aboutsummaryrefslogtreecommitdiffstats
path: root/src (follow)
Commit message (Collapse)AuthorAgeFilesLines
* input_plugin: add method check()Max Kellermann2011-09-166-0/+72
| | | | | To check for errors without reading. The decoder thread wants to do that, before it passes the input stream to the plugin.
* input/{soup,curl}: free unused postponed_errorMax Kellermann2011-09-162-0/+8
| | | | Fix memory leak.
* input/soup: move libsoup calls to the I/O threadMax Kellermann2011-09-161-4/+25
| | | | | libsoup's asynchronous API is not thread safe. By moving the calls into the I/O thread, several crash bugs will be fixed.
* input/soup: wait for _cancel_message() to completeMax Kellermann2011-09-161-5/+21
| | | | | Add the "complete" attribute, and set it in _session_callback(). _close() waits for it to become true.
* input/soup: return real GErrors to the callerMax Kellermann2011-09-161-3/+31
| | | | | Add attribute postponed_error, pass this GError to the caller.
* input/soup: set the "ready" flag on failureMax Kellermann2011-09-161-0/+2
| | | | Fix deadlock.
* input/soup: remove redundant flag "ready"Max Kellermann2011-09-161-6/+2
| | | | | We'll add proper locking now, and there's no need to carry this duplicate flag.
* input/soup: report failure in _read()Max Kellermann2011-09-161-0/+1
| | | | To report failures to callers who don't use _buffer().
* input/soup: unlock before cancelling the soup messageMax Kellermann2011-09-161-3/+5
| | | | Avoid recursive deadlock in _session_callback().
* decoder_api: remove the loop from decoder_read()Max Kellermann2011-09-161-16/+10
| | | | | This was useless, because input_stream_read() blocks until data is available. Calling it in a loop doesn't make sense.
* decoder_api: move code to decoder_check_cancel_read()Max Kellermann2011-09-161-12/+28
|
* input/curl: wait some more before resuming the streamMax Kellermann2011-09-161-1/+6
| | | | | Pausing and resuming after every little chunk adds lots of overhead. Add a lower level for resuming the stream.
* input/curl: clear the "paused" flag before resumingMax Kellermann2011-09-161-1/+1
| | | | | curl_easy_pause(CURLPAUSE_CONT) may have added enough data to go into pause again.
* input/curl: remove _schedule_update()Max Kellermann2011-09-161-48/+1
| | | | | Call _update_fds() directly. This is possible because it's only called from within the I/O thread.
* input/curl: unlock the mutex for io_thread_call()Max Kellermann2011-09-161-1/+4
| | | | Fix deadlock.
* input/soup: fix "unused local variable" warningsMax Kellermann2011-09-161-1/+1
|
* Merge branch 'v0.16.x'Max Kellermann2011-09-162-2/+3
|\
| * input/curl, output/pulse: fix "unused local variable" warningsMax Kellermann2011-09-162-2/+3
| |
* | input_stream: add "nonnull" attributesMax Kellermann2011-09-151-0/+12
| |
* | Merge branch 'v0.16.x'Max Kellermann2011-09-153-11/+35
|\| | | | | | | | | | | Conflicts: NEWS configure.ac
| * decoder/ffmpeg: flush the codec after seekingMax Kellermann2011-09-151-1/+3
| | | | | | | | | | Let the codec start with fresh buffers. This should fix the remaining seeking issues.
| * decoder/ffmpeg: explicitly specify the current stream for seekingMax Kellermann2011-09-151-2/+11
| | | | | | | | | | Use AVStream.time_base to convert the decoder_seek_where() value, and pass the current stream number to av_seek_frame().
| * decoder/ffmpeg: don't require key frame for seekingMax Kellermann2011-09-151-1/+2
| | | | | | | | Use flag AV_TIME_BASE.
| * decoder/ffmpeg: higher precision timestampsMax Kellermann2011-09-151-1/+2
| |
| * decoder/ffmpeg: move formula to time_from_ffmpeg()Max Kellermann2011-09-151-2/+10
| |
| * decoder/ffmpeg: add local variable "av_stream"Max Kellermann2011-09-151-3/+4
| | | | | | | | Code simplification.
| * input/rewind: copy the MIME type only onceMax Kellermann2011-09-151-1/+3
| | | | | | | | Reduce heap usage by reducing the number of malloc() / free() calls.
| * pcm_format: fix 32-to-24 bit conversion (the "silence" bug)Max Kellermann2011-09-081-2/+2
| | | | | | | | | | | | D'oh, we were reading 16 bit integers instead of 32 bit integers! That caused silence when trying to play a 32 bit input file on a 24 bit sound card (e.g. USB sound chips with 24 bit packed samples).
* | input/curl: per-request mutex/condMax Kellermann2011-09-151-33/+35
| | | | | | | | | | | | | | The global data structures are now lock-free, because they are accessed only from the I/O thread. By using per-request locks, we have finer grained locking, preparing for locks shared with the client.
* | input/curl: move all libCURL calls to the I/O threadMax Kellermann2011-09-151-29/+69
| | | | | | | | | | | | This adds some overheads for indirect calls to the I/O thread, but reduces the amount of global locks. Next step will be switching to per-request locks.
* | input/file: return NULL instead of "false"Max Kellermann2011-09-152-5/+5
| |
* | input/curl: don't set "ready" on _easy_free()Max Kellermann2011-09-151-3/+2
| | | | | | | | | | Move the assignment to _request_done(), right before the GCond is signalled. This might fix spurious wakeups.
* | input/curl: add assertions on io_thread_inside()Max Kellermann2011-09-151-0/+9
| |
* | input/curl: merge _request_abort() into _request_done()Max Kellermann2011-09-151-31/+19
| | | | | | | | | | This is a trivial function. Merge some duplicate code, e.g. the g_cond_broadcast() call.
* | udp_server: disable -Wmissing-field-initializersMax Kellermann2011-09-151-0/+7
| |
* | gcc.h: add macro GCC_CHECK_VERSIONMax Kellermann2011-09-151-1/+6
| |
* | input/curl: API documentation fixMax Kellermann2011-09-151-1/+1
| |
* | input_plugin: add method update()Max Kellermann2011-09-144-0/+33
| | | | | | | | | | Update the struct attributes, important for facades like the "rewind" plugin. To replace buffer().
* | input_internal: add assertionsMax Kellermann2011-09-141-0/+9
| |
* | input_stream: move input_stream_init(), _deinit() to _internal.cMax Kellermann2011-09-1414-20/+88
| |
* | decoder_internal: don't call input_stream_buffer()Max Kellermann2011-09-143-37/+12
| | | | | | | | | | | | This is not necessary since all relevant input plugins have been moved to the I/O thread, and there is no remaining useful buffer() implementation. This also fixes a busy loop when playing radio.
* | db_print: move code to function db_selection_print()Max Kellermann2011-09-137-132/+27
| | | | | | | | Use it in handle_lsinfo(), and eliminate some duplicate code.
* | db_visitor: add method playlist()Max Kellermann2011-09-133-0/+54
| |
* | db_plugin: add method visit()Max Kellermann2011-09-137-22/+147
| |
* | database: don't allow uri==NULLMax Kellermann2011-09-137-10/+34
| | | | | | | | Add nonnull attributes and fix all callers.
* | gcc.h: add macro gcc_nonnullMax Kellermann2011-09-131-0/+4
| |
* | gcc.h: change "mpd_" prefix to "gcc_"Max Kellermann2011-09-132-19/+19
| | | | | | | | This is specific to gcc, not to mpd.
* | db_plugin: add method get_song()Max Kellermann2011-09-133-2/+39
| | | | | | | | New db_get_song() implementation.
* | database: generate GError when database is disabledMax Kellermann2011-09-133-5/+20
| |
* | directory: don't visit "self" in _walk()Max Kellermann2011-09-132-4/+8
| | | | | | | | Let the caller decide if the current directory should be visited.