Skip to content
This repository


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Aug 11, 2013

  1. Updated German MythFrontend, MythArchive, MythGallery and MythMusic

    translation from Florian Bittner
    (The other files contained no new translations.)
    Thank you Florian!
    Closes #11738
    Signed-off-by: Nicolas Riendeau <>
    authored August 11, 2013 nriendeau committed August 11, 2013

Jul 03, 2013

  1. Clear additional field when resetting DVD VM to ensure playback doesn…

    …'t abort with "Error reading NAV packet".
    (cherry picked from commit bdc2b28)
    (cherry picked from commit c9accc2)
    authored July 01, 2013 tralph committed July 03, 2013
  2. John Poet

    dtvrecorder: Flush the payload buffer whenever a new payload starts.

    We start buffering at the beginning of each payload, and don't write out the
    data until we know if this payload has a keyframe.  This is so a new file
    (for back to back recordings) would start on a keyframe.
    Further, the keyframe offset recorded in the DB, is the offset to the
    beginning of the payload with the keyframe.  Or at least it was supposed to
    be.  Since we were not actually flushing any data currently in the buffer at
    the start of a payload, that offset could be pointing to a completely
    different payload!
    (cherry picked from commit 09b7873)
    (cherry picked from commit 10e87f5)
    authored June 28, 2013 tralph committed July 03, 2013

Jun 29, 2013

  1. rdv

    Add support for Miro v6.x.

    Signed-off-by: Raymond Wagner <>
    authored June 22, 2013 wagnerrp committed June 29, 2013

May 29, 2013

  1. Stuart Morgan

    Pull in fix for out-of-bounds array write in xine_demux_sputext.cpp f…

    …rom upstream. Coverity 746831
    (cherry picked from commit 7030916)
    authored May 29, 2013

Apr 22, 2013

  1. New Netherlands locale file

    Thank you!
    closes #11502
    Signed-off-by: Nicolas Riendeau <>
    authored April 22, 2013 nriendeau committed April 22, 2013

Apr 15, 2013

  1. k-ross

    Fix audio sync.

    Fixes #11440
    Signed-off-by: Jean-Yves Avenard <>
    (cherry picked from commit 9ec7421)
    authored April 15, 2013 jyavenard committed April 15, 2013
  2. libmyth: Avoid a divide by 0 exception in AudioOutputBase

    If MythMusic starts when the default audio is ALSA and if the device
    is in use by another process then snd_pcm_open can fail.  In this
    case AudioOutputBase::SetAudiotime can cause a divide by 0 exception
    because this->effdsp is 0.
    This patch checks the value of effdsp and avoids the division if it's 0.
    Signed-off-by: Jean-Yves Avenard <>
    (cherry picked from commit 84aac9d)
    authored June 24, 2012 jyavenard committed April 15, 2013

Mar 26, 2013

  1. John Poet

    Allow more than one keyframe per payload.

    See 69cd78b for fixes/0.26
    See edefeb6 for 0.27-pre
    Fixes #11459
    authored March 26, 2013

Mar 25, 2013

  1. Rob Smith

    Fixes #11463, this removes the binding Call-time pass-by-references

    authored March 25, 2013

Mar 22, 2013

  1. Jim Stichnoth

    Enable proper display of elapsed time when paused with VAAPI.

    Make sure av_pause_frame is updated so that the player can pick up the
    current frame number while paused.  Refs #10962
    (cherry picked from commit 841ff3a)
    (cherry picked from commit 6be8c10)
    authored March 22, 2013

Mar 08, 2013

  1. John Poet

    dtvrecorder: Don't try to buffer frames for audio-only streams.

    (cherry picked from commit 0d9312d)
    Signed-off-by: Taylor Ralph <>
    authored March 08, 2013 tralph committed March 08, 2013

Mar 03, 2013

  1. John Poet

    Cleanup some useless "discontinuity detected" messages following 87a15e1

    Closes #11328
    (cherry picked from commit efc1e46)
    Signed-off-by: Taylor Ralph <>
    authored March 02, 2013 tralph committed March 02, 2013
  2. John Poet

    DTVRecorder: buffer up each payload until we know if the payload is a

    Especially with H.264, we can be several packets into a payload before we
    know if it is a keyframe or just a frame or something else.  Always buffer
    the packets of a payload until we know what we are doing with the payload.
    If we need to switch ringbuffers, we now do it so the new file will start at
    the beginning of the payload holding a keyframe.  Each file will also start
    with a PAT/PMT if the "recorder" calls for it.
    (cherry picked from commit 87a15e1)
    Signed-off-by: Taylor Ralph <>
    authored March 02, 2013 tralph committed March 02, 2013

Mar 01, 2013

  1. Thomas Joiner

    backport 0026dda

    fix issue #10222: mythcommflag segfaulting
    The problem in issue #10222 is that it is adding the CODEC_FLAG2_FAST flag.
    This causes unsafe functions to be used, causing a segfault. The proper way to
    fix this would probably be to figure out how to test for the "experimental
    speedups" flag, however I don't know the codebase well enough for this.
    I tested how this would affect the speed of mythcommflag on one of my recordings.
    The results:
    Before removing flag:
    real 2m13.123s
    user 2m16.713s
    sys 0m3.502s
    After removing flag:
    real 2m13.021s
    user 2m15.446s
    sys 0m3.758s
    Signed-off-by: Taylor Ralph <>
    authored February 28, 2013 tralph committed February 28, 2013

Feb 26, 2013

  1. Jim Stichnoth

    Fix h.264 frame/keyframe identification in the player.

    If a NAL unit that determines a frame start appears as the last NAL
    unit in the packet, then the frame start is not identified until the
    start of the next packet.  If this new packet has its own NAL unit
    that determines a frame start, and that NAL unit is not at the end of
    the packet, then only one frame start would be counted rather than
    two.  This is fixed by returning the total number of frame starts,
    rather than a bool.
    In theory, the same problem could exist in the recorder, but it
    appears that the packets in the original recording may have some
    trailing NAL units that are stripped by the ffmpeg libraries when
    assembling the packet for the player, causing the situation not to
    (cherry picked from commit 86c22ed)
    Bumps the binary version.
    authored February 20, 2013

Feb 01, 2013

  1. Jim Stichnoth

    Enable more ffmpeg logging at startup of playback.

    Allow logging in avformat_find_stream_info() if
    "-v libav --loglevel debug" is specified.
    This proves useful for live TV debugging among others.
    (adapted from commit ae0dcaf)
    authored January 31, 2013

Jan 26, 2013

  1. David Engel

    Fix issues which sometimes prevente scheduling around live TV.

    Backport of SHA1:b05ecd46 to fixes/0.25.
    There are some cases where we don't attempt to schedule around live
    TV.  This change cleans up the logic which determines when to do that
    and fixes some of the bugs found in the process.
    NB: the case where a user enters live TV immediately before a
    recording is scheduled to start is still not handled gracefully.
    Fixing that is a lot tougher and will likely mean involving the
    scheduler in choosing which input to use for live TV.
    Refs #11207
    authored January 25, 2013

Jan 12, 2013

  1. Jim Stichnoth

    Fix "mythcommflag --rebuild" for non-HDPVR recordings.

    By default, only identify IDR keyframes for h.264 recordings.
    However, since DVB-S recordings may lack IDR frames, we check after
    1000 frames whether any keyframes have been found, and if not, back up
    and restart using the more lenient setting.
    Adapted from 802e32b and
    authored January 12, 2013
  2. Jim Stichnoth

    Subtitles: fix missing CEA-708 closed captions.

    This is a quick hack version of
    7d9bf31, which fixes a CEA-708
    decoding bug exposed by recent changes to NBC's CEA-708 encoding.
    (cherry picked from commit eca41c8)
    authored January 11, 2013

Jan 03, 2013

  1. Jonatan Lindblad

    mythccextractor: Skip teletext page headers

    (cherry picked from commit 9519a10)
    authored January 03, 2013
  2. Jonatan Lindblad

    mythccextractor: Fix teletext subtitle duration

    Previously ProcessTeletext() didn't leave one subtitle behind.  This caused the
    length calculation in IngestSubtitle() to never be performed, which meant that
    the length was always set to the default value 750 ms.
    Fixes #11081.
    (cherry picked from commit 08ad6ce)
    (cherry picked from commit 018691f)
    authored January 03, 2013
  3. Jim Stichnoth

    Configure the h.264 parser to identify only IDR keyframes.

    This makes keyframes identified by "mythcommflag --rebuild" and
    mythtranscode much closer to reality, though there are still
    discrepancies compared to the recorder and raw ffmpeg.
    (cherry picked from commit 590f161)
    authored January 03, 2013

Jan 01, 2013

  1. Jim Stichnoth

    Fix an issue with keyframe positions produced by mythcommflag --rebuild.

    The seeking code in DecoderBase assumes there is a keyframe at frame
    0, otherwise it shifts all keyframe numbers by the value of the first
    keyframe number in the map.  This causes problems for recordings with
    variable keyframe distances.  To correct this, make sure mythcommflag
    --rebuild creates an explicit entry for frame 0.
    Note: this does not fix the issue of mythcommflag --rebuild adding
    bogus keyframes to the seektable for h.264 recordings.
    Cherry-picked from a small part of
    49dbed5 .
    (cherry picked from commit ff4edc7)
    authored January 01, 2013

Dec 18, 2012

  1. Stuart Morgan

    Fix copy/paste error in MythUIType::SetMinArea(). Coverity 746771

    authored December 18, 2012
  2. Stuart Morgan

    Fix copy/paste error in MythUIType::AdjustMinArea(). Coverity 746772

    authored December 18, 2012

Dec 05, 2012

  1. Stuart Morgan

    Fix .empty() being called when .clear() was meant in cddb.cpp. Spotte…

    …d by cppcheck
    authored December 05, 2012

Dec 01, 2012

  1. Jean-Yves Avenard

    Fix cross-compilation on mac.

    Depending on the version of Qt you are using QMAKE_OBJECTIVE_CXXFLAGS may be ignored and QMAKE_OBJECTIVE_CFLAGS is used instead.
    So define both.
    [cherry-picked from 75e47c5]
    authored December 01, 2012

Nov 30, 2012

  1. Jean-Yves Avenard

    Fix AirTunes so it works with iTunes 11

    iTunes used to only send an AES-128 challenge once, and myth was expecting that the challenge came during the OPTIONS packet.
    The challenge is now sent with more packet, including during the ANNOUNCE one.
    So we move the handling of the challenge so it will work no matter which packet it comes with.
    Tested with iTunes 11, iTunes 10 and iOS 6.01
    [adapted from 06be922]
    authored November 30, 2012
  2. Jean-Yves Avenard

    Fix compilation with clang/clang++ - configure: we can't use CFLAGS f…

    …or compiling C++ code, especially as CFLAGS would include -std=c99 which makes no sense for C++ - was compiled as a C++ file, not an objective-C++ file, gcc was forgiving about the error, clang++ isn't
    [cherry-picked from c3544ed]
    authored June 14, 2012
  3. Jean-Yves Avenard

    Fix compilation against SDK 10.7 when using Xcode 4.4

    [cherry-picked from 31e5b9d]
    authored June 14, 2012

Nov 29, 2012

  1. Set palette correctly when DVD menu highlight changes to ensure highl…

    …ights are the right colour/have the correct transparency.
    Fixes #11265
    Signed-off-by: Stuart Morgan <>
    (cherry picked from commit 5df1322)
    authored November 29, 2012 stuartm committed November 29, 2012
  2. iandall

    Fix deletes on Slave backends. Remove references to deleted_set and

    initialise dont_expire_set properly which conflicted with 'deleted'
    recgroup behaviour since 0.25. Fixes #10704
    Signed-off-by: Stuart Morgan <>
    (cherry picked from commit e150a19)
    authored November 29, 2012 stuartm committed November 29, 2012
  3. Stuart Morgan

    Treat POLLHUP as EOF not an error, fixes IVTV (PVR-) recorders failin…

    …g. Based on patch from Fixes #10732
    (cherry picked from commit 3d13d79)
    authored November 29, 2012

Nov 23, 2012

  1. daniel-kristjansson

    A RecordingRule is owned by it's RecordingInfo and should not be free…

    …'d elsewhere.
Something went wrong with that request. Please try again.