The pcmbuffer wasn't reset after subsequent seeks (i.e. all but the first one) when seeking while paused. This caused the buffer to be filled only once and so the wrong sound was played upon resuming. Now we make sure the pcmbuffer is always reset when not playing (a more detailed explaination is in the task).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21570 a1c6a512-1295-4272-9138-f99709370657
Prevents playing the remainder of the pcmbuf on manual stop, as it isn't needed in this case. Now the remainder is only played when the music automatically plays out.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21569 a1c6a512-1295-4272-9138-f99709370657
Ensure that the PCM buffer is flushed at the end of the last track in a
playlist.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21316 a1c6a512-1295-4272-9138-f99709370657
* Use events to notify things when the track has changed instead of the nasty has_track_changed()
* Event for when the mp3entry for the next track is avilable (which allows alot more tags to be static which means less redrawing in the WPS)
* virtually guarentee that the mp3entry sturct returned by audio_current/next_track() is going to be valid for the duration of the current track. The only time it wont be now is during the time between the codec finishing the previous track and the next track actually starting (~2s), but this is not an issue as long as it is called again when the TRACK_CHANGED event happens (or just use the pointer that gives)
It is still possible to confuse the WPS with the next tracks id3 info being displayed but this should fix itself up faster than it used to (and be harder to do)
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@20633 a1c6a512-1295-4272-9138-f99709370657
This also appears to fix the constant boosting part of FS#8999 (at least I couldn't reproduce). It does certainly not fix that buffering is stucked.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@20093 a1c6a512-1295-4272-9138-f99709370657
Use a smaller PCM buffer on targets with 2MB or less ram.
(FS#9703)
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@19743 a1c6a512-1295-4272-9138-f99709370657
This should be a good first step to allow multi-driver targets, like the Elio (ATA/SD), or the D2 (NAND/SD).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@18960 a1c6a512-1295-4272-9138-f99709370657
Based on a patch by Tomasz Malesinski
* Merge Codec buffer and Malloc buffer into one large buffer.
* The new merged buffer is now 1MB on targets with lots of memory.
* Renamed codec_get_memory to codec_get_buffer and made it behave more.
like plugin_get_buffer.
* Bumped Codec api and min api versions.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@18834 a1c6a512-1295-4272-9138-f99709370657
* Remove gui_splash()
* Rename gui_syncsplash() to splashf() and remove its voice
capabilities.
* Rename the internal splash() to splash_internal() and introduce an
externally visible splash() that handles simple splashing without
printf functionality e.g. splash(HZ, ID2P(LANG_FOO)); or splash(HZ,
"foo"); if a LANG_* id is passed it will be voiced.
* Adjust all places that called gui_syncsplash() to use the correct
variant from above.
* Export both new functions to plugins and adjust places calling
rb->splash() to use the correct variant so that we now have naming
consistency between the core and plugins.
* Fix one latent bug that would cause my sim to crash with the above
changes and correct P2STR and P2ID macros, thanks to pondlife.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@18282 a1c6a512-1295-4272-9138-f99709370657
-Allows loading multiple thumbnails back-to-back in the one thumbnail buffer.
-Mutex to prevent race conditions with talk queue indices and
thumbnail buffer state.
-Synchronous shutup.
-Shutup is a noop if no voice is queued.
-mp3_play_stop() does nothing until the audio thread is ready.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@18046 a1c6a512-1295-4272-9138-f99709370657
The issue happened when the value of 'filling' was STATE_FINISHED: the low_buffer_callback wouldn't do anything even when there would still be data remaining to be buffered, leading to a shortage of audio data. Only making the callback act even when filling is STATE_FINISHED (the simple fix) isn't right because of cases when the last track in the playlist is fully buffered. Therefore I added a new state to distinguish between when the last track is fully buffered (STATE_FINISHED) and when it isn't (STATE_END_OF_PLAYLIST).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@17875 a1c6a512-1295-4272-9138-f99709370657
