FLAC 0.10 released. This is probably the final beta. There have been many improvements in the last two months:
It should be a short hop to 1.0.
- Both the encoder and decoder have been significantly sped up. Aside from C improvements, the code base now has an assembly infrastructure that allows assembly routines for different architectures to be easily integrated. Many key routines have now have faster IA-32 implementations (thanks to Miroslav).
- A new metadata block SEEKTABLE has been defined to hold an arbitrary number of seek points, which speeds up seeking within a stream.
- flac now has a command-line usage similar to 'gzip'; make sure to see the latest documentation for the new usage. It also attempts to preserve the input file's timestamp and permissions.
- The -# options in flac have been tweaked to yield the best compression-to-encode-time ratios. The new default is -5.
- flac can now usually autodetect WAVE files when encoding so that -fw is usually not needed when encoding from stdin.
- The WAVE reader in flac now just ignores (with a warning) unsupported sub-chunks instead of aborting with an error.
- Added an option '--delete-input-file' to flac which automatically deletes the input after a successful encode/decode.
- Added an option '-o' to flac to force the output file name (the old usage of "flac - outputfilename" is no longer supported).
- Changed the XMMS plugin to send smaller chunks of samples (now 512) so that visualization is not slow.
- Fixed a bug in the stream decoder where the decoded samples counter got corrupted after a seek.
FLAC 0.9 released. There were some format changes that broke backwards compatibility but these should be the last (see below). Also, there have been several bug fixes and some new features:
FLAC is on track for an official 1.0 release soon.
- FLAC's sync code has been lengthened to 14 bits from 9 bits. This should enable a faster and more robust synchronization mechanism.
- Two reserved bits were added to the frame header.
- A CRC-16 was added to the FLAC frame footer, and the decoder now does frame integrity checking based on the CRC.
- The format now includes a new subframe field to indicate when a subblock has one or more 0 LSBs for all samples. This increases compression on some kinds of data.
- Added two options to the analysis mode, one for including the residual signal in the analysis file, and one for generating gnuplot files of each subframe's residual distribution with some statistics. See the latest documentation.
- XMMS plugin now supports 8-bit files.
- Fixed a bug in the Winamp2 plugin where the audio sounded garbled.
- Fixed a bug in the Winamp2 plugin where Winamp would hang sporadically at the end of a track (c.f. bug #231197).
FLAC 0.8 released. This release is a result of extensive testing and fixes several bugs encountered when pushing the encoder to the limit. I'm pretty confident in the stability of the encoder/decoder now for all kinds of input. There have also been several features added. Here is a complete list of the changes since 0.7:
FLAC is very close to being ready for an official release. The only known problems left are with the Winamp plugins, which should be fixed soon, and pipes with MSVC.
- Created a new utility called metaflac. It is a metadata editor for .flac files. Right now it just lists the contents of the metadata blocks but eventually it will allow update/insertion/deletion.
- Added two new metadata blocks: PADDING which has an obvious function, and APPLICATION, which is meant to be open to third party applications. See the latest format docs for more info, or the new id registration page.
- Added a -P option to flac to reserve a PADDING block when encoding.
- Added support for 24-bit files to flac (the FLAC format always supported it).
- Started the Winamp3 plugin.
- Greatly expanded the test suite, adding more streams (24-bit streams, noise streams, non-audio streams, more patterns) and more option combinations to the encoder. The test suite runs about 30 streams and over 5000 encodings now.
- Fixed a bug in libFLAC that happened when using an exhaustive LPC coefficient quantization search with 8 bps input.
- Fixed a bug in libFLAC where the error estimation in the fixed predictor could overflow.
- Fixed a bug in libFLAC where LPC was attempted even when the autocorrelation coefficients implied it wouldn't help.
- Reworked the LPC coefficient quantizer, which also fixed another bug that might occur in rare cases.
- Really fixed the '-V overflow' bug (c.f. bug #231976).
- Fixed a bug in flac related to the decode buffer sizing.
FLAC 0.7 released. This is mainly a bug fix release, specifically:
- Fixed a bug that happened when both -fr and --seek were used at the same time.
- Fixed a bug with -p (c.f. bug #230992).
- Fixed a bug that happened when using large (>32K) blocksizes and -V (c.f. bug #231976).
- Fixed a bug where encoder was double-closing a file.
- Expanded the test suite.
- Added more optimization flags for gcc, which should speed up flac.
FLAC 0.6 released. The encoder is now much faster. The -m option has been sped up by 4x and -r improved, meaning that in the default compression mode (-6), encoding should be at least 3 times faster. Other changes:
- Some bugs related to flac and pipes were fixed (see here for the discussion).
- A "loose mid-side" (-M) option to the encoder has been added, which adaptively switches between independent and mid-side coding, instead of the exhaustive search that -m does.
- An analyze mode (-a) has been added to flac. This is useful mainly for developers; currently it will dump info about each frame and subframe to a file. It's a text file in a format that can be easily processed by scripts; a separate analysis program is in the works.
- The source now has an autoconf/libtool-based build system. This should allow the source to build "out-of-the-box" on many more platforms.
FLAC 0.5 released. This is the first beta version of FLAC. Being beta, there will be no changes to the format that will break older streams, unless a serious bug involving the format is found. What this means is that, barring such a bug, streams created with 0.5 will be decodable by future versions. This version also includes some new features:
- An MD5 signature of the unencoded audio is computed during encoding, and stored in the Encoding metadata block in the stream header. When decoding, flac will now compute the MD5 signature of the decoded data and compare it against the signature in the stream header.
- A test mode (-t) has been added to flac. It works like decode mode but doesn't write an output file.
- FLAC 0.4 released. This version fixes a bug in the constant subframe detection. More importantly, a verify option (-V) has been added to flac that verifies the encoding process. With this option turned on, flac will create a parallel decoder while encoding to make sure that the encoded output decodes to exactly match the original input. In this way, any unknown bug in the encoder will be caught and flac will abort with an error message.
- FLAC debuts on SourceForge. The FLAC project is now being hosted on SourceForge. Visit the FLAC project page to join the mailing list or sign up as a developer.