gri3-wm/src/main.c

966 lines
38 KiB
C
Raw Normal View History

/*
* vim:ts=4:sw=4:expandtab
*
* i3 - an improved dynamic tiling window manager
* © 2009 Michael Stapelberg and contributors (see also: LICENSE)
*
* main.c: Initialization, main loop
*
*/
#include "all.h"
#include <ev.h>
2010-11-27 00:26:51 +01:00
#include <fcntl.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <libgen.h>
2013-06-05 14:59:05 +02:00
#include "shmlog.h"
#ifdef I3_ASAN_ENABLED
#include <sanitizer/lsan_interface.h>
#endif
#include "sd-daemon.h"
/* The original value of RLIMIT_CORE when i3 was started. We need to restore
* this before starting any other process, since we set RLIMIT_CORE to
* RLIM_INFINITY for i3 debugging versions. */
struct rlimit original_rlimit_core;
/** The number of file descriptors passed via socket activation. */
int listen_fds;
/* We keep the xcb_prepare watcher around to be able to enable and disable it
* temporarily for drag_pointer(). */
static struct ev_prepare *xcb_prepare;
extern Con *focused;
char **start_argv;
xcb_connection_t *conn;
/* The screen (0 when you are using DISPLAY=:0) of the connection 'conn' */
int conn_screen;
/* Display handle for libstartup-notification */
SnDisplay *sndisplay;
/* The last timestamp we got from X11 (timestamps are included in some events
* and are used for some things, like determining a unique ID in startup
* notification). */
xcb_timestamp_t last_timestamp = XCB_CURRENT_TIME;
xcb_screen_t *root_screen;
xcb_window_t root;
/* Color depth, visual id and colormap to use when creating windows and
* pixmaps. Will use 32 bit depth and an appropriate visual, if available,
* otherwise the root windows default (usually 24 bit TrueColor). */
uint8_t root_depth;
xcb_visualtype_t *visual_type;
xcb_colormap_t colormap;
struct ev_loop *main_loop;
xcb_key_symbols_t *keysyms;
2013-06-05 14:59:05 +02:00
/* Default shmlog size if not set by user. */
const int default_shmlog_size = 25 * 1024 * 1024;
/* The list of key bindings */
struct bindings_head *bindings;
/* The list of exec-lines */
struct autostarts_head autostarts = TAILQ_HEAD_INITIALIZER(autostarts);
/* The list of exec_always lines */
struct autostarts_always_head autostarts_always = TAILQ_HEAD_INITIALIZER(autostarts_always);
/* The list of assignments */
struct assignments_head assignments = TAILQ_HEAD_INITIALIZER(assignments);
/* The list of workspace assignments (which workspace should end up on which
* output) */
struct ws_assignments_head ws_assignments = TAILQ_HEAD_INITIALIZER(ws_assignments);
2010-11-27 00:26:51 +01:00
/* We hope that those are supported and set them to true */
bool xcursor_supported = true;
bool xkb_supported = true;
2010-11-27 00:26:51 +01:00
bool force_xinerama = false;
/*
* This callback is only a dummy, see xcb_prepare_cb.
* See also man libev(3): "ev_prepare" and "ev_check" - customise your event loop
*
*/
static void xcb_got_event(EV_P_ struct ev_io *w, int revents) {
/* empty, because xcb_prepare_cb are used */
}
/*
* Called just before the event loop sleeps. Ensures xcbs incoming and outgoing
* queues are empty so that any activity will trigger another event loop
* iteration, and hence another xcb_prepare_cb invocation.
*
*/
static void xcb_prepare_cb(EV_P_ ev_prepare *w, int revents) {
/* Process all queued (and possibly new) events before the event loop
sleeps. */
xcb_generic_event_t *event;
while ((event = xcb_poll_for_event(conn)) != NULL) {
if (event->response_type == 0) {
if (event_is_ignored(event->sequence, 0))
DLOG("Expected X11 Error received for sequence %x\n", event->sequence);
else {
xcb_generic_error_t *error = (xcb_generic_error_t *)event;
DLOG("X11 Error received (probably harmless)! sequence 0x%x, error_code = %d\n",
error->sequence, error->error_code);
}
free(event);
continue;
}
/* Strip off the highest bit (set if the event is generated) */
int type = (event->response_type & 0x7F);
handle_event(type, event);
free(event);
}
/* Flush all queued events to X11. */
xcb_flush(conn);
}
/*
* Enable or disable the main X11 event handling function.
* This is used by drag_pointer() which has its own, modal event handler, which
* takes precedence over the normal event handler.
*
*/
void main_set_x11_cb(bool enable) {
DLOG("Setting main X11 callback to enabled=%d\n", enable);
if (enable) {
ev_prepare_start(main_loop, xcb_prepare);
/* Trigger the watcher explicitly to handle all remaining X11 events.
* drag_pointer()s event handler exits in the middle of the loop. */
ev_feed_event(main_loop, xcb_prepare, 0);
} else {
ev_prepare_stop(main_loop, xcb_prepare);
}
}
/*
* Exit handler which destroys the main_loop. Will trigger cleanup handlers.
*
*/
static void i3_exit(void) {
/* 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. */
#if EV_VERSION_MAJOR >= 4
ev_loop_destroy(main_loop);
#endif
if (*shmlogname != '\0') {
fprintf(stderr, "Closing SHM log \"%s\"\n", shmlogname);
fflush(stderr);
shm_unlink(shmlogname);
}
ipc_shutdown(SHUTDOWN_REASON_EXIT);
unlink(config.ipc_socket_path);
}
/*
* (One-shot) Handler for all signals with default action "Core", see signal(7)
*
* Unlinks the SHM log and re-raises the signal.
*
*/
static void handle_core_signal(int sig, siginfo_t *info, void *data) {
if (*shmlogname != '\0') {
shm_unlink(shmlogname);
}
raise(sig);
}
/*
* (One-shot) Handler for all signals with default action "Term", see signal(7)
*
* Exits the program gracefully.
*
*/
static void handle_term_signal(struct ev_loop *loop, ev_signal *signal, int revents) {
/* We exit gracefully here in the sense that cleanup handlers
* installed via atexit are invoked. */
exit(128 + signal->signum);
}
/*
* Set up handlers for all signals with default action "Term", see signal(7)
*
*/
static void setup_term_handlers(void) {
static struct ev_signal signal_watchers[6];
size_t num_watchers = sizeof(signal_watchers) / sizeof(signal_watchers[0]);
/* We have to rely on libev functionality here and should not use
* sigaction handlers because we need to invoke the exit handlers
* and cannot do so from an asynchronous signal handling context as
* not all code triggered during exit is signal safe (and exiting
* the main loop from said handler is not easily possible). libev's
* signal handlers does not impose such a constraint on us. */
ev_signal_init(&signal_watchers[0], handle_term_signal, SIGHUP);
ev_signal_init(&signal_watchers[1], handle_term_signal, SIGINT);
ev_signal_init(&signal_watchers[2], handle_term_signal, SIGALRM);
ev_signal_init(&signal_watchers[3], handle_term_signal, SIGTERM);
ev_signal_init(&signal_watchers[4], handle_term_signal, SIGUSR1);
ev_signal_init(&signal_watchers[5], handle_term_signal, SIGUSR1);
for (size_t i = 0; i < num_watchers; i++) {
ev_signal_start(main_loop, &signal_watchers[i]);
/* The signal handlers should not block ev_run from returning
* and so none of the signal handlers should hold a reference to
* the main loop. */
ev_unref(main_loop);
}
}
int main(int argc, char *argv[]) {
/* Keep a symbol pointing to the I3_VERSION string constant so that we have
* it in gdb backtraces. */
static const char *_i3_version __attribute__((used)) = I3_VERSION;
char *override_configpath = NULL;
bool autostart = true;
char *layout_path = NULL;
bool delete_layout_path = false;
bool disable_randr15 = false;
char *fake_outputs = NULL;
bool disable_signalhandler = false;
bool only_check_config = false;
static struct option long_options[] = {
2010-04-13 16:48:42 +02:00
{"no-autostart", no_argument, 0, 'a'},
{"config", required_argument, 0, 'c'},
{"version", no_argument, 0, 'v'},
{"moreversion", no_argument, 0, 'm'},
{"more-version", no_argument, 0, 'm'},
{"more_version", no_argument, 0, 'm'},
2010-04-13 16:48:42 +02:00
{"help", no_argument, 0, 'h'},
{"layout", required_argument, 0, 'L'},
{"restart", required_argument, 0, 0},
2010-04-13 16:48:42 +02:00
{"force-xinerama", no_argument, 0, 0},
{"force_xinerama", no_argument, 0, 0},
{"disable-randr15", no_argument, 0, 0},
{"disable_randr15", no_argument, 0, 0},
{"disable-signalhandler", no_argument, 0, 0},
{"shmlog-size", required_argument, 0, 0},
{"shmlog_size", required_argument, 0, 0},
{"get-socketpath", no_argument, 0, 0},
{"get_socketpath", no_argument, 0, 0},
{"fake_outputs", required_argument, 0, 0},
{"fake-outputs", required_argument, 0, 0},
{"force-old-config-parser-v4.4-only", no_argument, 0, 0},
{0, 0, 0, 0}};
int option_index = 0, opt;
setlocale(LC_ALL, "");
/* Get the RLIMIT_CORE limit at startup time to restore this before
* starting processes. */
getrlimit(RLIMIT_CORE, &original_rlimit_core);
/* Disable output buffering to make redirects in .xsession actually useful for debugging */
if (!isatty(fileno(stdout)))
setbuf(stdout, NULL);
srand(time(NULL));
/* Init logging *before* initializing debug_build to guarantee early
* (file) logging. */
init_logging();
2013-06-05 15:08:47 +02:00
/* On release builds, disable SHM logging by default. */
shmlog_size = (is_debug_build() || strstr(argv[0], "i3-with-shmlog") != NULL ? default_shmlog_size : 0);
start_argv = argv;
while ((opt = getopt_long(argc, argv, "c:CvmaL:hld:V", long_options, &option_index)) != -1) {
switch (opt) {
case 'a':
LOG("Autostart disabled using -a\n");
autostart = false;
break;
case 'L':
FREE(layout_path);
layout_path = sstrdup(optarg);
delete_layout_path = false;
break;
case 'c':
FREE(override_configpath);
override_configpath = sstrdup(optarg);
break;
case 'C':
LOG("Checking configuration file only (-C)\n");
only_check_config = true;
break;
case 'v':
printf("i3 version %s © 2009 Michael Stapelberg and contributors\n", i3_version);
exit(EXIT_SUCCESS);
break;
case 'm':
printf("Binary i3 version: %s © 2009 Michael Stapelberg and contributors\n", i3_version);
display_running_version();
exit(EXIT_SUCCESS);
break;
case 'V':
set_verbosity(true);
break;
case 'd':
LOG("Enabling debug logging\n");
set_debug_logging(true);
break;
case 'l':
/* DEPRECATED, ignored for the next 3 versions (3.e, 3.f, 3.g) */
break;
case 0:
if (strcmp(long_options[option_index].name, "force-xinerama") == 0 ||
strcmp(long_options[option_index].name, "force_xinerama") == 0) {
force_xinerama = true;
ELOG("Using Xinerama instead of RandR. This option should be "
"avoided at all cost because it does not refresh the list "
"of screens, so you cannot configure displays at runtime. "
"Please check if your driver really does not support RandR "
"and disable this option as soon as you can.\n");
break;
} else if (strcmp(long_options[option_index].name, "disable-randr15") == 0 ||
strcmp(long_options[option_index].name, "disable_randr15") == 0) {
disable_randr15 = true;
break;
} else if (strcmp(long_options[option_index].name, "disable-signalhandler") == 0) {
disable_signalhandler = true;
break;
} else if (strcmp(long_options[option_index].name, "get-socketpath") == 0 ||
strcmp(long_options[option_index].name, "get_socketpath") == 0) {
char *socket_path = root_atom_contents("I3_SOCKET_PATH", NULL, 0);
if (socket_path) {
printf("%s\n", socket_path);
exit(EXIT_SUCCESS);
}
exit(EXIT_FAILURE);
} else if (strcmp(long_options[option_index].name, "shmlog-size") == 0 ||
strcmp(long_options[option_index].name, "shmlog_size") == 0) {
shmlog_size = atoi(optarg);
/* Re-initialize logging immediately to get as many
* logmessages as possible into the SHM log. */
init_logging();
LOG("Limiting SHM log size to %d bytes\n", shmlog_size);
break;
} else if (strcmp(long_options[option_index].name, "restart") == 0) {
FREE(layout_path);
layout_path = sstrdup(optarg);
delete_layout_path = true;
break;
} else if (strcmp(long_options[option_index].name, "fake-outputs") == 0 ||
strcmp(long_options[option_index].name, "fake_outputs") == 0) {
LOG("Initializing fake outputs: %s\n", optarg);
fake_outputs = sstrdup(optarg);
break;
} else if (strcmp(long_options[option_index].name, "force-old-config-parser-v4.4-only") == 0) {
2012-12-24 15:46:57 +01:00
ELOG("You are passing --force-old-config-parser-v4.4-only, but that flag was removed by now.\n");
break;
}
/* fall-through */
default:
fprintf(stderr, "Usage: %s [-c configfile] [-d all] [-a] [-v] [-V] [-C]\n", argv[0]);
fprintf(stderr, "\n");
2011-10-02 19:55:15 +02:00
fprintf(stderr, "\t-a disable autostart ('exec' lines in config)\n");
fprintf(stderr, "\t-c <file> use the provided configfile instead\n");
fprintf(stderr, "\t-C validate configuration file and exit\n");
fprintf(stderr, "\t-d all enable debug output\n");
2011-10-02 19:55:15 +02:00
fprintf(stderr, "\t-L <file> path to the serialized layout during restarts\n");
fprintf(stderr, "\t-v display version and exit\n");
fprintf(stderr, "\t-V enable verbose mode\n");
fprintf(stderr, "\n");
fprintf(stderr, "\t--force-xinerama\n"
"\tUse Xinerama instead of RandR.\n"
"\tThis option should only be used if you are stuck with the\n"
"\told nVidia closed source driver (older than 302.17), which does\n"
"\tnot support RandR.\n");
2011-10-02 19:55:15 +02:00
fprintf(stderr, "\n");
fprintf(stderr, "\t--get-socketpath\n"
"\tRetrieve the i3 IPC socket path from X11, print it, then exit.\n");
fprintf(stderr, "\n");
fprintf(stderr, "\t--shmlog-size <limit>\n"
"\tLimits the size of the i3 SHM log to <limit> bytes. Setting this\n"
"\tto 0 disables SHM logging entirely.\n"
"\tThe default is %d bytes.\n",
shmlog_size);
fprintf(stderr, "\n");
fprintf(stderr, "If you pass plain text arguments, i3 will interpret them as a command\n"
"to send to a currently running i3 (like i3-msg). This allows you to\n"
"use nice and logical commands, such as:\n"
"\n"
"\ti3 border none\n"
"\ti3 floating toggle\n"
"\ti3 kill window\n"
"\n");
exit(EXIT_FAILURE);
}
}
if (only_check_config) {
exit(parse_configuration(override_configpath, false) ? 0 : 1);
}
/* If the user passes more arguments, we act like i3-msg would: Just send
* the arguments as an IPC message to i3. This allows for nice semantic
* commands such as 'i3 border none'. */
if (optind < argc) {
/* We enable verbose mode so that the user knows whats going on.
* This should make it easier to find mistakes when the user passes
* arguments by mistake. */
set_verbosity(true);
LOG("Additional arguments passed. Sending them as a command to i3.\n");
char *payload = NULL;
while (optind < argc) {
if (!payload) {
payload = sstrdup(argv[optind]);
} else {
char *both;
sasprintf(&both, "%s %s", payload, argv[optind]);
free(payload);
payload = both;
}
optind++;
}
DLOG("Command is: %s (%zd bytes)\n", payload, strlen(payload));
char *socket_path = root_atom_contents("I3_SOCKET_PATH", NULL, 0);
if (!socket_path) {
ELOG("Could not get i3 IPC socket path\n");
return 1;
}
int sockfd = socket(AF_LOCAL, SOCK_STREAM, 0);
if (sockfd == -1)
err(EXIT_FAILURE, "Could not create socket");
struct sockaddr_un addr;
memset(&addr, 0, sizeof(struct sockaddr_un));
addr.sun_family = AF_LOCAL;
strncpy(addr.sun_path, socket_path, sizeof(addr.sun_path) - 1);
2016-11-03 08:18:18 +01:00
FREE(socket_path);
if (connect(sockfd, (const struct sockaddr *)&addr, sizeof(struct sockaddr_un)) < 0)
err(EXIT_FAILURE, "Could not connect to i3");
if (ipc_send_message(sockfd, strlen(payload), I3_IPC_MESSAGE_TYPE_RUN_COMMAND,
(uint8_t *)payload) == -1)
err(EXIT_FAILURE, "IPC: write()");
2016-11-03 08:18:18 +01:00
FREE(payload);
uint32_t reply_length;
uint32_t reply_type;
uint8_t *reply;
int ret;
if ((ret = ipc_recv_message(sockfd, &reply_type, &reply_length, &reply)) != 0) {
if (ret == -1)
err(EXIT_FAILURE, "IPC: read()");
return 1;
}
if (reply_type != I3_IPC_REPLY_TYPE_COMMAND)
errx(EXIT_FAILURE, "IPC: received reply of type %d but expected %d (COMMAND)", reply_type, I3_IPC_REPLY_TYPE_COMMAND);
printf("%.*s\n", reply_length, reply);
2016-11-03 08:18:18 +01:00
FREE(reply);
return 0;
}
/* Enable logging to handle the case when the user did not specify --shmlog-size */
init_logging();
/* Try to enable core dumps by default when running a debug build */
2012-05-09 19:12:20 +02:00
if (is_debug_build()) {
struct rlimit limit = {RLIM_INFINITY, RLIM_INFINITY};
setrlimit(RLIMIT_CORE, &limit);
/* The following code is helpful, but not required. We thus dont pay
* much attention to error handling, non-linux or other edge cases. */
LOG("CORE DUMPS: You are running a development version of i3, so coredumps were automatically enabled (ulimit -c unlimited).\n");
size_t cwd_size = 1024;
char *cwd = smalloc(cwd_size);
char *cwd_ret;
while ((cwd_ret = getcwd(cwd, cwd_size)) == NULL && errno == ERANGE) {
cwd_size = cwd_size * 2;
cwd = srealloc(cwd, cwd_size);
}
if (cwd_ret != NULL)
LOG("CORE DUMPS: Your current working directory is \"%s\".\n", cwd);
int patternfd;
if ((patternfd = open("/proc/sys/kernel/core_pattern", O_RDONLY)) >= 0) {
memset(cwd, '\0', cwd_size);
if (read(patternfd, cwd, cwd_size) > 0)
/* a trailing newline is included in cwd */
LOG("CORE DUMPS: Your core_pattern is: %s", cwd);
close(patternfd);
}
free(cwd);
}
LOG("i3 %s starting\n", i3_version);
conn = xcb_connect(NULL, &conn_screen);
if (xcb_connection_has_error(conn))
errx(EXIT_FAILURE, "Cannot open display\n");
sndisplay = sn_xcb_display_new(conn, NULL, NULL);
/* Initialize the libev event loop. This needs to be done before loading
* the config file because the parser will install an ev_child watcher
* for the nagbar when config errors are found. */
main_loop = EV_DEFAULT;
if (main_loop == NULL)
die("Could not initialize libev. Bad LIBEV_FLAGS?\n");
root_screen = xcb_aux_get_screen(conn, conn_screen);
root = root_screen->root;
/* Place requests for the atoms we need as soon as possible */
#define xmacro(atom) \
xcb_intern_atom_cookie_t atom##_cookie = xcb_intern_atom(conn, 0, strlen(#atom), #atom);
#include "atoms.xmacro"
#undef xmacro
root_depth = root_screen->root_depth;
colormap = root_screen->default_colormap;
visual_type = xcb_aux_find_visual_by_attrs(root_screen, -1, 32);
if (visual_type != NULL) {
root_depth = xcb_aux_get_depth_of_visual(root_screen, visual_type->visual_id);
colormap = xcb_generate_id(conn);
xcb_void_cookie_t cm_cookie = xcb_create_colormap_checked(conn,
XCB_COLORMAP_ALLOC_NONE,
colormap,
root,
visual_type->visual_id);
xcb_generic_error_t *error = xcb_request_check(conn, cm_cookie);
if (error != NULL) {
ELOG("Could not create colormap. Error code: %d\n", error->error_code);
exit(EXIT_FAILURE);
}
} else {
visual_type = get_visualtype(root_screen);
}
init_dpi();
DLOG("root_depth = %d, visual_id = 0x%08x.\n", root_depth, visual_type->visual_id);
DLOG("root_screen->height_in_pixels = %d, root_screen->height_in_millimeters = %d\n",
root_screen->height_in_pixels, root_screen->height_in_millimeters);
DLOG("One logical pixel corresponds to %d physical pixels on this display.\n", logical_px(1));
2011-06-10 18:27:20 +02:00
xcb_get_geometry_cookie_t gcookie = xcb_get_geometry(conn, root);
xcb_query_pointer_cookie_t pointercookie = xcb_query_pointer(conn, root);
/* Setup NetWM atoms */
#define xmacro(name) \
do { \
xcb_intern_atom_reply_t *reply = xcb_intern_atom_reply(conn, name##_cookie, NULL); \
if (!reply) { \
ELOG("Could not get atom " #name "\n"); \
exit(-1); \
} \
A_##name = reply->atom; \
free(reply); \
} while (0);
#include "atoms.xmacro"
#undef xmacro
load_configuration(conn, override_configpath, false);
if (config.ipc_socket_path == NULL) {
/* Fall back to a file name in /tmp/ based on the PID */
if ((config.ipc_socket_path = getenv("I3SOCK")) == NULL)
config.ipc_socket_path = get_process_filename("ipc-socket");
else
config.ipc_socket_path = sstrdup(config.ipc_socket_path);
}
if (config.force_xinerama) {
force_xinerama = true;
}
xcb_void_cookie_t cookie;
cookie = xcb_change_window_attributes_checked(conn, root, XCB_CW_EVENT_MASK, (uint32_t[]){ROOT_EVENT_MASK});
xcb_generic_error_t *error = xcb_request_check(conn, cookie);
if (error != NULL) {
ELOG("Another window manager seems to be running (X error %d)\n", error->error_code);
#ifdef I3_ASAN_ENABLED
__lsan_do_leak_check();
#endif
return 1;
}
2011-06-10 18:27:20 +02:00
xcb_get_geometry_reply_t *greply = xcb_get_geometry_reply(conn, gcookie, NULL);
if (greply == NULL) {
ELOG("Could not get geometry of the root window, exiting\n");
return 1;
}
DLOG("root geometry reply: (%d, %d) %d x %d\n", greply->x, greply->y, greply->width, greply->height);
xcursor_load_cursors();
2010-11-27 00:26:51 +01:00
/* Set a cursor for the root window (otherwise the root window will show no
cursor until the first client is launched). */
if (xcursor_supported)
xcursor_set_root_cursor(XCURSOR_CURSOR_POINTER);
else
xcb_set_root_cursor(XCURSOR_CURSOR_POINTER);
const xcb_query_extension_reply_t *extreply;
extreply = xcb_get_extension_data(conn, &xcb_xkb_id);
xkb_supported = extreply->present;
if (!extreply->present) {
DLOG("xkb is not present on this server\n");
} else {
DLOG("initializing xcb-xkb\n");
xcb_xkb_use_extension(conn, XCB_XKB_MAJOR_VERSION, XCB_XKB_MINOR_VERSION);
xcb_xkb_select_events(conn,
XCB_XKB_ID_USE_CORE_KBD,
XCB_XKB_EVENT_TYPE_STATE_NOTIFY | XCB_XKB_EVENT_TYPE_MAP_NOTIFY | XCB_XKB_EVENT_TYPE_NEW_KEYBOARD_NOTIFY,
0,
XCB_XKB_EVENT_TYPE_STATE_NOTIFY | XCB_XKB_EVENT_TYPE_MAP_NOTIFY | XCB_XKB_EVENT_TYPE_NEW_KEYBOARD_NOTIFY,
0xff,
0xff,
NULL);
Use libxkbcommon for translating keysyms, support all XKB groups. fixes #1835 This commit improves the translation of keysyms to keycodes by loading keymaps using libxkbcommon-x11 and using libxkbcommon for figuring out the keymap, depending on each keybinding’s modifiers. This way, the upper layers of complex layouts are now usable with i3’s bindsym directive, such as de_neo’s layer 3 and higher. Furthermore, the commit generalizes the handling of different XKB groups. We formerly had support only for two separate groups, the default group 1, and group 2. While Mode_switch is only one way to switch to group 2, we called the binding option Mode_switch. With this commit, the new names Group1, Group2 (an alias for Mode_switch), Group3 and Group4 are introduced for configuring bindings. This is only useful for advanced keyboard layouts, such as people loading two keyboard layouts and switching between them (us, ru seems to be a popular combination). When grabbing keys, one can only specify the modifier mask, but not an XKB state mask (or value), so we still dynamically unbind and re-bind keys whenever the XKB group changes. The commit was manually tested using the following i3 config: bindsym Group4+n nop heya from group 4 bindsym Group3+n nop heya from group 3 bindsym Group2+n nop heya from group 2 bindsym n nop heya bindsym shift+N nop explicit shift binding bindsym shift+r nop implicit shift binding bindcode Group2+38 nop fallback overwritten in group 2 only bindcode 38 nop fallback …with the following layout: setxkbmap -layout "us,ua,ru,de" -variant ",winkeys,,neo" \ -option "grp:shift_caps_toggle,grp_led:scroll" \ -model pc104 -rules evdev By default (xkb group 1, us layout), pressing “n” will result in the “heya” message appearing. Pressing “a” will result in the “fallback” message appearing. “j” is not triggered. By pressing Shift+CapsLock you switch to the next group (xkb group 2, ua layout). Pressing “a” will result in the “fallback overwritten in group 2 only” message, pressing “n” will still result in “heya”. “j” is not triggered. In the next group (xkb group 3, ru layout), pressing “a” will result in the “fallback” message again, pressing “n” will result in “heya”, “j” is not triggered. In the last group (xkb group 4, de_neo layout), pressing “a” will still result in “fallback”, pressing “n” will result in “heya”, pressing “j” will result in “heya from group 4”. Pressing shift+n results in “explicit shift binding”, pressing shift+r results in “implicit shift binding”. This ensures that keysym translation falls back to looking at non-shift keys (“r” can be used instead of ”R”) and that the order of keybindings doesn’t play a role (“bindsym n” does not override “bindsym shift+n”, even though it’s specified earlier in the config). The fallback behavior ensures use-cases such as ticket #1775 are still covered. Only binding keys when the X server is in the corresponding XKB group ensures use-cases such as ticket #585 are still covered.
2015-08-23 22:49:32 +02:00
/* Setting both, XCB_XKB_PER_CLIENT_FLAG_GRABS_USE_XKB_STATE and
* XCB_XKB_PER_CLIENT_FLAG_LOOKUP_STATE_WHEN_GRABBED, will lead to the
* X server sending us the full XKB state in KeyPress and KeyRelease:
* https://sources.debian.net/src/xorg-server/2:1.17.2-1.1/xkb/xkbEvents.c/?hl=927#L927
*/
xcb_xkb_per_client_flags_reply_t *pcf_reply;
/* The last three parameters are unset because they are only relevant
* when using a feature called automatic reset of boolean controls:
* https://www.x.org/releases/X11R7.7/doc/kbproto/xkbproto.html#Automatic_Reset_of_Boolean_Controls
Use libxkbcommon for translating keysyms, support all XKB groups. fixes #1835 This commit improves the translation of keysyms to keycodes by loading keymaps using libxkbcommon-x11 and using libxkbcommon for figuring out the keymap, depending on each keybinding’s modifiers. This way, the upper layers of complex layouts are now usable with i3’s bindsym directive, such as de_neo’s layer 3 and higher. Furthermore, the commit generalizes the handling of different XKB groups. We formerly had support only for two separate groups, the default group 1, and group 2. While Mode_switch is only one way to switch to group 2, we called the binding option Mode_switch. With this commit, the new names Group1, Group2 (an alias for Mode_switch), Group3 and Group4 are introduced for configuring bindings. This is only useful for advanced keyboard layouts, such as people loading two keyboard layouts and switching between them (us, ru seems to be a popular combination). When grabbing keys, one can only specify the modifier mask, but not an XKB state mask (or value), so we still dynamically unbind and re-bind keys whenever the XKB group changes. The commit was manually tested using the following i3 config: bindsym Group4+n nop heya from group 4 bindsym Group3+n nop heya from group 3 bindsym Group2+n nop heya from group 2 bindsym n nop heya bindsym shift+N nop explicit shift binding bindsym shift+r nop implicit shift binding bindcode Group2+38 nop fallback overwritten in group 2 only bindcode 38 nop fallback …with the following layout: setxkbmap -layout "us,ua,ru,de" -variant ",winkeys,,neo" \ -option "grp:shift_caps_toggle,grp_led:scroll" \ -model pc104 -rules evdev By default (xkb group 1, us layout), pressing “n” will result in the “heya” message appearing. Pressing “a” will result in the “fallback” message appearing. “j” is not triggered. By pressing Shift+CapsLock you switch to the next group (xkb group 2, ua layout). Pressing “a” will result in the “fallback overwritten in group 2 only” message, pressing “n” will still result in “heya”. “j” is not triggered. In the next group (xkb group 3, ru layout), pressing “a” will result in the “fallback” message again, pressing “n” will result in “heya”, “j” is not triggered. In the last group (xkb group 4, de_neo layout), pressing “a” will still result in “fallback”, pressing “n” will result in “heya”, pressing “j” will result in “heya from group 4”. Pressing shift+n results in “explicit shift binding”, pressing shift+r results in “implicit shift binding”. This ensures that keysym translation falls back to looking at non-shift keys (“r” can be used instead of ”R”) and that the order of keybindings doesn’t play a role (“bindsym n” does not override “bindsym shift+n”, even though it’s specified earlier in the config). The fallback behavior ensures use-cases such as ticket #1775 are still covered. Only binding keys when the X server is in the corresponding XKB group ensures use-cases such as ticket #585 are still covered.
2015-08-23 22:49:32 +02:00
* */
pcf_reply = xcb_xkb_per_client_flags_reply(
conn,
xcb_xkb_per_client_flags(
conn,
XCB_XKB_ID_USE_CORE_KBD,
XCB_XKB_PER_CLIENT_FLAG_GRABS_USE_XKB_STATE | XCB_XKB_PER_CLIENT_FLAG_LOOKUP_STATE_WHEN_GRABBED,
XCB_XKB_PER_CLIENT_FLAG_GRABS_USE_XKB_STATE | XCB_XKB_PER_CLIENT_FLAG_LOOKUP_STATE_WHEN_GRABBED,
0 /* uint32_t ctrlsToChange */,
0 /* uint32_t autoCtrls */,
0 /* uint32_t autoCtrlsValues */),
NULL);
if (pcf_reply == NULL ||
!(pcf_reply->value & XCB_XKB_PER_CLIENT_FLAG_GRABS_USE_XKB_STATE)) {
ELOG("Could not set XCB_XKB_PER_CLIENT_FLAG_GRABS_USE_XKB_STATE\n");
}
if (pcf_reply == NULL ||
!(pcf_reply->value & XCB_XKB_PER_CLIENT_FLAG_LOOKUP_STATE_WHEN_GRABBED)) {
ELOG("Could not set XCB_XKB_PER_CLIENT_FLAG_LOOKUP_STATE_WHEN_GRABBED\n");
}
free(pcf_reply);
xkb_base = extreply->first_event;
}
restore_connect();
property_handlers_init();
ewmh_setup_hints();
keysyms = xcb_key_symbols_alloc(conn);
xcb_numlock_mask = aio_get_mod_mask_for(XCB_NUM_LOCK, keysyms);
Use libxkbcommon for translating keysyms, support all XKB groups. fixes #1835 This commit improves the translation of keysyms to keycodes by loading keymaps using libxkbcommon-x11 and using libxkbcommon for figuring out the keymap, depending on each keybinding’s modifiers. This way, the upper layers of complex layouts are now usable with i3’s bindsym directive, such as de_neo’s layer 3 and higher. Furthermore, the commit generalizes the handling of different XKB groups. We formerly had support only for two separate groups, the default group 1, and group 2. While Mode_switch is only one way to switch to group 2, we called the binding option Mode_switch. With this commit, the new names Group1, Group2 (an alias for Mode_switch), Group3 and Group4 are introduced for configuring bindings. This is only useful for advanced keyboard layouts, such as people loading two keyboard layouts and switching between them (us, ru seems to be a popular combination). When grabbing keys, one can only specify the modifier mask, but not an XKB state mask (or value), so we still dynamically unbind and re-bind keys whenever the XKB group changes. The commit was manually tested using the following i3 config: bindsym Group4+n nop heya from group 4 bindsym Group3+n nop heya from group 3 bindsym Group2+n nop heya from group 2 bindsym n nop heya bindsym shift+N nop explicit shift binding bindsym shift+r nop implicit shift binding bindcode Group2+38 nop fallback overwritten in group 2 only bindcode 38 nop fallback …with the following layout: setxkbmap -layout "us,ua,ru,de" -variant ",winkeys,,neo" \ -option "grp:shift_caps_toggle,grp_led:scroll" \ -model pc104 -rules evdev By default (xkb group 1, us layout), pressing “n” will result in the “heya” message appearing. Pressing “a” will result in the “fallback” message appearing. “j” is not triggered. By pressing Shift+CapsLock you switch to the next group (xkb group 2, ua layout). Pressing “a” will result in the “fallback overwritten in group 2 only” message, pressing “n” will still result in “heya”. “j” is not triggered. In the next group (xkb group 3, ru layout), pressing “a” will result in the “fallback” message again, pressing “n” will result in “heya”, “j” is not triggered. In the last group (xkb group 4, de_neo layout), pressing “a” will still result in “fallback”, pressing “n” will result in “heya”, pressing “j” will result in “heya from group 4”. Pressing shift+n results in “explicit shift binding”, pressing shift+r results in “implicit shift binding”. This ensures that keysym translation falls back to looking at non-shift keys (“r” can be used instead of ”R”) and that the order of keybindings doesn’t play a role (“bindsym n” does not override “bindsym shift+n”, even though it’s specified earlier in the config). The fallback behavior ensures use-cases such as ticket #1775 are still covered. Only binding keys when the X server is in the corresponding XKB group ensures use-cases such as ticket #585 are still covered.
2015-08-23 22:49:32 +02:00
if (!load_keymap())
die("Could not load keymap\n");
translate_keysyms();
Use libxkbcommon for translating keysyms, support all XKB groups. fixes #1835 This commit improves the translation of keysyms to keycodes by loading keymaps using libxkbcommon-x11 and using libxkbcommon for figuring out the keymap, depending on each keybinding’s modifiers. This way, the upper layers of complex layouts are now usable with i3’s bindsym directive, such as de_neo’s layer 3 and higher. Furthermore, the commit generalizes the handling of different XKB groups. We formerly had support only for two separate groups, the default group 1, and group 2. While Mode_switch is only one way to switch to group 2, we called the binding option Mode_switch. With this commit, the new names Group1, Group2 (an alias for Mode_switch), Group3 and Group4 are introduced for configuring bindings. This is only useful for advanced keyboard layouts, such as people loading two keyboard layouts and switching between them (us, ru seems to be a popular combination). When grabbing keys, one can only specify the modifier mask, but not an XKB state mask (or value), so we still dynamically unbind and re-bind keys whenever the XKB group changes. The commit was manually tested using the following i3 config: bindsym Group4+n nop heya from group 4 bindsym Group3+n nop heya from group 3 bindsym Group2+n nop heya from group 2 bindsym n nop heya bindsym shift+N nop explicit shift binding bindsym shift+r nop implicit shift binding bindcode Group2+38 nop fallback overwritten in group 2 only bindcode 38 nop fallback …with the following layout: setxkbmap -layout "us,ua,ru,de" -variant ",winkeys,,neo" \ -option "grp:shift_caps_toggle,grp_led:scroll" \ -model pc104 -rules evdev By default (xkb group 1, us layout), pressing “n” will result in the “heya” message appearing. Pressing “a” will result in the “fallback” message appearing. “j” is not triggered. By pressing Shift+CapsLock you switch to the next group (xkb group 2, ua layout). Pressing “a” will result in the “fallback overwritten in group 2 only” message, pressing “n” will still result in “heya”. “j” is not triggered. In the next group (xkb group 3, ru layout), pressing “a” will result in the “fallback” message again, pressing “n” will result in “heya”, “j” is not triggered. In the last group (xkb group 4, de_neo layout), pressing “a” will still result in “fallback”, pressing “n” will result in “heya”, pressing “j” will result in “heya from group 4”. Pressing shift+n results in “explicit shift binding”, pressing shift+r results in “implicit shift binding”. This ensures that keysym translation falls back to looking at non-shift keys (“r” can be used instead of ”R”) and that the order of keybindings doesn’t play a role (“bindsym n” does not override “bindsym shift+n”, even though it’s specified earlier in the config). The fallback behavior ensures use-cases such as ticket #1775 are still covered. Only binding keys when the X server is in the corresponding XKB group ensures use-cases such as ticket #585 are still covered.
2015-08-23 22:49:32 +02:00
grab_all_keys(conn);
2011-01-05 00:16:10 +01:00
bool needs_tree_init = true;
if (layout_path != NULL) {
LOG("Trying to restore the layout from \"%s\".\n", layout_path);
2011-06-10 18:27:20 +02:00
needs_tree_init = !tree_restore(layout_path, greply);
if (delete_layout_path) {
2011-01-05 00:16:10 +01:00
unlink(layout_path);
const char *dir = dirname(layout_path);
/* possibly fails with ENOTEMPTY if there are files (or
* sockets) left. */
rmdir(dir);
}
2011-01-05 00:16:10 +01:00
}
if (needs_tree_init)
2011-06-10 18:27:20 +02:00
tree_init(greply);
free(greply);
2011-01-05 00:16:10 +01:00
/* Setup fake outputs for testing */
if (fake_outputs == NULL && config.fake_outputs != NULL)
fake_outputs = config.fake_outputs;
if (fake_outputs != NULL) {
fake_outputs_init(fake_outputs);
FREE(fake_outputs);
config.fake_outputs = NULL;
} else if (force_xinerama) {
/* Force Xinerama (for drivers which don't support RandR yet, esp. the
* nVidia binary graphics driver), when specified either in the config
* file or on command-line */
xinerama_init();
} else {
DLOG("Checking for XRandR...\n");
randr_init(&randr_base, disable_randr15 || config.disable_randr15);
}
/* We need to force disabling outputs which have been loaded from the
* layout file but are no longer active. This can happen if the output has
* been disabled in the short time between writing the restart layout file
* and restarting i3. See #2326. */
if (layout_path != NULL && randr_base > -1) {
Con *con;
TAILQ_FOREACH(con, &(croot->nodes_head), nodes) {
Output *output;
TAILQ_FOREACH(output, &outputs, outputs) {
if (output->active || strcmp(con->name, output_primary_name(output)) != 0)
continue;
/* This will correctly correlate the output with its content
* container. We need to make the connection to properly
* disable the output. */
if (output->con == NULL) {
output_init_con(output);
output->changed = false;
}
output->to_be_disabled = true;
randr_disable_output(output);
}
}
}
FREE(layout_path);
scratchpad_fix_resolution();
xcb_query_pointer_reply_t *pointerreply;
Output *output = NULL;
if (!(pointerreply = xcb_query_pointer_reply(conn, pointercookie, NULL))) {
ELOG("Could not query pointer position, using first screen\n");
} else {
DLOG("Pointer at %d, %d\n", pointerreply->root_x, pointerreply->root_y);
output = get_output_containing(pointerreply->root_x, pointerreply->root_y);
if (!output) {
ELOG("ERROR: No screen at (%d, %d), starting on the first screen\n",
pointerreply->root_x, pointerreply->root_y);
output = get_first_output();
}
con_focus(con_descend_focused(output_get_content(output->con)));
2016-01-09 12:07:47 +01:00
free(pointerreply);
}
tree_render();
/* Create the UNIX domain socket for IPC */
int ipc_socket = ipc_create_socket(config.ipc_socket_path);
if (ipc_socket == -1) {
ELOG("Could not create the IPC socket, IPC disabled\n");
} else {
struct ev_io *ipc_io = scalloc(1, sizeof(struct ev_io));
ev_io_init(ipc_io, ipc_new_client, ipc_socket, EV_READ);
ev_io_start(main_loop, ipc_io);
}
/* Also handle the UNIX domain sockets passed via socket activation. The
* parameter 1 means "remove the environment variables", we dont want to
* pass these to child processes. */
listen_fds = sd_listen_fds(0);
if (listen_fds < 0)
ELOG("socket activation: Error in sd_listen_fds\n");
else if (listen_fds == 0)
DLOG("socket activation: no sockets passed\n");
else {
int flags;
for (int fd = SD_LISTEN_FDS_START;
fd < (SD_LISTEN_FDS_START + listen_fds);
fd++) {
DLOG("socket activation: also listening on fd %d\n", fd);
/* sd_listen_fds() enables FD_CLOEXEC by default.
* However, we need to keep the file descriptors open for in-place
* restarting, therefore we explicitly disable FD_CLOEXEC. */
if ((flags = fcntl(fd, F_GETFD)) < 0 ||
fcntl(fd, F_SETFD, flags & ~FD_CLOEXEC) < 0) {
ELOG("Could not disable FD_CLOEXEC on fd %d\n", fd);
}
struct ev_io *ipc_io = scalloc(1, sizeof(struct ev_io));
ev_io_init(ipc_io, ipc_new_client, fd, EV_READ);
ev_io_start(main_loop, ipc_io);
}
}
/* Set up i3 specific atoms like I3_SOCKET_PATH and I3_CONFIG_PATH */
x_set_i3_atoms();
ewmh_update_workarea();
/* Set the ewmh desktop properties. */
ewmh_update_current_desktop();
ewmh_update_number_of_desktops();
ewmh_update_desktop_names();
ewmh_update_desktop_viewport();
struct ev_io *xcb_watcher = scalloc(1, sizeof(struct ev_io));
xcb_prepare = scalloc(1, sizeof(struct ev_prepare));
ev_io_init(xcb_watcher, xcb_got_event, xcb_get_file_descriptor(conn), EV_READ);
ev_io_start(main_loop, xcb_watcher);
ev_prepare_init(xcb_prepare, xcb_prepare_cb);
ev_prepare_start(main_loop, xcb_prepare);
xcb_flush(conn);
/* What follows is a fugly consequence of X11 protocol race conditions like
* the following: In an i3 in-place restart, i3 will reparent all windows
* to the root window, then exec() itself. In the new process, it calls
* manage_existing_windows. However, in case any application sent a
* generated UnmapNotify message to the WM (as GIMP does), this message
* will be handled by i3 *after* managing the window, thus i3 thinks the
* window just closed itself. In reality, the message was sent in the time
* period where i3 wasnt running yet.
*
* To prevent this, we grab the server (disables processing of any other
* connections), then discard all pending events (since we didnt do
* anything, there cannot be any meaningful responses), then ungrab the
* server. */
xcb_grab_server(conn);
{
xcb_aux_sync(conn);
xcb_generic_event_t *event;
while ((event = xcb_poll_for_event(conn)) != NULL) {
if (event->response_type == 0) {
free(event);
continue;
}
/* Strip off the highest bit (set if the event is generated) */
int type = (event->response_type & 0x7F);
/* We still need to handle MapRequests which are sent in the
* timespan starting from when we register as a window manager and
* this piece of code which drops events. */
if (type == XCB_MAP_REQUEST)
handle_event(type, event);
free(event);
}
manage_existing_windows(root);
}
xcb_ungrab_server(conn);
if (autostart) {
LOG("This is not an in-place restart, copying root window contents to a pixmap\n");
xcb_screen_t *root = xcb_aux_get_screen(conn, conn_screen);
uint16_t width = root->width_in_pixels;
uint16_t height = root->height_in_pixels;
xcb_pixmap_t pixmap = xcb_generate_id(conn);
xcb_gcontext_t gc = xcb_generate_id(conn);
xcb_create_pixmap(conn, root->root_depth, pixmap, root->root, width, height);
xcb_create_gc(conn, gc, root->root,
XCB_GC_FUNCTION | XCB_GC_PLANE_MASK | XCB_GC_FILL_STYLE | XCB_GC_SUBWINDOW_MODE,
(uint32_t[]){XCB_GX_COPY, ~0, XCB_FILL_STYLE_SOLID, XCB_SUBWINDOW_MODE_INCLUDE_INFERIORS});
xcb_copy_area(conn, root->root, pixmap, gc, 0, 0, 0, 0, width, height);
xcb_change_window_attributes(conn, root->root, XCB_CW_BACK_PIXMAP, (uint32_t[]){pixmap});
xcb_flush(conn);
xcb_free_gc(conn, gc);
xcb_free_pixmap(conn, pixmap);
}
#if defined(__OpenBSD__)
if (pledge("stdio rpath wpath cpath proc exec unix", NULL) == -1)
err(EXIT_FAILURE, "pledge");
#endif
if (!disable_signalhandler)
setup_signal_handler();
else {
struct sigaction action;
action.sa_sigaction = handle_core_signal;
action.sa_flags = SA_NODEFER | SA_RESETHAND | SA_SIGINFO;
sigemptyset(&action.sa_mask);
/* Catch all signals with default action "Core", see signal(7) */
if (sigaction(SIGQUIT, &action, NULL) == -1 ||
sigaction(SIGILL, &action, NULL) == -1 ||
sigaction(SIGABRT, &action, NULL) == -1 ||
sigaction(SIGFPE, &action, NULL) == -1 ||
sigaction(SIGSEGV, &action, NULL) == -1)
ELOG("Could not setup signal handler.\n");
}
setup_term_handlers();
2010-12-30 21:09:32 +01:00
/* Ignore SIGPIPE to survive errors when an IPC client disconnects
2015-03-12 05:41:43 +01:00
* while we are sending them a message */
2010-12-30 21:09:32 +01:00
signal(SIGPIPE, SIG_IGN);
2010-11-23 00:40:05 +01:00
/* Autostarting exec-lines */
if (autostart) {
while (!TAILQ_EMPTY(&autostarts)) {
struct Autostart *exec = TAILQ_FIRST(&autostarts);
2010-11-23 00:40:05 +01:00
LOG("auto-starting %s\n", exec->command);
start_application(exec->command, exec->no_startup_id);
FREE(exec->command);
TAILQ_REMOVE(&autostarts, exec, autostarts);
FREE(exec);
2010-11-23 00:40:05 +01:00
}
}
/* Autostarting exec_always-lines */
while (!TAILQ_EMPTY(&autostarts_always)) {
struct Autostart *exec_always = TAILQ_FIRST(&autostarts_always);
LOG("auto-starting (always!) %s\n", exec_always->command);
start_application(exec_always->command, exec_always->no_startup_id);
FREE(exec_always->command);
TAILQ_REMOVE(&autostarts_always, exec_always, autostarts_always);
FREE(exec_always);
}
/* Start i3bar processes for all configured bars */
Barconfig *barconfig;
TAILQ_FOREACH(barconfig, &barconfigs, configs) {
char *command = NULL;
sasprintf(&command, "%s --bar_id=%s --socket=\"%s\"",
barconfig->i3bar_command ? barconfig->i3bar_command : "i3bar",
barconfig->id, current_socketpath);
LOG("Starting bar process: %s\n", command);
start_application(command, true);
free(command);
}
2017-11-18 18:48:38 +01:00
/* Make sure to destroy the event loop to invoke the cleanup callbacks
* when calling exit() */
atexit(i3_exit);
ev_loop(main_loop, 0);
}