Hi Andrew, On Thu, Sep 10, 2015 at 8:37 AM, Clément Bœsch <u...@pkh.me> wrote:
> On Thu, Sep 10, 2015 at 08:35:21PM +1000, Andrew Armstrong wrote: > > Hello, > > > > A recent Chrome update today has caused seeking in audio files streamed > on > > the web (eg, via SoundCloud and others) to no longer work. Instead, the > > file is slowly buffered until the required position in the data stream is > > reached. > > > > I added some notes about this problem at the Chrome bug report list at > > https://code.google.com/p/chromium/issues/detail?id=530043 > > > > I'm not sure how to directly test this in the ffmpeg toolset, perhaps it > > has been corrected in a recent commit? Please see below for my analysis. > > > > > > *ANALYSIS* > > ffmpeg has been updated in the new Chrome release. Various mp3 seeking > > commits were included that made large changes, breaking this feature. > Other > > commits have since been added that have not been merged in, I do not know > > if these fix the issue - potentially. > > > > I noticed a new seek "mode" was added to ffmpeg for mp3's that causes > > seeking to either by disabled or use the "slow but accurate" mode of > > reading data until the byte position is available, instead of causing a > > seek like before. > > > > This commit, which was not part of the merge, is also of interest > > > https://github.com/FFmpeg/FFmpeg/commit/c43bd08f8b043df7e18110e5344283c37b8380c1 > > > > *DETAILS* > > I noticed that ffmpeg was updated, and ffmpeg had a lot of changes > > regarding mp3 seeking, and more changes after the commit that Chrome > merged. > > > > 1. Chrome stable release diff: > > > https://chromium.googlesource.com/chromium/src/+log/44.0.2403.157..45.0.2454.85?pretty=fuller&n=10000 > > > > 2.These are the ffmpeg DEPS submodule versions > > 44x ffmpeg deps: ffmpeg cc2ec2825b0cc25cf27c5843847e7028c1cdb075 > > 45x ffmpeg deps: ffmpeg 833732528c1873f37b490b289eeaded2ae86349c > > > > 3. Diff of ffmpeg from the previous stable release to this one: > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+log/cc2ec2825b0cc25cf27c5843847e7028c1cdb075..833732528c1873f37b490b289eeaded2ae86349c > > > > 4. Commits affecting mp3 seeking in this diff: > > avformat/mp3: large id3 tags break concatenated file detection > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+/537ab680534e53bd298ba3f62d4aabb56afcd403 > > > > avformat/mp3dec: fix gapless audio when seeking in CBR mode > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+/92eef26e67ea8d2265854594344f8db17b9ce299 > > > > avformat/mp3dec: Allow forcing the use of the xing TOC for CBR files > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+/a742a0536dbb8b5a280b21ffd76c8b4acdbd20a6 > > > > avformat/mp3dec: offset seek index to end of id3v2 tag > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+/8b76c0eb561b0313e2a27950fe9d2bc5e4780dd8 > > > > avformat/mp3: skip junk at the beginning of mp3 files > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+/2b3e9bbfb529e6bde238aeb511b55ebe461664c8 > > > > avformat/mp3dec: allow enabling generic seek mode > > > https://chromium.googlesource.com/chromium/third_party/ffmpeg.git/+/c3a73666ad1eee93e49f25efae30fda5556c228e > > > > 5. I notice that further commits, that were not merged, have been made to > > ffmpeg seeking at > > > https://github.com/FFmpeg/FFmpeg/commit/c43bd08f8b043df7e18110e5344283c37b8380c1 > > > > So which commit is responsible of the regression? Also, what code does Chrome invoke to seek? I assume it calls av_seek_frame() - with which arguments? Ronald _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
