with which they were called in the debug
symbols. Therefore, this will make backtraces show something like
../i3-4.2/src/main.c instead of src/main.c.
See also http://stackoverflow.com/questions/6473561/
before:
$ time CC=clang make -j16
CC=clang make -j16 6,04s user 0,86s system 468% cpu 1,471 total
CC=clang make -j16 6,05s user 0,87s system 468% cpu 1,477 total
CC=clang make -j16 6,15s user 0,86s system 464% cpu 1,510 total
CC=clang make -j16 6,05s user 0,93s system 467% cpu 1,493 total
CC=clang make -j16 6,10s user 0,84s system 461% cpu 1,507 total
$ time CC=gcc make -j16
CC=gcc make -j16 9,91s user 1,43s system 508% cpu 2,231 total
CC=gcc make -j16 10,02s user 1,37s system 500% cpu 2,275 total
CC=gcc make -j16 9,80s user 1,60s system 507% cpu 2,245 total
CC=gcc make -j16 10,02s user 1,44s system 506% cpu 2,264 total
CC=gcc make -j16 9,99s user 1,45s system 505% cpu 2,261 total
after:
$ time CC=clang make -j16
CC=clang make -j16 3,41s user 0,83s system 375% cpu 1,131 total
CC=clang make -j16 3,29s user 0,90s system 373% cpu 1,122 total
CC=clang make -j16 3,35s user 0,77s system 369% cpu 1,116 total
CC=clang make -j16 3,36s user 0,78s system 374% cpu 1,105 total
CC=clang make -j16 3,46s user 0,75s system 373% cpu 1,126 total
$ time CC=gcc make -j16
CC=gcc make -j16 10,74s user 1,44s system 494% cpu 2,462 total
CC=gcc make -j16 10,68s user 1,54s system 497% cpu 2,453 total
CC=gcc make -j16 10,60s user 1,60s system 488% cpu 2,499 total
CC=gcc make -j16 10,63s user 1,51s system 485% cpu 2,502 total
CC=gcc make -j16 10,70s user 1,51s system 497% cpu 2,453 total
Therefore, we enable pre-compiled headers only when CC=clang.
This is useful for third-party scripts which require certain features
and want to error out cleanly when they are run with an old i3 version.
Additionally, i3 --version might be different from what’s actually
running (an old version of the binary), so i3-msg -t get_version will be
the best way to figure out the i3 version you are actually running from
this commit on.
With this commit, the "default" layout is replaced by the splith and
splitv layouts. splith is equivalent to default with orientation
horizontal and splitv is equivalent to default with orientation
vertical.
The "split h" and "split v" commands continue to work as before, they
split the current container and you will end up in a split container
with layout splith (after "split h") or splitv (after "split v").
To change a splith container into a splitv container, use either "layout
splitv" or "layout toggle split". The latter command is used in the
default config as mod+l (previously "layout default"). In case you have
"layout default" in your config file, it is recommended to just replace
it by "layout toggle split", which will work as "layout default" did
before when pressing it once, but toggle between horizontal/vertical
when pressing it repeatedly.
The rationale behind this commit is that it’s cleaner to have all
parameters that influence how windows are rendered in the layout itself
rather than having a special parameter in combination with only one
layout. This enables us to change existing split containers in all cases
without breaking existing features (see ticket #464). Also, users should
feel more confident about whether they are actually splitting or just
changing an existing split container now.
As a nice side-effect, this commit brings back the "layout toggle"
feature we once had in i3 version 3 (see the userguide).
AFAIK, it is safe to use in-place restart to upgrade into versions
after this commit (switching to an older version will break your layout,
though).
Fixes#464
e.g. pressing Mod1+x when having the following in your configfile:
bindsym Mod1+x some invalid command
will lead to an i3-nagbar instance popping up, offering you to view the
error log (which will contain parser errors from this commit on).
This workaround is necessary for terminal emulators which parse -e in a
different way: some accept a list of arguments (-e command arg1 arg2 …),
some accept only one argument (-e "command arg1 arg2 …"). Therefore, we
just create a script and pass that as the one and only argument.
CPPFLGES, CFLAGS and LDFLAGS should be user variables
We now provide default flags but use I3_*FLAGS flags for our own needed
flags
Also reoder lib flags a bit
In certain situations (when you have a h-split within a h-split) you
couldn’t properly resize previously. This commit makes the resize
command properly traverse up the containers.
fixes#754
This fixes a race condition when running the tests. I think that the X11
server has more time to clean up the resources when we do an explicit
disconnect. The symptom I was seeing was that sometimes, i3 couldn’t
become the window manager on one of the Xdummy instances.
Depending on the memory layout, it could happen that bind->command was
exchanged with something else while the parser still accessed it.
Therefore, we now copy the command and let the parser use that copy.
When moving window from other (not current) workspace to another
workspace with criteria we should stay on current workspace.
And we should exit early when criteria was specified but didn't
match any window.
This is now restricted according to the already defined fullscreen
focus constraints. Test case 157 was removed, as we don't prevent
level up/down in fullscreen anymore. Those commands are properly
tested in fullscreen by test case 156.
Fixes: #612
Basically, a focus change can't escape a fullscreen container. The
only exception is per-output fullscreen containers, as you should
be able to focus a container in a different workspace in this case.
This is an improvement on 4eab046e, now considering the difference
between global and per-output fullscreen and taking the tree
structure into account to determine what escaping the fullscreen
container means. It only affects targeted focus commands in the
form "for_window [...] focus", but it lays the foundation for
forthcoming fixes to all other focus commands.
If the target is in a different workspace, there's no reason why
we wouldn't allow the user to focus it. We already allow this when
focusing a workspace, for example.
calling workspace by number now also checks for switching back and forth
and creates a new workspace if no workspace starting with that number is
found
also removed the obsolete tree_render() in favor of setting
cmd_output->needs_tree_render to true
When calculating coordinates we should multiply at first otherwise
we lose precision when i3 is compiled without sse2 support.
The following code prints "Res1: 348 Res2: 349" when compiled with
-O0 -mno-sse2 and "Res1: 349 Res2: 349" with -O0 -msee2.
Note that -msse2 is default flag on 64bit OSes.
int main() {
double a = 349.0 / 768;
double b = 349.0 * 768;
int res1 = a * 768;
int res2 = b / 768;
printf("Res1: %d Res2: %d\n", res1, res2);
return 0;
}
Thanks guys for helping me to hunt down this one.
See also:
http://article.gmane.org/gmane.linux.kernel/1268792
The C compiler will handle (void) as "no arguments" and () as "variadic
function" (equivalent to (...)) which might lead to subtle errors, such
as the one which was fixed with commit 0ea64ae4.
This fixes focus problems with Eclipse. Apparently, Eclipse waits for getting
notified about the focus, and since it used non-managed windows, i3 didn’t care
to update the focus.
Fixes: #621, #675
Fixes: #668
Calling tree_close with dont_kill_parent=true will avoid it from closing the
workspace if it’s empty (and it’s temporarily empty, because 'floating disable'
detaches, then re-attaches the window).
Initially I thought using the second precision time() function is good enough,
but to make t/113-urgent.t considerably faster (>2s vs. 0.08s), we put in a
little more effort and use gettimeofday. Otherwise, this test blocks the whole
testsuite from completing much faster on modern machines :).
This change has two implications:
1) tree_render() will now be called precisely once for input which consists of
multiple commands (like "focus left; focus right"). Also, the caller of
parse_command() has to call it. This makes us able to fix tickets such as
ticket #608 (where multiple tree_render() calls are noticable).
2) The output of a command is now a JSON array of return values of the
individual subcommands. In the case of "focus left; focus right", this is:
[{"success":true}, {"success":true}]
While this is incompatible with what i3 returned before, the return value of
commands was undocumented and therefore not subject to our API stability.
You need to specify the --enable-32bit-visual flag when starting i3. This is
done because everything feels sluggish on my system when using a 32 bit visual
instead of a 24 bit visual. Fast > fancy.
Fixes floating containers seemingly showing up in the wrong
workspace after moving workspaces containing floating containers.
We must *always* fix the coordinates of floating containers when
moving workspaces across outputs. That's because the coordinates
of floating containers are *not* relative to the workspaces.
Currently it supports the following options:
"oldest": match the first window that triggered an urgent event
"latest": match the last window that triggered an urgent event
If you had workspace 1, 2, 3, 4 on LVDS1 and you enabled HDMI2 (where workspace
1 to workspace 5 are assigned to HDMI2), i3 would look for a new workspace for
LVDS1 (since all workspaces were moved), create workspace 5, move that over due
to assignment and then create workspace 6. Effectively, you would end up with
an empty workspace 5.
Fixes a bug where splitting then moving in the other orientation
(e.g. v-splitting and moving right) would result in the old
indicators not disappearing.
This re-introduces borders around the workspace buttons in i3bar.
No additional pixels will be consumed (you will not lose any space for your
windows).
This fixes problems with the Oracle JRE7, which checks the current focus after
receiving WM_TAKE_FOCUS and just does nothing when the focus is on one of its
windows. Hopefully it doesn’t introduce any regressions :).
On the rationale of using a custom parser instead of a lex/yacc one, see this
quote from src/commands_parser.c:
We use a hand-written parser instead of lex/yacc because our commands are
easy for humans, not for computers. Thus, it’s quite hard to specify a
context-free grammar for the commands. A PEG grammar would be easier, but
there’s downsides to every PEG parser generator I have come accross so far.
This parser is basically a state machine which looks for literals or strings
and can push either on a stack. After identifying a literal or string, it
will either transition to the current state, to a different state, or call a
function (like cmd_move()).
Special care has been taken that error messages are useful and the code is
well testable (when compiled with -DTEST_PARSER it will output to stdout
instead of actually calling any function).
During the migration phase (I plan to completely switch to this parser before
4.2 will be released), the new parser will parse every command you send to
i3 and save the resulting call stack. Then, the old parser will parse your
input and actually execute the commands. Afterwards, both call stacks will be
compared and any differences will be logged.
The new parser works with 100% of the test suite and produces identical call
stacks.
The problem was that the workspace was considered empty for a brief period of
time when entering floating mode. This happened when you assigned Gimp to a
workspace which is not in use yet.
With this commit, i3 will now use either $XDG_RUNTIME_DIR/i3 (XDG_RUNTIME_DIR
is only writable by the user, so this is not a problem) or a secure temporary
location in /tmp, following the pattern /tmp/i3-<user>.XXXXXX
Fixes: #585
This commit fixes the problem of i3 wrongly grabbing/interpreting (!) some key
bindings. Basically, when you have, say, "bindsym Mod1+4 workspace 4", but you
also have "bindsym Mod1+semicolon focus right" (both are default), and your
keyboard layout has semicolon on Mode_switch + 4, the "workspace 4" keybinding
was shadowed by the "focus right" keybinding, because that also resolves to
semicolon.
So, from now on, i3 will only consider column 0 and 1 for normal bindings and
column 2 and 3 for bindings using Mode_switch (columns as seen in xmodmap
-pke).
After a reload, the drawing parameters for the decorations might
have changed, so we need to invalidate the cache and force a redraw
of the currently visible decorations. Also, don't leak the previous
font when reloading by freeing it before parsing the config.
Abstracted draw_text and predict_text_width into libi3. Use
predict_text_width from libi3 in i3 too. This required tracking
xcb_connection in a xcb_connection_t *conn variable that libi3
expects to be available in i3bar.
Also prints out useful stuff:
CORE DUMPS: You are running a development version of i3, so coredumps were
automatically enabled (ulimit -c unlimited).
CORE DUMPS: Your current working directory is "/home/michael/i3".
CORE DUMPS: Your core_pattern is: /tmp/%e.core.%p
i3 (tree) version 4.0.2-479-g26ab2ac (2011-11-08, branch "next") starting
This does not affect child processes of i3.
The intention of this change is to make debugging easier – it’s one less thing
users of the development version have to worry about when trying to help with
debugging.
Fixes#533
The problem was that the code was always executed. While it *attaches* the new
container to the workspace container, it also sets current = NULL and thus
always appends the container instead of inserting it after the currently
focused child. So now, we just don’t execute that code at all for
workspace_layout == default.
Following bug:
1) Assign workspace 9 to output HDMI2
2) On HDMI2, be on workspace 1
3) Focus a different output, say LVDS1
4) Execute i3 'workspace 9'
5) Something happens, but you end up back on ws 1
(this is due to an EnterNotify being generated when warping)
This means you can now specify an amount of pixels for the move command. The
default is 10. Note that this of course only works for floating windows.
Example:
move left 20 px
Previously, we'd only close an empty workspace when we moved away
from it. Now we also close it when the last client exits, as long
as that workspace is not visible anymore.
Also, the API changed a bit. There are two functions now, both assume you
already got the keysyms (which is the case for i3 and i3-config-wizard),
one gets the modifier mapping for you (aio_get_mod_mask_for) while the other
assumes you also got that. No roundtrips are required for the latter.
i3bar previously used get_colorpixel on strings without the leading # (ff0000
instead of #ff0000). Since it uses libi3’s get_colorpixel now we needed to
update a few places.
Some of them are useless nowadays, others very unlikely to be a problem.
Those which might still be interesting somewhen in the future are just
commented out.
This fixes a problem with workspace assignments.
I tested i3 -C with three user configs (Thanks SardemFF7, julien, xeen)
and did not notice any problems.
localtime_r does not have the side-effect of behaving like it called tzset(),
in particular it will save one stat(/etc/localtime) syscall. This is not a big
deal, but it makes the strace output cleaner and thus more useful :).
Comment in the code makes the fix clear, I think.
Steps to reproduce the issue:
1) On the right monitor, open two terminals (h-split workspace)
2) 'i3 mark foo' on the right terminal
3) Focus the left terminal
4) Switch to the left monitor
5) i3 '[con_mark="foo"] focus'
6) the left window will contain the pointer (and is focused for a short period
of time, or even permanently due to race conditions)
From i3 --help:
If you pass plain text arguments, i3 will interpret them as a command
to send to a currently running i3 (like i3-msg). This allows you to
use nice and logical commands, such as:
i3 border none
i3 floating toggle
i3 kill window
In order to not depend on X11 just for getting the socket paths, scripts or
other programs can now use i3 --get-socketpath. Since i3 must be present on the
computer anyways, this saves one dependency :).
The former two provide fallbacks in case $PAGER or $EDITOR is not set (which
might be more common than you think, because they have to be set in
~/.xsession, not in the shell configuration!) while the latter tries to launch
a terminal emulator. The scripts are most prominently used in i3-nagbar, which
alerts the user when the configuration is broken for some reason. Also,
i3-sensible-terminal is used in the default configuration.
This commit does not rely on the shell supporting ${PAGER:-less} anymore, which
is not the case for 'fish'.
This is mainly useful for the testsuite. The tests can wait until i3 processed
all X11 events and then continue. This eliminates sleep() calls which leads to
a more robust and faster testsuite.
The configuration option does the same as the commandline parameter, except
it can be easily set by the user (e.g. you are using KDM and can't start a
session through ~/.xsession).
Signed-off-by: Michael Walle <michael@walle.cc>
This commit makes the coordinates proportional when moving floating windows.
That is, if you have a window at the bottom of your 1920 px monitor and move it
to your 800 px monitor, it will be at the bottom of the 800 px monitor (and not
out of bounds).
My testcase was putting a floating window on the left output, but overlapping a
little to the right output. Then switch to a workspace on the right output.
From the source:
We need ev >= 4 for the following code. Since it is not *that* important
(it only makes sure that there are no i3-nagbar instances left behind) we
still support old systems with libev 3.
Instead of using a quoted string to specify the class / title, the assign
command now uses criteria, just like the for_window command or the command
scopes.
An example comes here:
# Assign all Chromium windows (including popups) to workspace 1: www
assign [class="^Chromium$"] → 1: www
# Make the main browser window borderless
for_window [class="^Chromium$" title=" - Chromium$"] border none
This gives you more control over the matching process due to various reasons:
1) Criteria work case-sensitive by default. Use the (?i) option if you want a
case-insensitive match, like this:
assign [class="(?i)^ChroMIUM$"] → 1
2) class and instance of WM_CLASS can now be matched separately. For example,
when starting urxvt -name irssi, xprop will report this:
WM_CLASS(STRING) = "irssi", "URxvt"
The first part of this is the instance ("irssi"), the second part is the
class ("URxvt").
An appropriate assignment looks like this:
assign [class="^URxvt$" instance="irssi"] → 2
3) You can now freely use a forward slash (/) in all strings since that is no
longer used to separate class from title (in-band signaling is bad, mhkay?).
This should fix focus problems where EnterNotify events were generated while
moving windows. Hopefully it does not trigger any regressions :).
Fixes#500
Fixes#452
This might also fix problems with Firefox where Firefox only displays a white
window (my guess is that it mapped a floating window outside of the actually
visible area for some reason).
- Introduce warp_to static variable in x.c that stores the coordinates
to warp to as a Rect.
- Add x_set_warp_to function to set this variable. Use in _tree_next,
workspace_show, and con_move_to_workspace.
- In x_push_chanages, if warp_to is set, then call xcb_warp_pointer_rect
and then reset it to NULL.
This fixes all know bugs for pointer warping for me.
1. Fix focus when moving to same workspace.
If we have a single window on a workspace and we switch to the same
worksapce, focus_next will be the workspace container, rather than the
current window, so simply call con_descend_focused to ensure we set the
focus to a window.
2. Fix focus when moving a container to a visible workspace.
Call workspace_show before we attaching to new visible workspace, so we
don't get in the weird situation where target workspace has focused
window, but it isn't considered focused.
It now uses the container orientation (if it is appropriate, the last focused
one otherwise) to recurse. This works better if the target workspace is in
vertical orientation when you use right/left or if it is in horizontal
orientation but you use up/down.
Modify _tree_next() so that when we reach the workspace container:
1. Find the next corresponding output (screen) using the added
get_output_next().
2. If there is another output, find the visible workspace.
3. Call workspace_show on found workspace.
4. Find the appropriate window to focus (leftmost/rightmost, etc.) using
con_descend_direction, and then focus it.
I've only tested on horizontal monitors (left/right).
Generally, the traversal goes: numbered workspaces in order, and then
named workspaces in the order in which they appear in the tree.
Example:
Output 1: Output 2:
1 3 D C 2 4 B A
Traversal: 1, 2, 3, 4, D, C, B, A, 1, ...
Note, after the numbered workspaces, we traverse the named workspaces
from output 1, and then output 2, etc.
This fixes a race where we created cursors on the Xlib connection, flushed,
then used the cursor on the XCB connection. Even though we flushed, the X
server did not process the requests yet and therefore returned a BadCursor
error.
This bugfix uses the Xlib connection for setting the root window cursor which
will ensure that the requests are properly serialized.
An easy test for this (on my machine) is the following ~/.xsession:
xsetroot -cursor_name cross
exec i3
If you see a cross cursor instead of the pointer, the race happens. You’ll see
a error_code=6 error in your ~/.xsession-errors.
Runs the same code as for non-stacked / non-tabbed containers, but on the
parent of the container (which is the stacked / tabbed con, since route_click
gets called with the child).
These errors can happen because a DestroyWindow request by a client will
trigger an UnmapNotify, then a DestroyNotify. We cannot distinguish this
UnmapNotify from an UnmapNotify not followed by a DestroyNotify, so we just try
to send the ReparentWindow / ChangeProperty and ignore the errors, if any.
Think of the following layout:
-------------
| tab | |
| con | win |
| | |
-------------
The tabbed container on the left has two children. Assume you have focused the
second/right child in the tabbed container. i3 used to focus the first/left
container of the tabbed container when using 'focus right' (it wrapped focus).
With this commit, the default behaviour is to instead focus the window on the
right of the screen.
The intention is to make focus switching more intuitive, especially with tabbed
containers supporting 'focus left'/'focus right' in tree. You should end up
using less 'focus parent' :).
You can force the old behaviour with 'force_focus_wrapping true' in your
config.
Code coverage is 62.5% with this commit.
An example to set all XTerms floating:
for_window [class="XTerm"] mode floating
To make all urxvts use a 1-pixel border:
for_window [class="urxvt"] border 1pixel
A less useful, but rather funny example:
for_window [title="x200: ~/work"] mode floating
The commands are not completely arbitrary. The commands above were tested,
others may need some fixing. Internally, windows are compared against your
criteria (class, title, …) when they are initially managed and whenever one of
the relevant values change. Then, the specified command is run *once* (per
window). It gets prefixed with a criteria to make it match only the specific
window that triggered it. So, if you configure "mode floating", i3 runs
something like '[id="8393923"] mode floating'.
New containers were previously attached directly to the workspace instead of to
the previously focused place in the workspace (for example a stacked con).
Fixes: #376
Use 'kill window' to kill a specific window (for example only one specific
popup), use 'kill client' to kill the whole application (or X11 connection to
be specific).
When a tabbed container had more than one child and at least the first one
supported WM_DELETE, i3 entered an endless loop when killing that tabbed
container. This was due to tree_close only sending WM_DELETE without actually
removing the child, while the loop in tree_close assumed that with every call
of tree_close one child would be removed.
Actually, commit 1c5adc6c35 commented out code
without ever fixing it. I think this was responsible for the 'workspace
switching sometimes does not work' bug. My observations:
Had it again today and analyzed a log of it. Looks like after unmapping the
windows on one workspace (in my case: chromium, eclipse, urxvt, focus on
eclipse) we get UnmapNotify events for chromium and eclipse, but then we get an
EnterNotify for the terminal (due to unmapping the other windows and therefore
mapping the terminal under the cursor), only afterwards the UnmapNotify
follows.
So, there are two things wrong with that:
• We handle EnterNotifys for unmapped windows
• Unmapping windows sometimes works in a sequence, sometimes the sequence gets
split. Not sure why (if unmapping can take longer for some windows or if our
syncing is wrong -- but i checked the latter briefly and it looks correct).
Maybe GrabServer helps?
• We don’t ignore EnterNotify events caused by UnmapNotifies. We used to, but
then there was a different problem and we decided to solve the EnterNotify
problem in another way, which actually never happened (commit
1c5adc6c35).
According to the Xlib Programming Manual section 10.7.2 [1], these events are
generated when keyboard grabs activate/deactivate, while we are only interested
in focus changes which are done by other programs independend from the
keyboard.
[1] http://tronche.com/gui/x/xlib/events/input-focus/grab.html
This involves:
• Compiling with xcb-util instead of xcb-{atom,aux} (they merged the libraries)
• Not using xcb-{event,property} anymore (code removed upstream)
• Not using the predefined WINDOW, CARDINEL, … atoms (removed upstream)
• Using the new xcb_icccm_* data types/functions instead of just xcb_*
(for example xcb_icccm_get_wm_hints instead of xcb_get_wm_hints)
Also I refactored the atoms to use x-macros.
The Clang Static Analyzer uncovered those issues:
- The variable "changed" in handlers.c is written to, but it's
never read since that specific write, so the write is not
necessary.
- In util.c, "tail" may be NULL. In that case, we shouldn't pass
it to strlen because strlen's behavior is not defined when s is
NULL.
- In util.c, "write_index" is incremented twice. It's never used
anymore after being incremented once, so the second increment is
not necessary.
Due to lots of cases which were added and added to tree_move(), the function
was not really easy to understand. For this refactoring, I wrote tree_move()
from scratch, thinking about (hopefully) all cases. The testsuite still passes.
The move command also has different parameters now. Instead of the hard to
understand 'before v' stuff, we use 'move [left|right|up|down]'.
If we're not killing the mapped window and we're not killing the
parent window either in tree_close, then there's no reason to try
to change the focus. This fixes focus issues when moving a
container around another container (move up, left, bottom, right).
The case of an X11 server having multiple displays is handled correctly by the
code in src/mainx.c. However, due to some functions not being correctly
refactored and still getting the first screen (and also the first root window)
from the XCB connection, i3 was operating on the wrong root window.
Deiz