[MPlayer-dev-eng] [PATCH] LibFaad2 update

[Diego Biurrun]

César Ramos writes:
> - I make test: It work with all sample ftp files. 
> - Added license files.
> 
> 
> Changes: 
>     - specrec.c: fixed compilation without PS
>     - sbr_fbt.c, sbr_hfgen.c: fixed possible lockup with bad data.
>     - cfft.c,h, common.c,h, filtbank.c,h, mdct.h, sbr_qmf.h, structs.h,
> decoder.c, specrec.c:
>          Removed USE_SSE
>     - specrec.c: small optimizations to inverse quant
>     - neaacdec.h, decoder.c, syntax.c, structs.h: added output flag for PS
>     - common.c: added log2_fix()
>     - sbr_syntax.c: bugfix with PS_DEC and DRM_PS defines
>     - cfft.c, cfft_tab.h: added size 128 fixed point FFT tables for HVXC
>     - sbr_syntax.c: allow only 1 PS data element per SBR extension data
>     - specrec.c: safeguard against PS suddenly going on
>     - drm_dec.c: make sure we still copy to the right channel if there's no
> data
>     - hcr.c: bugfix
>     - hcr.c: Rewritten and optimized
>     - drm_dec.*, sbr_dec.c, syntax.c: improved error concealment
>     - sbr_dec.c, sbr_syntax.c: duplicate code removal
>     - syntax.c: re-enable SBR CRC checking for DRM
>     - drm_dec.*: changed delay filter to 2 samples plus fractional delay, as
> in normal PS AAC
>                 (DRM standard is wrong here)

Patch rejected, it is missing the following important bugfix for SBR
files:


Subject: [Mplayer-cvslog] CVS: main/libfaad2 common.h,1.6,1.7
Date: Mon, 27 Sep 2004 04:11:26 +0200 (CEST)

Modified Files:
	common.h 
Log Message:
SBR code does NOT work with fixed point (uses floats, slow as hell)

Index: common.h
===================================================================
RCS file: /cvsroot/mplayer/main/libfaad2/common.h,v
retrieving revision 1.6
retrieving revision 1.7
diff -u -r1.6 -r1.7
--- common.h	24 Sep 2004 17:31:33 -0000	1.6
+++ common.h	27 Sep 2004 02:11:24 -0000	1.7
@@ -124,6 +124,9 @@
 # ifdef MAIN_DEC
 #  undef MAIN_DEC
 # endif
+# ifdef SBR_DEC
+#  undef SBR_DEC
+# endif
 #endif // FIXED_POINT
 
 #ifdef DRM


Have you run benchmarks comparing this updated version to what we
currently have in CVS?  Does it compile and run under Cygwin and
MinGW?  We have added some portability patches to the vanilla
sources.  I should add some sort of local diff to CVS someday..

Oh, and never EVER start a new thread by replying to a message.  If
you are too lazy to type an address into your mailer, how are we to
know that you are not negligent in the rest of the things you do?
This is not meant as an offense, but this is the way this kind of
behavior comes across.

Regards

Diego