mpv/common/msg.c

899 lines
28 KiB
C
Raw Normal View History

/*
* This file is part of mpv.
*
* mpv is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* mpv is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with mpv. If not, see <http://www.gnu.org/licenses/>.
*/
#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
#include <unistd.h>
#include <assert.h>
#include <pthread.h>
#include <stdint.h>
#include "mpv_talloc.h"
#include "misc/bstr.h"
#include "osdep/atomic.h"
#include "common/common.h"
#include "common/global.h"
#include "misc/bstr.h"
#include "options/options.h"
#include "options/path.h"
#include "osdep/terminal.h"
#include "osdep/io.h"
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
#include "osdep/threads.h"
#include "osdep/timer.h"
#include "libmpv/client.h"
#include "msg.h"
#include "msg_control.h"
#define TERM_BUF 100
struct mp_log_root {
struct mpv_global *global;
pthread_mutex_t lock;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_t log_file_lock;
pthread_cond_t log_file_wakeup;
// --- protected by lock
char **msg_levels;
bool use_terminal; // make accesses to stderr/stdout
2013-12-21 22:11:12 +00:00
bool module;
bool show_time;
int blank_lines; // number of lines usable by status
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
int status_lines; // number of current status lines
2013-12-21 22:11:12 +00:00
bool color;
int verbose;
bool really_quiet;
2013-12-21 22:11:12 +00:00
bool force_stderr;
struct mp_log_buffer **buffers;
int num_buffers;
struct mp_log_buffer *early_buffer;
FILE *stats_file;
bstr buffer;
2013-12-21 22:11:12 +00:00
// --- must be accessed atomically
/* This is incremented every time the msglevels must be reloaded.
* (This is perhaps better than maintaining a globally accessible and
* synchronized mp_log tree.) */
atomic_ulong reload_counter;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
// --- owner thread only (caller of mp_msg_init() etc.)
char *log_path;
char *stats_path;
pthread_t log_file_thread;
// --- owner thread only, but frozen while log_file_thread is running
FILE *log_file;
struct mp_log_buffer *log_file_buffer;
// --- protected by log_file_lock
bool log_file_thread_active; // also termination signal for the thread
};
struct mp_log {
struct mp_log_root *root;
const char *prefix;
const char *verbose_prefix;
int max_level; // minimum log level for this instance
int level; // minimum log level for any outputs
int terminal_level; // minimum log level for terminal output
atomic_ulong reload_counter;
char *partial;
};
struct mp_log_buffer {
struct mp_log_root *root;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_t lock;
// --- protected by lock
struct mp_log_buffer_entry **entries; // ringbuffer
int capacity; // total space in entries[]
int entry0; // first (oldest) entry index
int num_entries; // number of valid entries after entry0
uint64_t dropped; // number of skipped entries
bool silent;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
// --- immutable
void (*wakeup_cb)(void *ctx);
void *wakeup_cb_ctx;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
int level;
};
static const struct mp_log null_log = {0};
struct mp_log *const mp_null_log = (struct mp_log *)&null_log;
static bool match_mod(const char *name, const char *mod)
{
if (!strcmp(mod, "all"))
return true;
// Path prefix matches
bstr b = bstr0(name);
return bstr_eatstart0(&b, mod) && (bstr_eatstart0(&b, "/") || !b.len);
}
static void update_loglevel(struct mp_log *log)
{
struct mp_log_root *root = log->root;
pthread_mutex_lock(&root->lock);
log->level = MSGL_STATUS + root->verbose; // default log level
for (int n = 0; root->msg_levels && root->msg_levels[n * 2 + 0]; n++) {
if (match_mod(log->verbose_prefix, root->msg_levels[n * 2 + 0]))
log->level = mp_msg_find_level(root->msg_levels[n * 2 + 1]);
}
log->terminal_level = log->level;
for (int n = 0; n < log->root->num_buffers; n++) {
int buffer_level = log->root->buffers[n]->level;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
if (buffer_level == MP_LOG_BUFFER_MSGL_LOGFILE)
buffer_level = MSGL_DEBUG;
if (buffer_level != MP_LOG_BUFFER_MSGL_TERM)
log->level = MPMAX(log->level, buffer_level);
}
if (log->root->log_file)
log->level = MPMAX(log->level, MSGL_DEBUG);
if (log->root->stats_file)
log->level = MPMAX(log->level, MSGL_STATS);
log->level = MPMIN(log->level, log->max_level);
if (root->really_quiet)
log->level = -1;
atomic_store(&log->reload_counter, atomic_load(&log->root->reload_counter));
pthread_mutex_unlock(&root->lock);
}
// Set (numerically) the maximum level that should still be output for this log
// instances. E.g. lev=MSGL_WARN => show only warnings and errors.
void mp_msg_set_max_level(struct mp_log *log, int lev)
{
if (!log->root)
return;
pthread_mutex_lock(&log->root->lock);
log->max_level = MPCLAMP(lev, -1, MSGL_MAX);
pthread_mutex_unlock(&log->root->lock);
update_loglevel(log);
}
// Get the current effective msg level.
// Thread-safety: see mp_msg().
int mp_msg_level(struct mp_log *log)
{
struct mp_log_root *root = log->root;
if (!root)
return -1;
if (atomic_load_explicit(&log->reload_counter, memory_order_relaxed) !=
atomic_load_explicit(&root->reload_counter, memory_order_relaxed))
{
update_loglevel(log);
}
return log->level;
}
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
// Reposition cursor and clear lines for outputting the status line. In certain
// cases, like term OSD and subtitle display, the status can consist of
// multiple lines.
static void prepare_status_line(struct mp_log_root *root, char *new_status)
{
FILE *f = stderr;
size_t new_lines = 1;
char *tmp = new_status;
while (1) {
tmp = strchr(tmp, '\n');
if (!tmp)
break;
new_lines++;
tmp++;
}
size_t old_lines = root->status_lines;
if (!new_status[0] && old_lines == 0)
return; // nothing to clear
size_t clear_lines = MPMIN(MPMAX(new_lines, old_lines), root->blank_lines);
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
// clear the status line itself
fprintf(f, "\r\033[K");
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
// and clear all previous old lines
for (size_t n = 1; n < clear_lines; n++)
fprintf(f, "\033[A\r\033[K");
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
// skip "unused" blank lines, so that status is aligned to term bottom
for (size_t n = new_lines; n < clear_lines; n++)
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
fprintf(f, "\n");
root->status_lines = new_lines;
root->blank_lines = MPMAX(root->blank_lines, new_lines);
}
static void flush_status_line(struct mp_log_root *root)
{
// If there was a status line, don't overwrite it, but skip it.
if (root->status_lines)
fprintf(stderr, "\n");
root->status_lines = 0;
root->blank_lines = 0;
}
void mp_msg_flush_status_line(struct mp_log *log)
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
{
if (log->root) {
pthread_mutex_lock(&log->root->lock);
flush_status_line(log->root);
pthread_mutex_unlock(&log->root->lock);
}
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
}
void mp_msg_set_term_title(struct mp_log *log, const char *title)
{
if (log->root && title) {
// Lock because printf to terminal is not necessarily atomic.
pthread_mutex_lock(&log->root->lock);
fprintf(stderr, "\e]0;%s\007", title);
pthread_mutex_unlock(&log->root->lock);
}
}
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
bool mp_msg_has_status_line(struct mpv_global *global)
{
struct mp_log_root *root = global->log->root;
pthread_mutex_lock(&root->lock);
bool r = root->status_lines > 0;
pthread_mutex_unlock(&root->lock);
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
return r;
}
static void set_term_color(FILE *stream, int c)
{
if (c == -1) {
fprintf(stream, "\033[0m");
} else {
fprintf(stream, "\033[%d;3%dm", c >> 3, c & 7);
}
}
static void set_msg_color(FILE* stream, int lev)
{
static const int v_colors[] = {9, 1, 3, -1, -1, 2, 8, 8, 8, -1};
set_term_color(stream, v_colors[lev]);
}
static void pretty_print_module(FILE* stream, const char *prefix, bool use_color, int lev)
{
// Use random color based on the name of the module
if (use_color) {
size_t prefix_len = strlen(prefix);
unsigned int mod = 0;
for (int i = 0; i < prefix_len; ++i)
mod = mod * 33 + prefix[i];
set_term_color(stream, (mod + 1) % 15 + 1);
}
fprintf(stream, "%10s", prefix);
if (use_color)
set_term_color(stream, -1);
fprintf(stream, ": ");
if (use_color)
set_msg_color(stream, lev);
}
static bool test_terminal_level(struct mp_log *log, int lev)
{
return lev <= log->terminal_level && log->root->use_terminal &&
!(lev == MSGL_STATUS && terminal_in_background());
}
static void print_terminal_line(struct mp_log *log, int lev,
char *text, char *trail)
{
if (!test_terminal_level(log, lev))
return;
struct mp_log_root *root = log->root;
FILE *stream = (root->force_stderr || lev == MSGL_STATUS) ? stderr : stdout;
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
if (lev != MSGL_STATUS)
player: redo terminal OSD and status line handling The terminal OSD code includes the handling of the terminal status line, showing player OSD messages on the terminal, and showing subtitles on terminal (the latter two only if there is no video window, or if terminal OSD is forced). This didn't handle some corner cases correctly. For example, showing an OSD message on the terminal always cleared the previous line, even if the line was an important message (or even just the command prompt, if most other messages were silenced). Attempt to handle this correctly by keeping track of how many lines the terminal OSD currently consists of. Since there could be race conditions with other messages being printed, implement this in msg.c. Now msg.c expects that MSGL_STATUS messages rewrite the status line, so the caller is forced to use a single mp_msg() call to set the status line. Instead of littering print_status() all over the place, update the status only once per playloop iteration in update_osd_msg(). In audio- only mode, the status line might now be a little bit off, but it's perhaps ok. Print the status line only if it has changed, or if another message was printed. This might help with extremely slow terminals, although in audio+video mode, it'll still be updated very often (A-V sync display changes on every frame). Instead of hardcoding the terminal sequences, use terminfo/termcap to get the sequences. Remove the --term-osd-esc option, which allowed to override the hardcoded escapes - it's useless now. The fallback for terminals with no escape sequences for moving the cursor and clearing a line is removed. This somewhat breaks status line display on these terminals, including the MS Windows console: instead of querying the terminal size and clearing the line manually by padding the output with spaces, the line is simply not cleared. I don't expect this to be a problem on UNIX, and on MS Windows we could emulate escape sequences. Note that terminal OSD (other than the status line) was broken anyway on these terminals. In osd.c, the function get_term_width() is not used anymore, so remove it. To remind us that the MS Windows console apparently adds a line break when writint the last column, adjust screen_width in terminal- win.c accordingly.
2014-01-13 19:05:41 +00:00
flush_status_line(root);
2013-12-21 22:11:12 +00:00
if (root->color)
set_msg_color(stream, lev);
if (root->show_time)
fprintf(stream, "[%10.6f] ", (mp_time_us() - MP_START_TIME) / 1e6);
const char *prefix = log->prefix;
if ((lev >= MSGL_V) || root->verbose || root->module)
prefix = log->verbose_prefix;
if (prefix) {
if (root->module) {
pretty_print_module(stream, prefix, root->color, lev);
} else {
fprintf(stream, "[%s] ", prefix);
}
}
fprintf(stream, "%s%s", text, trail);
2013-12-21 22:11:12 +00:00
if (root->color)
set_term_color(stream, -1);
fflush(stream);
}
static struct mp_log_buffer_entry *log_buffer_read(struct mp_log_buffer *buffer)
{
assert(buffer->num_entries);
struct mp_log_buffer_entry *res = buffer->entries[buffer->entry0];
buffer->entry0 = (buffer->entry0 + 1) % buffer->capacity;
buffer->num_entries -= 1;
return res;
}
static void write_msg_to_buffers(struct mp_log *log, int lev, char *text)
{
struct mp_log_root *root = log->root;
for (int n = 0; n < root->num_buffers; n++) {
struct mp_log_buffer *buffer = root->buffers[n];
bool wakeup = false;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_lock(&buffer->lock);
int buffer_level = buffer->level;
if (buffer_level == MP_LOG_BUFFER_MSGL_TERM)
buffer_level = log->terminal_level;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
if (buffer_level == MP_LOG_BUFFER_MSGL_LOGFILE)
buffer_level = MPMAX(log->terminal_level, MSGL_DEBUG);
if (lev <= buffer_level && lev != MSGL_STATUS) {
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
if (buffer->level == MP_LOG_BUFFER_MSGL_LOGFILE) {
// If the buffer is full, block until we can write again.
bool dead = false;
while (buffer->num_entries == buffer->capacity && !dead) {
// Temporary unlock is OK; buffer->level is immutable, and
// buffer can't go away because the global log lock is held.
pthread_mutex_unlock(&buffer->lock);
pthread_mutex_lock(&root->log_file_lock);
if (root->log_file_thread_active) {
pthread_cond_wait(&root->log_file_wakeup,
&root->log_file_lock);
} else {
dead = true;
}
pthread_mutex_unlock(&root->log_file_lock);
pthread_mutex_lock(&buffer->lock);
}
}
if (buffer->num_entries == buffer->capacity) {
struct mp_log_buffer_entry *skip = log_buffer_read(buffer);
talloc_free(skip);
buffer->dropped += 1;
}
struct mp_log_buffer_entry *entry = talloc_ptrtype(NULL, entry);
*entry = (struct mp_log_buffer_entry) {
.prefix = talloc_strdup(entry, log->verbose_prefix),
.level = lev,
.text = talloc_strdup(entry, text),
};
int pos = (buffer->entry0 + buffer->num_entries) % buffer->capacity;
buffer->entries[pos] = entry;
buffer->num_entries += 1;
if (buffer->wakeup_cb && !buffer->silent)
wakeup = true;
}
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_unlock(&buffer->lock);
if (wakeup)
buffer->wakeup_cb(buffer->wakeup_cb_ctx);
}
}
static void dump_stats(struct mp_log *log, int lev, char *text)
{
struct mp_log_root *root = log->root;
if (lev == MSGL_STATS && root->stats_file)
fprintf(root->stats_file, "%"PRId64" %s\n", mp_time_us(), text);
}
void mp_msg_va(struct mp_log *log, int lev, const char *format, va_list va)
{
if (!mp_msg_test(log, lev))
return; // do not display
struct mp_log_root *root = log->root;
pthread_mutex_lock(&root->lock);
root->buffer.len = 0;
if (log->partial[0])
bstr_xappend_asprintf(root, &root->buffer, "%s", log->partial);
log->partial[0] = '\0';
bstr_xappend_vasprintf(root, &root->buffer, format, va);
char *text = root->buffer.start;
if (lev == MSGL_STATS) {
dump_stats(log, lev, text);
} else if (lev == MSGL_STATUS && !test_terminal_level(log, lev)) {
/* discard */
} else {
if (lev == MSGL_STATUS)
prepare_status_line(root, text);
// Split away each line. Normally we require full lines; buffer partial
// lines if they happen.
while (1) {
char *end = strchr(text, '\n');
if (!end)
break;
char *next = &end[1];
char saved = next[0];
next[0] = '\0';
print_terminal_line(log, lev, text, "");
write_msg_to_buffers(log, lev, text);
next[0] = saved;
text = next;
}
if (lev == MSGL_STATUS) {
if (text[0])
print_terminal_line(log, lev, text, "\r");
} else if (text[0]) {
int size = strlen(text) + 1;
if (talloc_get_size(log->partial) < size)
log->partial = talloc_realloc(NULL, log->partial, char, size);
memcpy(log->partial, text, size);
}
}
pthread_mutex_unlock(&root->lock);
}
static void destroy_log(void *ptr)
{
struct mp_log *log = ptr;
// This is not managed via talloc itself, because mp_msg calls must be
// thread-safe, while talloc is not thread-safe.
talloc_free(log->partial);
}
// Create a new log context, which uses talloc_ctx as talloc parent, and parent
// as logical parent.
// The name is the prefix put before the output. It's usually prefixed by the
// parent's name. If the name starts with "/", the parent's name is not
// prefixed (except in verbose mode), and if it starts with "!", the name is
2013-08-05 17:05:48 +00:00
// not printed at all (except in verbose mode).
// If name is NULL, the parent's name/prefix is used.
// Thread-safety: fully thread-safe, but keep in mind that talloc is not (so
// talloc_ctx must be owned by the current thread).
struct mp_log *mp_log_new(void *talloc_ctx, struct mp_log *parent,
const char *name)
{
assert(parent);
struct mp_log *log = talloc_zero(talloc_ctx, struct mp_log);
if (!parent->root)
return log; // same as null_log
talloc_set_destructor(log, destroy_log);
log->root = parent->root;
log->partial = talloc_strdup(NULL, "");
log->max_level = MSGL_MAX;
if (name) {
if (name[0] == '!') {
name = &name[1];
} else if (name[0] == '/') {
name = &name[1];
log->prefix = talloc_strdup(log, name);
} else {
log->prefix = parent->prefix
? talloc_asprintf(log, "%s/%s", parent->prefix, name)
: talloc_strdup(log, name);
}
log->verbose_prefix = parent->prefix
? talloc_asprintf(log, "%s/%s", parent->prefix, name)
: talloc_strdup(log, name);
if (log->prefix && !log->prefix[0])
log->prefix = NULL;
if (!log->verbose_prefix[0])
log->verbose_prefix = "global";
} else {
log->prefix = talloc_strdup(log, parent->prefix);
log->verbose_prefix = talloc_strdup(log, parent->verbose_prefix);
}
return log;
}
void mp_msg_init(struct mpv_global *global)
{
assert(!global->log);
struct mp_log_root *root = talloc_zero(NULL, struct mp_log_root);
*root = (struct mp_log_root){
.global = global,
.reload_counter = ATOMIC_VAR_INIT(1),
};
pthread_mutex_init(&root->lock, NULL);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_init(&root->log_file_lock, NULL);
pthread_cond_init(&root->log_file_wakeup, NULL);
struct mp_log dummy = { .root = root };
struct mp_log *log = mp_log_new(root, &dummy, "");
2013-12-21 20:40:45 +00:00
global->log = log;
}
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
static void *log_file_thread(void *p)
{
struct mp_log_root *root = p;
mpthread_set_name("log-file");
pthread_mutex_lock(&root->log_file_lock);
while (root->log_file_thread_active) {
struct mp_log_buffer_entry *e =
mp_msg_log_buffer_read(root->log_file_buffer);
if (e) {
pthread_mutex_unlock(&root->log_file_lock);
fprintf(root->log_file, "[%8.3f][%c][%s] %s",
(mp_time_us() - MP_START_TIME) / 1e6,
mp_log_levels[e->level][0], e->prefix, e->text);
fflush(root->log_file);
pthread_mutex_lock(&root->log_file_lock);
talloc_free(e);
// Multiple threads might be blocked if the log buffer was full.
pthread_cond_broadcast(&root->log_file_wakeup);
} else {
pthread_cond_wait(&root->log_file_wakeup, &root->log_file_lock);
}
}
pthread_mutex_unlock(&root->log_file_lock);
return NULL;
}
static void wakeup_log_file(void *p)
{
struct mp_log_root *root = p;
pthread_mutex_lock(&root->log_file_lock);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_cond_broadcast(&root->log_file_wakeup);
pthread_mutex_unlock(&root->log_file_lock);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
}
// Only to be called from the main thread.
static void terminate_log_file_thread(struct mp_log_root *root)
{
bool wait_terminate = false;
pthread_mutex_lock(&root->log_file_lock);
if (root->log_file_thread_active) {
root->log_file_thread_active = false;
pthread_cond_broadcast(&root->log_file_wakeup);
wait_terminate = true;
}
pthread_mutex_unlock(&root->log_file_lock);
if (wait_terminate)
pthread_join(root->log_file_thread, NULL);
mp_msg_log_buffer_destroy(root->log_file_buffer);
root->log_file_buffer = NULL;
if (root->log_file)
fclose(root->log_file);
root->log_file = NULL;
}
// If opt is different from *current_path, update *current_path and return true.
// No lock must be held; passed values must be accessible without.
static bool check_new_path(struct mpv_global *global, char *opt,
char **current_path)
{
void *tmp = talloc_new(NULL);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
bool res = false;
char *new_path = mp_get_user_path(tmp, global, opt);
if (!new_path)
new_path = "";
char *old_path = *current_path ? *current_path : "";
if (strcmp(old_path, new_path) != 0) {
talloc_free(*current_path);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
*current_path = NULL;
if (new_path && new_path[0])
*current_path = talloc_strdup(NULL, new_path);
res = true;
}
talloc_free(tmp);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
return res;
}
void mp_msg_update_msglevels(struct mpv_global *global, struct MPOpts *opts)
{
struct mp_log_root *root = global->log->root;
2013-12-21 22:11:12 +00:00
pthread_mutex_lock(&root->lock);
2013-12-21 22:11:12 +00:00
root->verbose = opts->verbose;
root->really_quiet = opts->msg_really_quiet;
2013-12-21 22:11:12 +00:00
root->module = opts->msg_module;
root->use_terminal = opts->use_terminal;
root->show_time = opts->msg_time;
if (root->use_terminal)
root->color = opts->msg_color && isatty(STDOUT_FILENO);
2013-12-21 22:11:12 +00:00
m_option_type_msglevels.free(&root->msg_levels);
m_option_type_msglevels.copy(NULL, &root->msg_levels, &opts->msg_levels);
2013-12-21 22:11:12 +00:00
atomic_fetch_add(&root->reload_counter, 1);
pthread_mutex_unlock(&root->lock);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
if (check_new_path(global, opts->log_file, &root->log_path)) {
terminate_log_file_thread(root);
if (root->log_path) {
root->log_file = fopen(root->log_path, "wb");
if (root->log_file) {
root->log_file_buffer =
mp_msg_log_buffer_new(global, 100, MP_LOG_BUFFER_MSGL_LOGFILE,
wakeup_log_file, root);
root->log_file_thread_active = true;
if (pthread_create(&root->log_file_thread, NULL, log_file_thread,
root))
{
root->log_file_thread_active = false;
terminate_log_file_thread(root);
}
} else {
mp_err(global->log, "Failed to open log file '%s'\n",
root->log_path);
}
}
}
if (check_new_path(global, opts->dump_stats, &root->stats_path)) {
bool open_error = false;
pthread_mutex_lock(&root->lock);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
if (root->stats_file)
fclose(root->stats_file);
root->stats_file = NULL;
if (root->stats_path) {
root->stats_file = fopen(root->stats_path, "wb");
open_error = !root->stats_file;
}
pthread_mutex_unlock(&root->lock);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
if (open_error) {
mp_err(global->log, "Failed to open stats file '%s'\n",
root->stats_path);
}
}
}
2013-12-21 22:11:12 +00:00
void mp_msg_force_stderr(struct mpv_global *global, bool force_stderr)
{
struct mp_log_root *root = global->log->root;
pthread_mutex_lock(&root->lock);
2013-12-21 22:11:12 +00:00
root->force_stderr = force_stderr;
pthread_mutex_unlock(&root->lock);
}
// Only to be called from the main thread.
bool mp_msg_has_log_file(struct mpv_global *global)
{
struct mp_log_root *root = global->log->root;
return !!root->log_file;
2013-12-21 22:11:12 +00:00
}
void mp_msg_uninit(struct mpv_global *global)
{
struct mp_log_root *root = global->log->root;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
terminate_log_file_thread(root);
mp_msg_log_buffer_destroy(root->early_buffer);
assert(root->num_buffers == 0);
if (root->stats_file)
fclose(root->stats_file);
talloc_free(root->stats_path);
talloc_free(root->log_path);
m_option_type_msglevels.free(&root->msg_levels);
pthread_mutex_destroy(&root->lock);
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_destroy(&root->log_file_lock);
pthread_cond_destroy(&root->log_file_wakeup);
talloc_free(root);
global->log = NULL;
}
void mp_msg_set_early_logging(struct mpv_global *global, bool enable)
{
struct mp_log_root *root = global->log->root;
pthread_mutex_lock(&root->lock);
if (enable != !!root->early_buffer) {
if (enable) {
pthread_mutex_unlock(&root->lock);
struct mp_log_buffer *buf =
mp_msg_log_buffer_new(global, TERM_BUF, MP_LOG_BUFFER_MSGL_TERM,
NULL, NULL);
pthread_mutex_lock(&root->lock);
assert(!root->early_buffer); // no concurrent calls to this function
root->early_buffer = buf;
} else {
struct mp_log_buffer *buf = root->early_buffer;
root->early_buffer = NULL;
pthread_mutex_unlock(&root->lock);
mp_msg_log_buffer_destroy(buf);
return;
}
}
pthread_mutex_unlock(&root->lock);
}
struct mp_log_buffer *mp_msg_log_buffer_new(struct mpv_global *global,
int size, int level,
void (*wakeup_cb)(void *ctx),
void *wakeup_cb_ctx)
{
struct mp_log_root *root = global->log->root;
pthread_mutex_lock(&root->lock);
if (level == MP_LOG_BUFFER_MSGL_TERM) {
size = TERM_BUF;
// The first thing which creates a terminal-level log buffer gets the
// early log buffer, if it exists. This is supposed to enable a script
// to grab log messages from before it was initialized. It's OK that
// this works only for 1 script and only once.
if (root->early_buffer) {
struct mp_log_buffer *buffer = root->early_buffer;
root->early_buffer = NULL;
buffer->wakeup_cb = wakeup_cb;
buffer->wakeup_cb_ctx = wakeup_cb_ctx;
pthread_mutex_unlock(&root->lock);
return buffer;
}
}
assert(size > 0);
struct mp_log_buffer *buffer = talloc_ptrtype(NULL, buffer);
*buffer = (struct mp_log_buffer) {
.root = root,
.level = level,
.entries = talloc_array(buffer, struct mp_log_buffer_entry *, size),
.capacity = size,
.wakeup_cb = wakeup_cb,
.wakeup_cb_ctx = wakeup_cb_ctx,
};
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_init(&buffer->lock, NULL);
MP_TARRAY_APPEND(root, root->buffers, root->num_buffers, buffer);
atomic_fetch_add(&root->reload_counter, 1);
pthread_mutex_unlock(&root->lock);
return buffer;
}
void mp_msg_log_buffer_set_silent(struct mp_log_buffer *buffer, bool silent)
{
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_lock(&buffer->lock);
buffer->silent = silent;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_unlock(&buffer->lock);
}
void mp_msg_log_buffer_destroy(struct mp_log_buffer *buffer)
{
if (!buffer)
return;
struct mp_log_root *root = buffer->root;
pthread_mutex_lock(&root->lock);
for (int n = 0; n < root->num_buffers; n++) {
if (root->buffers[n] == buffer) {
MP_TARRAY_REMOVE_AT(root->buffers, root->num_buffers, n);
goto found;
}
}
abort();
found:
while (buffer->num_entries)
talloc_free(log_buffer_read(buffer));
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_destroy(&buffer->lock);
talloc_free(buffer);
atomic_fetch_add(&root->reload_counter, 1);
pthread_mutex_unlock(&root->lock);
}
// Return a queued message, or if the buffer is empty, NULL.
// Thread-safety: one buffer can be read by a single thread only.
struct mp_log_buffer_entry *mp_msg_log_buffer_read(struct mp_log_buffer *buffer)
{
struct mp_log_buffer_entry *res = NULL;
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_lock(&buffer->lock);
if (!buffer->silent && buffer->num_entries) {
if (buffer->dropped) {
res = talloc_ptrtype(NULL, res);
*res = (struct mp_log_buffer_entry) {
.prefix = "overflow",
.level = MSGL_FATAL,
.text = talloc_asprintf(res,
"log message buffer overflow: %"PRId64" messages skipped\n",
buffer->dropped),
};
buffer->dropped = 0;
} else {
res = log_buffer_read(buffer);
}
}
msg: make --log-file buffered through a thread Until now --log-file performed a blocking write to the log file, which made any calling thread block for I/O. It even explicitly flushed after every line (to make it tail-able, or to ensure a hard crash wouldn't lose any of the output). This wasn't so good, because it could cause real playback problems, which made it infeasible to enable it by default. Try to buffer it through a ring buffer and a thread. There's no other choice but to use a thread, since async I/O on files is generally a big and unportable pain. (We very much prefer portable pain.) Fortunately, there's already a ring buffer (mp_log_buffer, normally for the client API logging hook). This still involves some pretty messy locking. Give each mp_log_buffer its own lock to make this easier. This still makes calling threads block if the log buffer is full (unlike with client API log buffers, which just drop messages). I don't want log messages to get lost for this purpose. This also made locking pretty complicated (without it, mp_log_buffer wouldn't have needed its own lock). Maybe I'll remove this blocking again when it turns out to be nonsense. (We could avoid wasting an entire thread by "reusing" some other thread. E.g. pick some otherwise not real time thread, and make it react to the log buffer's wakeup callback. But let's not. It's complicated to abuse random threads for this. It'd also raise locking complexity, because we still want it to block on a full buffer.)
2020-01-29 22:34:59 +00:00
pthread_mutex_unlock(&buffer->lock);
return res;
}
// Thread-safety: fully thread-safe, but keep in mind that the lifetime of
// log must be guaranteed during the call.
// Never call this from signal handlers.
void mp_msg(struct mp_log *log, int lev, const char *format, ...)
{
va_list va;
va_start(va, format);
mp_msg_va(log, lev, format, va);
va_end(va);
}
const char *const mp_log_levels[MSGL_MAX + 1] = {
[MSGL_FATAL] = "fatal",
[MSGL_ERR] = "error",
[MSGL_WARN] = "warn",
[MSGL_INFO] = "info",
[MSGL_STATUS] = "status",
[MSGL_V] = "v",
[MSGL_DEBUG] = "debug",
[MSGL_TRACE] = "trace",
[MSGL_STATS] = "stats",
};
const int mp_mpv_log_levels[MSGL_MAX + 1] = {
[MSGL_FATAL] = MPV_LOG_LEVEL_FATAL,
[MSGL_ERR] = MPV_LOG_LEVEL_ERROR,
[MSGL_WARN] = MPV_LOG_LEVEL_WARN,
[MSGL_INFO] = MPV_LOG_LEVEL_INFO,
[MSGL_STATUS] = 0, // never used
[MSGL_V] = MPV_LOG_LEVEL_V,
[MSGL_DEBUG] = MPV_LOG_LEVEL_DEBUG,
[MSGL_TRACE] = MPV_LOG_LEVEL_TRACE,
[MSGL_STATS] = 0, // never used
};
int mp_msg_find_level(const char *s)
{
2014-01-16 20:34:47 +00:00
for (int n = 0; n < MP_ARRAY_SIZE(mp_log_levels); n++) {
if (mp_log_levels[n] && !strcmp(s, mp_log_levels[n]))
return n;
}
return -1;
}