- Jan 22, 2017
-
-
Mark Harris authored
The package_version file will be used if there are no tags, which is better than just the hash that is returned by --always.
-
Also appveyor.yml configuration file is moved to .appveyor.yml. Closes https://github.com/xiph/opus-tools/pull/14 Signed-off-by:
Mark Harris <mark.hsj@gmail.com>
-
Closes https://github.com/xiph/opus-tools/pull/12 Signed-off-by:
Mark Harris <mark.hsj@gmail.com>
-
Like with opus, this generates an unknown version for git archives (like the ones generated by GitHub). If you need a proper version either use releases, a proper git repo or craft your own package_version. Signed-off-by:
Mark Harris <mark.hsj@gmail.com>
-
Signed-off-by:
Mark Harris <mark.hsj@gmail.com>
-
Mark Harris authored
-
Mark Harris authored
-
- Jan 20, 2017
-
-
Mark Harris authored
Clearly indicate which options require a parameter and its usual format. Use imperative verb form consistently. Minor rearrangement to improve logical ordering. Move diagnostic options to the end. Move verbose --picture details to --help-picture so that basic usage information has a chance of fitting on one screen. Change use of obsolete term "MIME type" to "media type". Update man page to match.
-
Mark Harris authored
-
Mark Harris authored
When reading WAV/AIFF data with a known length, don't try to read (and then discard) data beyond that length, to avoid blocking or adding unnecessary latency when the data is not from a file.
-
Mark Harris authored
Avoid writing a final packet larger than 20ms that is larger than necessary, with 20ms or more that is beyond the end of the stream.
-
Mark Harris authored
-
Mark Harris authored
Even if it has only one channel, an input WAV file can contain a channel mask. A warning is issued if it is present but not a supported value. Don't warn about the previously accepted mask of 1 (FL), or 4 (FC). Arguably a channel mask of 1 (FL) should trigger the warning, just as 2 (FR) does, but some tools (including old opusdec when decoding to floating point WAV) actually write an incorrect channel mask of 1 for mono WAV files.
-
Mark Harris authored
Mono channel mask changed from 1 (front left) to 4 (front center) in WAV output.
-
- Jan 16, 2017
-
-
Ralph Giles authored
This avoids leaking the handle on error.
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
- Jan 14, 2017
-
-
Fix suggested by mark4o. Tested with 32 and 64-bit VS2015 and MinGW GCC 6.3. Signed-off-by:
Mark Harris <mark.hsj@gmail.com>
-
- Dec 31, 2016
-
-
Mark Harris authored
Fix errors (not affecting common values) in reading the sample rate from AIFF: the low 32 bits of the mantissa were incorrectly interpreted as signed, the first byte rather than first bit was being used as a sign, and non-integer rates were not properly rounded. Although the sample rate is validated prior to setting up the resampler, sample rates outside the range of an "int" could cause undefined behavior prior to that when cast from double to int for AIFF. Such rates could also be incorrect for WAV on 32-bit builds. Ensure that the rate is within the range of an int prior to casting.
-
- Nov 11, 2016
-
-
Signed-off-by:
Mark Harris <mark.hsj@gmail.com>
-
- Sep 23, 2016
-
-
Mark Harris authored
Don't treat unsigned fields as negative. Fixes crash due to negative totalsamples.
-
Mark Harris authored
-
Mark Harris authored
-
- Sep 22, 2016
-
-
Mark Harris authored
Fixes https://trac.xiph.org/ticket/2132, initial patch by rillke
-
- Sep 21, 2016
-
-
Ralph Giles authored
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Adopt markdown format Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
- Sep 20, 2016
-
-
Mark Harris authored
Don't print control characters and other invalid bytes from invalid WAV/AIFF fourccs to the terminal.
-
Mark Harris authored
Crash found by afl-fuzz.
-
Mark Harris authored
Properly handle chunks of any size from AIFF/WAV input files, including padding them to an even size as required. Also ensure that a relevant message is displayed on any AIFF/WAV open failure. See oggenc https://trac.xiph.org/ticket/1817, https://trac.xiph.org/ticket/1912, https://trac.xiph.org/ticket/2212 (CVE-2015-6749).
-
Mark Harris authored
Fix potential infinite loop on a specially-crafted input file if "long" is only 32 bits. Also on some systems, including some versions of Windows and 32-bit OS X, fseek() cannot handle large files even if the offset fits within a "long". These issues are addressed by consistently using the fseek/ftell variants that can handle large files, when available, for all input file processing rather than in only one place. Don't try to seek or use the result if ftell() returns an error (e.g. EOVERFLOW). If large file interfaces are not available, handle seeking in chunks rather than wrapping and potentially causing an infinite loop. When reading data until EOF, treat the ftell() value as only an estimate for percent progress; if the file grows, do not discard data after this point.
-
- Aug 27, 2016
-
-
Mark Harris authored
Interpret channel count properly for input format (signed for AIFF, unsigned for WAV) and reject invalid values. Fixes https://trac.xiph.org/ticket/2136 (CVE-2014-9639), https://trac.xiph.org/ticket/2137 (CVE-2014-9638).
-
- Jul 29, 2016
-
-
This enables automated testing of windows builds. Compile opus and libogg with same configuration and platform. Wasn't able to get the artifacts from the respective autobuilds. Package executables into artifact and change its name to contain version. Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Avoids trying to create and replace version.h more than once which led to file-locking errors with multicore builds. Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Fix x64 include paths which were using previous directory hierarchy. Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
Signed-off-by:
Ralph Giles <giles@thaumas.net>
-
- Jul 26, 2016
-
-