* ./modules/*: moved plugins to the new tree. Yet untested builds include

[vlc] / doc / mad / BUGS

diff --git a/doc/mad/BUGS b/doc/mad/BUGS
new file mode 100644 (file)
index 0000000..96c3931
--- /dev/null
+++ b/doc/mad/BUGS
@@ -0,0 +1,27 @@
+bug #09/11/2001-1: (FIXED 20/11/2001, by JP Saman)
+       Makefile.opts.in and Makefile.opts static libmad.a library is not filled in automatically by configure.
+       It should say in Makefile.opts LIB_MAD=lib/libmad.a
+
+bug #09/11/2001-2: (FIXED december 2001)
+       Libmad is not included in the vlc source code. It is assumed that the library is already available on the machine
+       where the plugin is compiled on. The configure option --with-libmad tells the script where libmad.a or libmad.so library
+       and mad.h include file are installed. E.g. if libmad.a is in /usr/lib and mad.h is in /usr/include then the configure
+       option should be --with-libmad=/usr
+
+bug #12/11/2001-1 (FIXED january 2002)
+       smoothen out audio in libmad_output callback function, by optimizing conversion routines. After investigation
+       and experimenting I found out that my input routine and output routine were not balanced correctly with
+       respect to buffer sizes. A lot of data got lost. I fixed that, but still need a bater scaling/smoothing routine
+       for the audio.
+       I tried a simpler audio scale function, but it did not improve the audio much. Not even on Intel machine, so I suspect the audio
+       decoding to need a bit too much time. The next thing to try is to steer the decoding process by hand (going to a lower level API).
+       Rewriting mad plugin to use a lower level API was not needed. I found the problem by reviewing libmad example code madlld and bbplay.
+       This resulted in a partial rewrite a some parts and a lot of testing. There are still some issues when trying on different CPU's
+       but this will sort itself gradually.
+
+bug #20/11/2001-1 (FIXED 25/11/2001, by JP Saman)
+       synchronization between audio and video output is lost. In libmad_output a timestamp is set on the audio data, but
+       it points to a later video frame, then when the audio data was read. How can I set the correct timestamp on the
+        audio fifo at the time it was read from the bitstream, so that it matches the video frame it belongs to?
+        Solution: Do it in two steps, in the input function save i_pts of fifo and in output function pass saved i_pts to
+        aout_fifo.