Mark Harris authored 10 years ago and Timothy B. Terriberry committed 10 years ago

Signed-off-by: Timothy B. Terriberry <tterribe@xiph.org>

Thanks to maikmerten for spotting it.

Mark Harris authored 11 years ago and Gregory Maxwell committed 11 years ago

Signed-off-by: Gregory Maxwell <greg@xiph.org>

Mark Harris authored 11 years ago and Gregory Maxwell committed 11 years ago

AIFF uses signed pcm for all sizes, whereas WAV is unsigned for 8-bit
and signed for other sizes.  For backward compatibility, continue to
use the WAV convention for raw input.

Signed-off-by: Gregory Maxwell <greg@xiph.org>

Mark Harris authored 11 years ago and Gregory Maxwell committed 11 years ago

Signed-off-by: Gregory Maxwell <greg@xiph.org>

Discarding the art is a special case because its often huge relative to the
 opus files.

Thanks for bitingsock on IRC for the report.

Signed-off-by: Ralph Giles <giles@mozilla.com>

Missed this in review.

If the METADATA_BLOCK_PICTURE field contained a non-empty mimetype
 or description field, then we would overwrite important pieces of
 the struct when encoding, making the tag useless.

Signed-off-by: Ralph Giles <giles@mozilla.com>

This provides a fallback commit id if no tags are available,
for example when doing a shallow clone.

Previously we linked to flac, but didn't set the enabling CFLAG.

wave_out.c also references PACKAGE_NAME and PACKAGE_VERSION.

GNU autotools used to enforce some coding standards, but I don't
see a warning about not having this or a changelog now.

This should have no real functional change at this point in time,
but it's clear that Ogg will never depend on Opus.  The reverse is
likely to remain true, but less axiomatic.

Only opusenc uses it at present.

Garf's new project files build things in win32/VSnnnn/ instead
of the top level. Also updates the ogg paths to use defines
instead of hard-coding release and platform tags.

It was still defining VERSION in the Visual Studio project files,
which hadn't been updated for this change.

Include the doc fixes from 3138cb7c,
even if people don't want the --quality alias after all.

This avoids at least one case where ./autogen.sh && ./configure && make
will re-run configure because the makefile rules updated something that
it depends upon. Pulling a new version from git will change the version
so we should update that at the first step rather than iterating after
the last one.

This reverts commit 3138cb7c.

This updates the configure.ac requirements for the last patch from
 JM, and also fixes the unused variable warnings it created.

Jean-Marc Valin authored 12 years ago and Gregory Maxwell committed 11 years ago

Channel mapping is now determined automatically by libopus.
This breaks the --uncoupled option because opusenc no longer
has any control over the mapping.

Automake 1.13 complains that INCLUDES is obsoleted now.

Test for the available possibilities in the code in the same order that
we use when including the headers.  This ensures that BSDs which test
positive for both sndio and oss compat won't include the headers for one
and the code for the other.  The sndio support was added in commit
064bb217, but only preferred sndio when
including the headers.

Enable the OSS code for all the possible HAVE_*_SOUNDCARD_H options.
A more comprehensive test and additional macros were added in commit
a0c50052, but only to include the
headers, not to also enable the relevant code.

Nothing uses it either.

It is declared to be obsolete and we already use AC_COMPILE_IFELSE
everywhere else, so finish the job to avoid confusion and people
pasting more instances of it in later.

Double quote all the parameters to AC_LANG_SOURCE and AC_LANG_PROGRAM.
This is actually required, even if you can get away with not doing it
sometimes, so again set a good example for future changes to follow,
to hopefully avoid people getting bitten harder than they need to be.

And the exp() check too.  Nothing in the code actually tests for that
and it's irrelevant whether it needs libm or not, since plenty of other
things definitely do need it.

So add the hardcoded -lm back where it is definitely needed (at the
very least for the speex resampler), and make sure it is in LIBS when
we test for lrint and f{min,max}f, since those tests do fail on Linux
if libm isn't linked.