After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 104854 - mp3 typefinding (and typefinding in general) too exact
mp3 typefinding (and typefinding in general) too exact
Status: RESOLVED FIXED
Product: GStreamer
Classification: Platform
Component: gst-plugins
0.6.0
Other other
: High major
: 0.6.x
Assigned To: Erik Walthinsen
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2003-01-30 22:18 UTC by Colin Walters
Modified: 2004-12-22 21:47 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
minor cleanups (570 bytes, patch)
2003-01-30 22:19 UTC, Colin Walters
none Details | Review
First attempt at checking the whole buffer. Please test. (2.74 KB, patch)
2003-01-31 01:03 UTC, Erik Walthinsen
none Details | Review

Description Colin Walters 2003-01-30 22:18:46 UTC
For internet radio, I often get a stream that is right in the middle of an
MP3.  The current mp3 typefinding code
(/foreign-cvs/cvs.sourceforge.net/gst-plugins/gst/mpegaudioparse/gstmp3types.c)
is too exact; only looks at the beginning of the buffer.

In general, it would be nice if we had a way to pass a 'fuzziness'
qualifier to typefinding functions.  For now though, I think we would not
break anything if we just made the mp3 typefinding function search through
more of the buffer for an mp3 frame header.  Most programs will already
konw the content from the filename extension.

By the way, I have a small patch to clean up that file, which I will attach.
Comment 1 Colin Walters 2003-01-30 22:19:30 UTC
Created attachment 13968 [details] [review]
minor cleanups
Comment 2 Erik Walthinsen 2003-01-31 01:00:27 UTC
Applied patch 13968 and GST_DEBUG enhancement as
gst-plugins/gst/mpegaudioparse/gstmp3types.h 1.11.8.1.
Comment 3 Erik Walthinsen 2003-01-31 01:03:39 UTC
Created attachment 13975 [details] [review]
First attempt at checking the whole buffer.  Please test.
Comment 4 Erik Walthinsen 2003-02-01 22:31:09 UTC
Delaying until 0.6.x because it is not required for either of the two
deliverables (gnome-media and nautilus-media), and there's a
significant false-positive risk.
Comment 5 Erik Walthinsen 2003-02-01 22:58:06 UTC
Since all these bugs are listed as needing to be fixed in 0.6.x, they are by
definition bugs in 0.6.0
Comment 6 Ronald Bultje 2003-04-07 17:04:03 UTC
Applied to 0.6.1 CVS.
Comment 7 Ronald Bultje 2003-04-09 21:05:49 UTC
Thanks to our statistical genious Benjamin (company), we found out
that the above has a chance of between 35 and 100% for false positives
on a random sequence of 4096 bytes (our typefind buffer size!), which
means it's unacceptable. So the last bit will be reverted from 0.6.1.
Comment 8 Ronald Bultje 2003-06-08 10:17:02 UTC
We actually fixed this some time ago, didn't we? I'm quite sure we
search three frames now, which is always enough and works for streams,
too.