GitLab

The following interleaving can cause problems:
The interrupt for SIGINT could set sig_request.cancel right after the consumer thread
checks it ("buf->cancel_flag || sig_request.cancel"), but before the consumer
thread enters the mutex in order to wait for more data ("COND_WAIT"). If the producer
thread reacts to the new situation immediately by exiting (which it can do because the
consumer thread hasn't entered the mutex yet), the consumer thread initiates a COND_WAIT
that might never be fulfilled because the producer thread is gone. The producer's intention
to kill the consumer using buffer_thread_kill doesn't work because it was executed after
the consumer's check for buf->cancel_flag and before the consumer's COND_WAIT.

Fix the problem by putting the cancelation flag checks in the same mutex lock zone as
the "COND_WAIT".

When setting up a new chain link, vcedit_open_callbacks
always tries to read a new chunk at the edit of the
stream identification loop. This could fail if the
whole segment fits within a single, already buffered
chunk.

Instead, check for a new available page before trying
to buffer more data. As long as there's a data page
available after the header pages, this will succeed,
exiting early without trying to load data unnecessarily.

Thanks to lsd for reporting the issue.
Signed-off-by: Thomas Daede <tdaede@mozilla.com>

Last-Update: 2011-07-20
Forwarded: no
Bug-Debian: https://bugs.debian.org/328266
Bug-Debian: https://bugs.debian.org/634855

In checking the quality setting, the program `oggenc'
performs a floating point comparison after down scaling
the given value to a tenth. This causes the inexact
internal representation to complain unnecessarily.
*
It is better to compare first, and normalize later,
since quality 10 is top notch!

Bug-Debian: https://bugs.debian.org/595104

This commit adds -d/--rm to vorbiscomment. The new option allows
removal of tags from the processed files. It allows both removal
of specific tags or all tags with a given namen. It also implies
--append (which should maybe be renamed to --update).

Manpage has been updated.
Signed-off-by: Ralph Giles <giles@thaumas.net>

Signed-off-by: Ralph Giles <giles@thaumas.net>

Mostly re-wrapping lines. These are changes from running
autogen.sh on a Fedora 24 system.

Fixes an automake warning.

This is a continuous integration service integrated with github.

This just invokes 'make check'. 'make distcheck' fails with a permission issue:

    cannot move `/home/travis/build/rillian/vorbis-tools/vorbis-tools-1.4.0/_build/po/sv.new.po' to `sv.po': Permission denied

This is bad style, especially since "goto fail".

... to follow the one in CHANGES.

MSVC doesn't have unistd.h

autoreconf looks for ACLOCAL_*FLAGS in Makefile.am and will
pass that on when it invokes aclocal. Furthermore, setting
a local like this doesn't make it available to subcommands,
so the line has no effect.

The automake dist target should take care of this for us now.

This is symlinked by automake and shouldn't be in-tree.

This simplifies compiling the utility stand-alone for testing
or in alternate build systems.

svn path=/trunk/vorbis-tools/; revision=19424

Input files with sampling rate 0 are useless and can make oggenc crash
because neither oggenc itself nor libvorbis are prepared for this case.
A sanity check lets the program refuse those inputs gracefully without crash.

Patch by Martin Steghöfer <martin@steghoefer.eu>

Fixes #2078
Bug-Debian: https://bugs.debian.org/716613

svn path=/trunk/vorbis-tools/; revision=19423

The ogg123 executable wasn't able to correctly play stereo speex
files that don't contain any inband intensity stereo signals at
all or whose first inband intensity stereo signal arrives after
the stream has already started. This was due to the stereo
information data structure not being initialized properly before
the first inband intensity stereo signal arrived. So in the
mentioned cases the speex decoder used uninitialized float
values for the stereo decoding. This patch fixes the problem by
using the proper initialization and deallocation functions from
libspeex.

Patch by Martin_Steghöfer <martin@steghoefer.eu>
Bug-Debian: https://bugs.debian.org/312185
Fixes #1676


svn path=/trunk/vorbis-tools/; revision=19405

ogg123: fix a format security issue in status.c, reported by downstream https://bugs.gentoo.org/520580


svn path=/trunk/vorbis-tools/; revision=19227

oggenc: fix crash on raw file close, reported by Hanno in issue #2009. pointer to a non-static struct was escaping its scope. Also fix a C99-ism.

svn path=/trunk/vorbis-tools/; revision=19117

svn path=/trunk/vorbis-tools/; revision=19090

Typo in the conditional to check the exit condition in ascii_strcasecmp.
It checks for (!*s1 || !*s1) instead of (!*s1 || !*s2). The typo did
not affect the result of the function as the loop is exited before
changing s1 or s2 anyway.

Fix ported from the copy in the flac library. Thanks to dcb314 for
reporting it here.

Closes https://trac.xiph.org/ticket/2007

svn path=/trunk/vorbis-tools/; revision=19089

svn path=/trunk/vorbis-tools/; revision=19046

Fixes an issue with the macro expecting pthread_t to be
convertable from an int, which isn't true on mingw where
pthread_t is a struct.

Also move the macro out of acinclude.m4 and into its own
file in the m4 subdirectory.

Thanks to Laura Michaels for reporting the issue.


svn path=/trunk/vorbis-tools/; revision=18791

Update configure.ac to, as ron puts it, "the 21st century" (mostly macro updates that appear to be symvol-pushing)


svn path=/trunk/vorbis-tools/; revision=18763

Patch to the vorbis-dev list. https://bugzilla.redhat.com/887540

svn path=/trunk/vorbis-tools/; revision=18747

ogginfo wasn't correcting for the granule position difference
 between 3.2.0 and 3.2.1 streams.
It was always assuming the 3.2.1 behavior, but since this code
 hasn't been changed since 2005 and the 3.2.1 bitstream bump was
 done in 2007, it was simply wrong before.

svn path=/trunk/vorbis-tools/; revision=18497

svn path=/trunk/vorbis-tools/; revision=18450

This isn't all of them, by any stretch, but they are
the ones which were easy to trigger when the output
file could not be opened. Helps with valgrind cleanness.

svn path=/trunk/vorbis-tools/; revision=18449

This option was accidentally disconnected in r16793.
This patch restores the option's functionality, and allows the code
 to distinguish between an actual unknown length (totalframes==-1)
 and an empty file (totalframes==0).
Fixes #1803.

svn path=/trunk/vorbis-tools/; revision=18383

resize_buffer() would only free the channel buffers when the
 channel count changed, not when the size of the buffer changed,
 even though it allocated new buffers in both cases.

svn path=/trunk/vorbis-tools/; revision=18382

patch from ph3r-der-loewe
See Trac #1786

svn path=/trunk/vorbis-tools/; revision=18121