Port to FLAC 1.1.3

[ankoe]

I think it makes sense to upgrade to flac 1.1.3, it has nice enhancements (details ​here) But version 1.1.3 is not compatible to 1.1.2, there is a ​porting guide

[VMiklos <vmiklos@…>]

here is a possible implementation:

​http://darcs.frugalware.org/repos/flac113/source/xapps-extra/mythplugins/mythplugins-0.20a-flac113.diff

[Jarod Wilson]

A good start, but this patch will hose over anyone using flac 1.0.4 to 1.1.2. I believe we'll need to add some #ifdef'ing to prevent breakage, which the flac folks were kind enough to give an example of here:

​http://flac.sourceforge.net/api/group__porting.html

I'll see if I can't make that happen...

[Jarod Wilson]

Okay, I've got a patch all #ifdef'd to hell and back that builds cleanly against flac 1.1.4 now, and *should* build against 1.1.2 as well (I'll try that shortly). Also need to see if flac encoding and decoding still works in both cases...

​http://wilsonet.com/jarod/mythmusic-0.20-flac113.patch

[Jarod Wilson]

Does indeed build against 1.1.2 as well, and test encodings pass basic sanity check. Committing shortly...

[Jarod Wilson]

(In [12810]) Add support for new flac API introduced in 1.1.3. Rather insane to have a huge API change from 1.1.2 to 1.1.3, but whatever... At least the flac folks wrote some decent docs on porting to the new API... Builds cleanly against both flac 1.1.2 and 1.1.4, and basic sanity checks show it appears to be doing the right thing. Closes #2957.

[Nigel]

Jarod, tried to get in before you committed, but failed :-( Was wondering if you could further simplify the #defines up the top by removing the duplication - using something like:

#if !defined(FLAC_API_VERSION_CURRENT) || FLAC_API_VERSION_CURRENT <= 7
  /* FLAC 1.0.4 to 1.1.2 */
  #include <FLAC/file_encoder.h>
  #define FLACENC FLAC__file_encoder
#else 
  /* FLAC 1.1.3 and up */
  #define NEWFLAC
  #include <FLAC/stream_encoder.h>
  #define FLACENC FLAC__stream_encoder
#endif
#define encoder_new() FLACENC_new()
#define encoder_setup(enc, streamable_subset, do_mid_side_stereo, \
            loose_mid_side_stereo, channels, bits_per_sample, \
            sample_rate, blocksize, max_lpc_order, \
            qlp_coeff_precision, do_qlp_coeff_prec_search, \
            do_escape_coding, do_exhaustive_model_search, \
            min_residual_partition_order, max_residual_partition_order, \
            rice_parameter_search_dist) \
            { \
              FLACENC_set_streamable_subset(enc, streamable_subset); \
...

et c. (same with FLACSeekable vs FLAC).

[Jarod Wilson]

I thought about doing that, and then I forgot I'd thought about doing that... I'll take a crack at cleaning it up some tomorrow, its already past my bedtime. :)

[mythtv@…]

Compilation fixes for commit 12810 and FLAC 1.1.2

[mythtv@…]

Commit 12810 doesn't compile for me. I'm using FLAC 1.1.2 and the problem is down to the lack of ';' at the end of the lines in the encoder_setup macro for FLAC versions 1.1.2 and below (see attached flacfix_r12810.patch). The patch also removes the '/' from the macro calls in the CPP files, this wasn't necessary to make it compile but the '/' aren't necessary to make it compile either.

[Stuart Auchterlonie]

(In [12814]) Refs #2957. Fixes compilation after [12810] with flac 1.1.2

[Jarod Wilson]

Gah. Now how the hell did that happen? I hit that same thing building against 1.1.2 and fixed it, but that must not have made it back over to my svn tree. Sorry 'bout that...

Didn't have time to poke at the duplication cleanups today, definitely should tomorrow though...

[Jarod Wilson]

Oh yeah... The back-slashes in the cpp files make vim syntax highlighting much happier, but either or. :)

[Jarod Wilson]

Patch to greatly reduce duplication and size of define blocks

[Jarod Wilson]

The above patch hasn't actually been tested yet, but it illustrates the approach I'm taking to reduce the duplication of defines. Will compile test shortly...

[Jarod Wilson]

Hrm. So the approach nigel suggested (which is what I implemented) isn't working for me. Had someone else look at it, who thought it should be working, but now believes it simply doesn't work (tried something similar w/a very simple test program across multiple versions of gcc). Here's an example of what I'm seeing:

Header define in flacdecoder.h:
#define StreamDecoder FLAC__SeekableStreamDecoder

Line of code in flacdecoder.cpp:
static StreamDecoderReadStatus flacread(blah)

Error when compiling:
flacdecoder.cpp:20: error: 'StreamDecoderReadStatus' does not name a type

I'm somewhat inclined to leave well-enough alone at this point, unless there's an easy/obvious way to do this that I'm just missing...

[Nigel]

Sorry. Yes, there is a trick to get GCC's CPP to join symbols - the ##:

#define FLACENC(a) FLAC__file_encoder ## a
#define encoder_new() FLACENC(_new)
#define encoder_setup(enc, streamable_subset) \
        FLACENC(_set_streamable_subset)(enc, BLAH, streamable_subset)

encoder_new();
encoder_setup(a,b);

But feel free to leave it as-is, because this trick probably requires GCC.

[Jarod Wilson]

Ah, I'd played around with ## a little bit, but probably didn't quite get it right. Makes sense how that would work now. But yeah, if its going to require trickery that may not work universally, let's just leave it as-is.