The testsuite already contains quite a number of SIGCHLD handler
installation/un-installations. Here is my attempt at an inventary.
1. complete-run.pl installs a SIGCHLD handler in the `start_xserver()` function
call, which prints an error and exits when all x server processes have exited.
2. In the TestWorker child process, a SIGCHLD handler is installed to reap dead
test child processes.
3. The TestWorker child process forks another child process for running the test
file, where the previously installed SIGCHLD handler (point 2) is unset.
This is where this commit comes in: it installs a SIGCHLD handler in the test
file child process, which will trigger when the i3 subprocess dies.
4. (For completeness: i3test.pm defines an END block where it unsets the
previous SIGCHLD handler before it kills the subprocesses.)
With this commit, when i3 segfaults, the output will look like this:
Writing logfile to 'testsuite-2022-09-10-21-14-46-4.20-103-gb242bceb/complete-run.log'...
[:100] /home/michael/i3/testcases/t/167-workspace_layout.t: BAILOUT
completed 0 of 1 tests
test /home/michael/i3/testcases/t/167-workspace_layout.t bailed out:
could not kill i3: No such process
fixes https://github.com/i3/i3/issues/4437
This commit should fix "move con to parent" trick (see below) in the
case when con->parent->parent is a workspace.
The trick:
mark _a, focus parent, focus parent, mark _b,
[con_mark=_a] move window to mark _b, [con_mark=_a] focus
The trick got broken in commit 626af81232e6ca81abea22267f3330c16d804596
in order to fix an i3 crash (#2003). Reverting said commit fixes the
trick. The crash is caused by the fact that empty workspace isn't
considered a split (checked in src/con.c:1324), so the moved window ends
up as a sibling of the target workspace, not as its child.
_NET_WM_STATE_FOCUSED is set on _NET_WM_STATE to indicate that the
window is focused. It must be set when the window is newly focused and
removed once the window no longer has focus.
> _NET_WM_STATE_FOCUSED indicates whether the window's decorations are
> drawn in an active state. Clients MUST regard it as a read-only hint.
> It cannot be set at map time or changed via a _NET_WM_STATE client
> message.
For example, this is used by GTK applications to show the decoration in
an active or inactive state. This change can be tested by opening a GTK
application (like evince), focusing the window and unfocusing the
window, and observing a change in the window decorations.
Fixes#2273
Issue #3049 describes a case where terminating i3 by means of SIGTERM
causes it to leak the runtime directory and all its contents. There are
multiple issues at play: first, any cleanup handlers registered via
atexit are never invoked when a signal terminates the program (see
atexit(3)). Hence, the log SHM log cleanup performed in i3_exit is not
invoked in that case. Second, compared to the shutdown path for the
'exit' command, we do not unlink the UNIX domain socket we create,
causing it to be leaked as well. Third, a handler for SIGTERM is not
registered at all despite handle_signal claiming to be the handler for
all 'Term' signals.
This change addresses all three problems and results in a graceful exit
including cleanup to happen when we receive a signal with the default
action 'Term'. It addresses issue #3049.
We need to set dont_map => 1 on the sync window to prevent an endless loop.
Further, t/219-ipc-window-focus.t made assumptions about windows being named
incrementally, and that assumption is broken by the sync window opened by the
first sync_with_i3 call from open_window, so use the more reliable ->name.
This way you can assign the test windows to an empty workspace to avoid
interacting with them (when xvfb-run is not an option):
assign [instance="i3test"] workspace testing
I previously tried to fix the check, but could only come up with a fix which
required removing our module pre-loading, which makes the tests considerably
more expensive. Instead, let’s just remove the check.
This tool is similar to xtrace in usage in that it intercepts traffic to
the X server. The motivating feature for writing the tool is its ability
to inject prepared reply messages instead of the server’s reply. In
this particular case, we’ll inject a RRGetMonitors reply to test i3’s
RandR 1.5 code paths.
The added testcase is a noop for now, but with the code that’s lingering
in the randr15 branch, i3 does actually detect monitors as per the
injected reply:
2016-11-20 21:10:05 - randr.c:__randr_query_outputs:618 -
RandR 1.5 available, querying monitors
2016-11-20 21:10:05 - randr.c:__randr_query_outputs:628 -
1 RandR monitors found (timestamp 0)
2016-11-20 21:10:05 - randr.c:__randr_query_outputs:646 -
name DP3, x 0, y 0, width 3840 px, height 2160 px, width 520 mm,
height 290 mm, primary 1, automatic 1
This is preparation work for issue #1799
We add $HOME to the environment variables we define for a test case
in order to redirect it from the user's actual home directory. This
is necessary because xcb-util-xrm will fall back to $HOME/.Xresources
when determining the DPI. If a user has this set to, e.g., 192 on their
machine, this would break tests.
Since tests shouldn't rely on the system they run in, we redirect the
home directory altogether to simulate a clean slate.
relates to #2465
Commit 287a0b4 introduced a segfault when validating the i3 config
as the root_screen will not be set in this case, causing a null
pointer dereference.
fixes#2144
Replace the XDummy script with Xephyr. This is done because of some
changes in the Xorg server that make XDummy difficult to use.
Rename library internal variables and function names to replace "xdummy"
with "xserver" to show this change (except for renaming the package and
lib file for better git history).
Rename the switch `--keep-xdummy-output` to `--keep-xserver-output`.
This switch should now be rarely used because Xephyr requires less set
up.
Replace "xdummy" with "xephyr" in comments and utility help
information. Update docs to show the new dependency.
fixes#1367
Since i3 honors the “Globally Active Input” focus model, we need to
explicitly state that we are not using that in our testcases :).
This requires X11::XCB from git to work (commit
71b25dcaafc509e710b8fd7de20c97ac3549fc39).
The suppression file makes valgrind output more readable by hiding
reports of memory leaks for GObject-related initialization functions in
Pango and Cairo.
