- don't close and reopen an audioOutput when it has a fixed output format, and closing and reopening the device is unneccessary when the input audio format changes
git-svn-id: https://svn.musicpd.org/mpd/trunk@4908 09075e82-0dd4-0310-85a5-a0d7c8717e4f
This patch should continue to allow mpd to play as well as
possible to icecast servers while avoiding stalls on local
devices. This has eliminated ALSA underrun errors
for me while streaming to a remote host while the network
connection was bad.
Of course, this makes opening a connection non-blocking, too,
so myShout_openShoutConn is a bit more complex.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4898 09075e82-0dd4-0310-85a5-a0d7c8717e4f
Make the unit tests compile correctly without using xstrdup.
Also, use "static inline" instead of "inline static": certain
compilers or cflags are likely to complain about the latter.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4892 09075e82-0dd4-0310-85a5-a0d7c8717e4f
Warren's fix in r4872 made phpMp work again, but also broke
the unit tests completely (they work in this version).
The version in 0.12.0 is far too buggy (it was from mpd-ke, what
do you expect?). This one passes all the unit tests that the
mpd-ke one passed, and should also work with phpMp when used
with PHP magic quotes.
This also means we can search on 100 (or more) tags at once, so
no more arbitrary limits other than system memory.
To run the unit tests, just do this:
gcc -o t -DUNIT_TEST=1 src/buffer2array.c && ./t && echo OK
git-svn-id: https://svn.musicpd.org/mpd/trunk@4874 09075e82-0dd4-0310-85a5-a0d7c8717e4f
I'm still not entirely certain why we index cb->metaChunkSet[]
with currentChunk (and not currentMetaChunk), but shank told me
that currentChunk is correct...
git-svn-id: https://svn.musicpd.org/mpd/trunk@4814 09075e82-0dd4-0310-85a5-a0d7c8717e4f
(based on suggested patch by Jan-Benedict Glaw):
> While hacking mpd, I noticed that an assert()ion in xrealloc is wrong.
> A null size is perfectly legal, so we shouldn't assert on that.
Since some C libraries return NULL when size == 0, we'll make
sure we get a free()-able pointer since some of those C
libraries also barf on free(NULL).
git-svn-id: https://svn.musicpd.org/mpd/trunk@4740 09075e82-0dd4-0310-85a5-a0d7c8717e4f
I'm checking for zero-size allocations and assert()-ing them,
so we can more easily get backtraces and debug problems, but we'll
also allow -DNDEBUG people to live on the edge if they wish.
We do not rely on errno when checking for OOM errors because
some implementations of malloc do not set it, and malloc
is commonly overridden by userspace wrappers.
I've spent some time looking through the source and didn't find any
obvious places where we would explicitly allocate 0 bytes, so we
shouldn't trip any of those assertions.
We also avoid allocating zero bytes because C libraries don't
handle this consistently (some return NULL, some not); and it's
dangerous either way.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4690 09075e82-0dd4-0310-85a5-a0d7c8717e4f
leave out initCommands to keep jat happy, and keep labels
at the left hand side
git-svn-id: https://svn.musicpd.org/mpd/trunk@4687 09075e82-0dd4-0310-85a5-a0d7c8717e4f
This finally fixes a bug from over two years ago playing a wave file
(oprah.wav) with the following characteristics (from sfinfo):
File Format Microsoft RIFF WAVE Format (wave)
Data Format 8-bit integer (unsigned, little endian)
Audio Data 986827 bytes begins at offset 58 (3a hex)
1 channel, 986827 frames
Sampling Rate 22050.00 Hz
Duration 44.754 seconds
Of course, this has been regression tested with all the files
that the previous commit got working. Thanks to Michael Pruett
(audiofile author) for the hint and shame on me for forgetting
about it for over two years :x
git-svn-id: https://svn.musicpd.org/mpd/trunk@4682 09075e82-0dd4-0310-85a5-a0d7c8717e4f
Use the 'Virtual' variants of afGetSampleFormat, afGetChannels,
afGetVirtualFrameSize in the audiofile library, since it already does
the necessary abstraction for us.
Of course, I've regression tested these changes against my
standard 44100Hz/2ch/16bit wave files and they continue to play
fine.
Files tested:
english.au (Linus Torvalds pronouncing 'Linux' in English)
B01.Red_Bright_Heart.au (Chinese opera, sounds correct to me even though
I don't actually understand the words)
git-svn-id: https://svn.musicpd.org/mpd/trunk@4681 09075e82-0dd4-0310-85a5-a0d7c8717e4f
This bug was NOT introduced in my OggFLAC additions, honest!
As far as I can see, it was introduced way back in r2482, but
nobody ever noticed until the post here:
http://www.musicpd.org/forum/index.php?topic=1152.0
While we're at it, clean up some of the variable typing.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4664 09075e82-0dd4-0310-85a5-a0d7c8717e4f
I'm not using __FUNCTION__ or __func__ because compiler support
for these is still a bit iffy as far as I know...
git-svn-id: https://svn.musicpd.org/mpd/trunk@4662 09075e82-0dd4-0310-85a5-a0d7c8717e4f
Unfortunately there doesn't seem to be an indent switch for this,
but we have find + perl:
find src -name '*.[ch]' | xargs perl -i -p -e \
's/^\s+(\w+):/$1:/g unless /^\s+default:/'
This is a followup to r4605, and there are no actual code
changes in this.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4661 09075e82-0dd4-0310-85a5-a0d7c8717e4f
Add -Wmissing-prototypes if compiling with gcc
Static where possible
git-svn-id: https://svn.musicpd.org/mpd/trunk@4657 09075e82-0dd4-0310-85a5-a0d7c8717e4f
size_t is bigger than int on most 64-bit machines, so cast
size_t to long when passing them to printf-like functions.
Ideally we'd use %z, but many compilers don't support it.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4656 09075e82-0dd4-0310-85a5-a0d7c8717e4f
First, make sure we call finishPlaylist() before
closeMp3Directory() since the latter will free non-SONG_TYPE_URL
songs in playlist, which causes an invalid read when we try to
look for SONG_TYPE_URL songs to free in finishPlaylist.
Secondly, make sure our children have all exited before freeing
the playerData. If we do not, slowly-delivered signals can
trigger a race condition in the signal handlers of the decode
and player processes which rely on getPlayerData. To avoid
waitpid-ing too long (or at all), move the freePlayerData() call
farther down in main() (this won't affect anything else)
to give the OS a better chance to deliver signals and finish running
sig handlers for terminated children.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4640 09075e82-0dd4-0310-85a5-a0d7c8717e4f
utf8ToFsCharset() and fsCharsetToUtf8() got very broken in r4311, and
resulted in several commits to fix those leaks. Unfortunately, not all
of those newly introduced leaks were fixed, nor was the result pretty.
Also, fixed a double-free in lsPlaylists(). This is very hard
to trigger (and therefore exploit) at the moment because we
check printDirectoryInfo() beforehand.
Intended behavior for utf8ToFsCharset() and fsCharsetToUtf8() as
God^H^H^Hshank originally intended is now documented in path.h
to prevent future errors like this.
mpd could still use some good valgrind testing before the 0.12.0
release.
<plug>In addition to reducing heap fragmentation, malloc
reductions from mpd-ke greatly reduces the chance of leaks from
happening due to programming errors.</plug>
git-svn-id: https://svn.musicpd.org/mpd/trunk@4639 09075e82-0dd4-0310-85a5-a0d7c8717e4f
Passing a ref to snd_pcm_hw_params_set_{buffer,period}_time_near
can modify our internal {period,buffer}_time members inside the
AlsaData structure, making re-initializing the device across
sample/bit rate and channel changes non-idempotent.
git-svn-id: https://svn.musicpd.org/mpd/trunk@4616 09075e82-0dd4-0310-85a5-a0d7c8717e4f
