If the song tag comes from a stream, and MPD playback restarts, MPD
would believe the tag should override the newly received tag. This
makes the previous tag appear stuck. This change passes the song tag
only if it's authoritative - i.e. if it's a song file.
Right after booting, the monotonic clock starts with a very small
value, and AudioOutput::LockUpdate() may believe that the fail_timer
has not recovered yet.
Ask FFmpeg to seek to the next packet boundary *before* the seek
position, so we don't miss audio data. Now we get too much, but we'll
solve that in the next commit.
The Connect method can be called between Schedule and lock. In that case, when
locked, the state is already set to CONNECTING of READY and the condition won't
be signaled anymore.
Not initialising granulepos leads to it having arbitrary values in the
encoded stream including possibly negative values which are not valid
and confuse opusdec. Explicitly initialise opus_encoder::granulepos
to avoid that problem.
The file handle is never reset to INVALID_HANDLE_VALUE, and thus the
destructor will assume the operation shall be cancelled and will
delete the temporary file.
This was a major breakage for saving the database file and the state
file.
Build a table of pre-existing tag types before adding new items. The
old way would check HasType() each time, which would return true after
the first instance of that tag type had been added, preventing
duplicate tag types to be merged.
This broke duplicate tag types loaded from the state file, because
this code path uses TagBuilder::Complement().
This is Darwin specific: the previous implementation was causing an integer
overflow when base.numer is very large. On PPC Darwin, the timebase info is 1000000000/33330116 and this is too large for integer arithmetic.
Fixes a problem with the "curl" input plugin: IsEOF() always returns
true because the "open" flag was cleared by
CurlInputStream::RequestDone() when end-of-stream was reached. This
flag stays false even when seeking to another position has succeeded.
This patch resets the "open" flag to true after seeking successfully.
NetBSD's pthread_setname_np() prototype is incompatible with the rest
of the world, and it requires to pass the string argument as a
non-const pointer. Instead of working around this misdesign, I hereby
disable the feature on NetBSD.
Add macro HAVE_THREAD_NAME which is set when any method to set the
thread name is available. Use that macro in FormatThreadName()
instead of just checking for HAVE_PTHREAD_SETNAME_NP.
Here's a change to dynamically allocate the DSD ID3 tag buffer.
Pretty much anything with cover art is going to exceed the existing,
static 4k limit... Here's a change to dynamically allocate the buffer
and sanity check it at some upper limit. I rather arbitrarily pulled
256k out of thin air just to keep a corrupt file from causing it to
trying to allocate a buffer larger than available memory.
When mounting had not yet finished, SocketMonitor::IsDefined() was
always false, due to the workaround at the beginning of the function
that calls SocketMonitor::Steal(). This commit drops the IsDefined()
check because it was never necessary and breaks reconnect.
nfs_destroy_context() will invoke all pending callbacks with
err==-EINTR. In CancellableCallback::Callback(), this will invoke
NfsConnection::DeferClose(), which however is only designed to be
called from nfs_service(). In non-debug mode, this will leak memory
because nfs_close_async() is never called.
Workaround: before nfs_destroy_context(), invoke nfs_close_async() on
all pending file handles.
The method NfsConnection::CancellableCallback::Callback() will always
invoke NfsConnection::Close() on the file handle, even if the void
pointer is not a nfsfh. This can happen if the Open() was not
successful, e.g. when the file does not exist.
Skipping those songs silently will confuse the client, because
commands specifying the song index within a playlist
(e.g. playlistdelete) will be out of sync.
This copies spl_print()'s behavior to playlist_file_print().
Version 2.5 fixed an API oddity, however it broke API compatibility,
at least with C++. Disable the workaround when a libavformat version
is detected that is recent enough.
The "::" to explicitly refer to the global namespace appeared like a
good idea in C++, but it breaks with C libraries that implement
standard functions using macros (e.g. musl).
There are code paths (mostly error cases) in which it is possible to
initialize an AudioOutput and then kill it without ever calling
audio_output_new(). In such a case, its destructor will attempt to
free a mixer that was never initialized, leading to an attempt to
take out a lock on a mutex that was similarly never initialized,
which hangs forever.
Fix by always initializing the mixer appropriately.
Clean up the "state" to indicate that there is no longer any
asynchronous operation. Fixes another NFS-related crash due to
cleanup of a non-existing asynchronous operation.
During the NfsLease::OnNfsConnectionFailed() call, the old (defunct)
nfs_context may be used to close file handles. Such code does not yet
exist, but will be added soon to fix other bugs.
libmp4v2 is licensed under MPL 1.1, which is incompatible with GPLv2.
Unfortunately, this means that we must remove the plugin.
More information can be found in the Debian bug report:
http://bugs.debian.org/767504
Commit d42c0f1dc5 added an OS X-specific
method of calling mpd_main_after_fork(), which uses Grand Central
Dispatch. Since this uses a block literal, it breaks compilation on
compilers which don't support the block extension, e.g. non-Apple
compilers. This affects users on older OS X releases with GCD (which
depend on older Clang releases, or Apple GCCs, which don't support the
C++11 features MPD needs); or which don't support GCD at all (10.5 and
lower).
This patch changes the #ifdef so that the non-GCD code is used
as it was on OS X before this patch if blocks aren't available, via
checking __BLOCKS__ macro.
The old formula calculates the output buffer size with "regular"
rounding (to the nearest integer), however sometimes, that is
insufficient and the last sample cannot be resampled. This causes
audible distortions. By changing the formula to consider the worst
case (always round up), this problem is eliminated.
While seeking, metadata must not be updated. ResponseBoundary() was
added in MPD 0.19.1, but I forgot to add the IsSeeking() check there.
This caused the "seekable" flag to reset.
On "list albumartist", songs that have no AlbumArtist tag will use the
Artist tag. However, if AlbumArtist is disabled via
"metadata_to_use", the TagBuilder::AddItem() call is ignored, and
PrintUniqueTag() attempts to print a nullptr string.
This commit fixes the problem by attempting the fallback only if
AlbumArtist is not disabled.
When uri_apply_base() was moved from db/upnp/Util.cpp to
util/UriUtil.cpp, the parameter order was changed, however without
swapping the parameters in the ContentDirectoryService constructor.
Many years ago, FAAD had a serious ABI bug: the NeAACDecInit()
prototype in its header declared the "samplerate" parameter to be
"unsigned long *", but internally, the function assumed it was
"uint32_t *" instead. On 32 bit machines, that was no difference, but
on 64 bit, this left one portion of the return value uninitialized;
and worse, on big-endian, the wrong word was filled. This bug had to
be worked around in MPD (commit 9c4e97a6).
A few months later, the bug was fixed in the FAAD CVS in commit 1.117
on file libfaad/decoder.c; the commit message was:
"Use public headers internally to prevent duplicate declarations"
The commit message was too brief at best; the problem was not
duplicate declarations, but a prototype mismatch. No mention of the
bug fix in the ChangeLog.
The MPD project never learned about this bug fix, and so MPD would
always pass a "uin32_t *" dressed up as a "unsigned long *". Nearly 6
years later, it's about time to fix this second ABI problem. Let's
kill the workaround!
Many years ago, FAAD had a serious ABI bug: the NeAACDecInit()
prototype in its header declared the "samplerate" parameter to be
"unsigned long *", but internally, the function assumed it was
"uint32_t *" instead. On 32 bit machines, that was no difference, but
on 64 bit, this left one portion of the return value uninitialized;
and worse, on big-endian, the wrong word was filled. This bug had to
be worked around in MPD (commit 9c4e97a6).
A few months later, the bug was fixed in the FAAD CVS in commit 1.117
on file libfaad/decoder.c; the commit message was:
"Use public headers internally to prevent duplicate declarations"
The commit message was too brief at best; the problem was not
duplicate declarations, but a prototype mismatch. No mention of the
bug fix in the ChangeLog.
The MPD project never learned about this bug fix, and so MPD would
always pass a "uin32_t *" dressed up as a "unsigned long *". Nearly 6
years later, it's about time to fix this second ABI problem. Let's
kill the workaround!
MPD handles all strings in UTF-8 internally. Those decoders which
read Latin-1 tags are supposed to implement the conversion, instead of
passing Latin-1 to TagBuilder::AddItem(). FixTagString() is simply
the wrong place to do that, and hard-coding Latin-1 is kind of
arbitrary.
Read all directory entries into memory and close the struct nfsdir
before returning the StorageDirectoryReader instance. This is what
libnfs does, anyway.
Creating a NfsStorage sets its own export_name as the "base". Now
NfsFileReader can use this information to derive the export_name to be
mounted, instead of guessing. This solves the "too many connection"
problem on the NFS server while updating the database.
Make the "open" method of plugins "file" and "archive" dummy methods
that always fail. Instead, let InputStream::Open() hard-code access
to these two plugins by using OpenLocalInputStream(). This allows
simplifyin the algorithm for falling back to probing archive plugins.
If an async opertion is in progress, nfs_close_async() will make
libnfs crash because the RPC callback will dereference an object that
was freed by nfs_close_async().
