Can you not update the download on the page and add a note saying "Releases downloaded prior to xx/xx/xx . . .", explaining? It will not be much hassle for users who wish to update their encoder, and it will be available to those outside of this fine community.

I agree : strict procedures are OK, but a bit of pragmatism is better. If there is a patch and if this patch works like Josh just said, then the patch should be available for download directly at FLAC's official site until the next minor release is out.

How about a 1.1.3a?

I wanted to check out the 'better multichannel support' of this new flac release so I did the following:

I extracted some PCM .wav files from a DVD-A disc I own, they're 24 bit 44.1KHz. I used Wavewizard v0.54b to merge the 6 .wav files according to Microsoft's standard order (L, R, C, LFE, RL, RR) into one big 5.1 24bit PCM .wav.

I then fed this file to convert using Foobar 0.9.4.2 set to preserve the bit depth and using the standard commandline: -s -8 - -o %d (even put my windows locale to English (UK)). This conversion failed and gives me the following error: Error writing to file (Encoder has terminated prematurely with code 1; please re-check parameters).

I tried feeding the same merged .wav file to the official flac frontend but it failed too, giving me both a warning (legacy WAVE file has format type 1 but bits-per-sample=24) and an error (WAVE has >2 channels but is not WAVE_FORMAT_EXTENSIBLE; cannot assign channels.

I've been converting my DVD-A discs using this very same method combined with FLAC 1.1.2 without any problems before. Moreover, putting back the 'old' flac.exe (1.1.2) to the Foobar folder results in a perfect encode.

What's up with the 'better mutlichannel support'?

(Regardless of this, thank you very much for creating this nice lossless encoder!)

The new flac 1.1.3 requires multichannel wav files to be in WAVE_FORMAT_EXTENSIBLE, but currently fb2k generates it in WAVE_FORMAT_EX for 'compatibility' reason.

Besides convincing Peter to add _EXTENSIBLE support to Converter, you can try the following undocumented option:

erm, doesn't the converter feed flac.exe the raw PCM data or am i missing something here?

Thanks Egor, that '--channel-map=none' options seems to do the trick . However, what I don't understand is that if Foobar seems to be the problem than why does the official flac frontend has the same problem? And is there any way I can create a multichannel .wav file from six mono .wav files that conform to the specification FLAC 1.1.3 demands?

/edit
Nice, FLAC 1.1.3 gives a 2% compression advantage compared to 1.1.2: respectively 1.04GB and 1.06GB on a 52 minute classical album @ 24 bit 44.1KHz.

Nope, fb2k just sends wav (not RAW) data to encoder's stdin.


What frontend do you mean?


The only freeware application was WaveWizard.

Well, the one that comes with the Windows Flac installer downloaded from http://flac.sourceforge.net/download.html. Seems to be created by 'Speek': http://members.home.nl/w.speek/flac.htm.

note that if you use --channel-map=none, the incoming channels should be in the correct order, otherwise they might be mapped incorrectly on decoding.

Josh

What's the simplest way to re-encode FLAC files with 1.1.3? I'm a windows user, and FLAC Frontend will not take FLAC as input to be encoded to FLAC. I want to keep all tags, etc, just encode with newer FLAC. (also, i'm not a command line master, so..)

The simplest way is foobar.

What kind of improvement is expected if all FLAC files encodedwith 1.1.2 at level 8 will be re-encoded with 1.1.3 at level 8? Some bytes, or will an album be some megabytes smaller?

depends on the material but probably 0.5-1.0%

This depends on the lenght of the album. I got an 1 MB difference on some while others gave me almost 20MB.

Cut and paste the following script (author unknown to me) into a text editor and name it something like "convert.cmd". Put in in the root of your flac directory. Doubleclick. And do something great of your day while reencoding.

Note: Of course you change the directory path of flac.exe to your specific.
Script takes care of the dot/comma bug.

EDIT: And it preserves the old tags.

You could also use Speek's Multi-Frontend.

The workaround is simple but the impact is large:
1. It affects a large portion of the user base.
2. It requires the users to read the known-bugs section.
3. It requires the users to verify what the decimal seperator is on the system, and then take action accordingly.
4. If they change locale, they must remember to change the FLAC parameter too.

Many people will be affected by this bug without knowing it. Many people will think they are affected and set the parameter to a faulty value.

That is the last I will say about that.

I'd even say "Many people are currently being affected by this (small) bug without knowing it". I won't say more.

This bug is not called out with regards to other versions (notable 1.1.2, 1.1.1) Can I safely assume that this section of the website is accurate, so earlier versions are free of this bug?

Triza

Does it preserve the tags?

Yes it does.

Yes they are BUT you should not switch back to an older version. The bug will not damage your files or do them any harm. The only issue is that the compression will not be as good as it could be, but it will still be as good as 1.1.2 (maybe a little bit better).
You still profit from the enhanced tagging features and the updated decoder.

If I use this script under Windows XP and I run this inside a map with FLAC files (no sub directories), then FLAC first encodes all files, but then starts again with the files it has already encoded. Am I doing something wrong?
BTW, it saves me 900KB of 300MB of files, so the improvement is not that much :-).

(Moved to a new topic)

John33, does your compile of flac 1.1.3 over at Rarewares take care of the dot/comma bug?

I've just uploaded a fresh compile which includes Josh's fix (I think! ). Can you try it and let me know, please?

Hi John33

I had just finished testing your first compile, as i suspected that it included the comma bug-fix, since it was compiled on the 8th January and the comma bug-fix submitted to CVS on the 5th January, but as you posted previously(and my test showed) it didn't. Anyway, i have just finished testing this new compile and it works perfectly and i got 0.5% better compression ratio on the default -5 setting. Also, could you please tell me if you have just included the single fix for the comma bug to the v1.1.3 release sources, or if you have done a compile of the newest CVS snapshot, so that the bug that would truncate output files after the first error when doing FLAC to FLAC transcoding of corrupt FLAC's also would be gone ?

Note: to others that want's to use this compile, then you will need to also download the libmmd.dll file, which also can be downloaded from rarewares.

Many thank's John33 for your continued efforts of making compiles available to us, it's much appreciated

Thank's in advance.


CU, Martin.

I included the single fix. However, if the consensus is that I should use the current CVS, I can do that easily enough. Is that safe, Josh?

BTW, thanks for the feedback on the compile.

Confirming that it works like a charm. Many thanks John, really appreciated! Speaking for myself, I see no use for a compile from the current CVS so no arm in the air here

I think so, I have not checked in anything but small bug fixes since the release. best to alter the vendor string in src/libFLAC/format.c though (like make the end say 20070113-john33 or something unique)

Josh

excuse me? can i make a feature request, i need messages of flac to be marked with colours. errors - red, ok's - green.

can you make this possible? it would be very suitable!

@John33

I don't think that you have to make a CVS release of flac either(unless you want to ), and i just asked, since i wanted to know what code i was using.

Btw, the bug-fixes that i know about from CVS is :

Fix bug where FLAC-to-FLAC transcoding of a corrupted FLAC file would truncate the transcoded file at the first error (SF#1615019: https://sourceforge.net/tracker/index.php?f...mp;atid=113478)

Fix bug where using -F with FLAC-to-FLAC transcoding of a corrupted FLAC would have no effect (SF#1615391: https://sourceforge.net/tracker/index.php?f...mp;atid=113478)

From the CVS repository of flac/src/flac/encode.c : http://flac.cvs.sourceforge.net/flac/flac/...code.c?view=log

And then of course the comma fix already included from the stream_encoder.c source.

CU, Martin.

OK, guys, thanks for the feedback. I'll not bother with the current CVS for now, but if it becomes appropriate at a later date just ask and all shall be provided!!

@Josh Coalson

Do you know why the optimized(ICL9.1-SSE2) flac.exe builds gives different filesizes ? There's no problem though, as i have verified that the decoded WAVs are bit-identical to the source files, but i was just currious about the exact reason for this.

PS. Many thank's for this new release which IMHO is just downright awesome and actually made me switch from WavPack to FLAC

Thank's again, mate


CU, Martin.


Edit: Of course i have remembered to add -A "tukey(0,5)" to the command-line of the stock compile when i compared the output filesizes of the stock compile VS the ICL9.1-SSE2 optimized compile, as the ICL9.1-SSE2 optimized compile's sources was taken from CVS after the 5th December 2005, so that it would include the comma decimal seperator bug-fix.

The answer from another topic (LAME 3.97: different binaries produce differente results!):

Yes, i know that this is the explenation for lossy codecs, but i would just think that introducing rounding errors into the calculations for lossless PCM samples would make the output lossy ?

Thank's for your reply, mate

CU, Martin.

libFLAC still uses floating point for the LPC analysis stage, so rounding errors can affect the filter creation. losslessness is achieved by substracting the filtered signal from the original and storing that, which is why it will always be lossless. i.e. if x is original signal, f() is filter, e is error, encoder computes e=x-f(x), and sends e and f only.

Josh

@Josh

Thank you very much for the explanation

Also have you any plans of adding wild-card support to flac.exe and metaflac.exe ?

@Egor

Sorry, mate - you where absolutely right


CU, Martin.

Just curious : what FLAC features made you switch ? What does FLAC currently have that WavPack doesn't ?

Thanks.

FLAC decodes faster and compresses better than WavPack's fastest decoding mode. FLAC has a CUESHEET metadata block which can decode single tracks out from the command line. FLAC has a PICTURE metadata block. FLAC can take FLAC files as input and transcode from FLAC to FLAC and also transfer tags. FLAC has a verify-while-encoding(-V) option. FLAC has better software support. FLAC has better hardware support.

I'm only interested in very fast decoding and not in lossy or hybrid mode or getting better compression in trade of lower decoding speed and then i saw that the new FLAC version both decoded faster while also compressing better than WavPack's fastest decoding mode. Then other features of the new FLAC version really impressed me like with it's added FLAC to FLAC transcoding while transfering all tags and a switch for embedding a cuesheet into a Vorbis comment(like WavPack) + even better than before error correction + giving better compression without decoding penalty.

CU, Martin.

Sorry, for repeating my question, but just in case that it was missed the first time around

@Josh Coalson

Would you please consider being so kind as to add wildcard support to flac.exe and metaflac.exe in a new version sometime ? Some days ago i wanted to test my FLAC archive before i re-encoded them again with the added -A "tukey(0,5)" switch, as i had forgotten to use that switch the first time around and had a comma as decimal seperator. Then i first rightclicked on my FLAC folder and selected "Open Command Window Here" and then i just typed "flac -t *.flac", but recieved an error message and so i understood that flac.exe didn't expanded wildcards unfortunetly. This then means that instead of being able to do this : "flac -t *.flac", then we have to do this instead : for %G in (*.flac) do flac -t "%G". Luckilly then Tycho has compiled a globbing utility which code is listed on the net and which i now use instead, but IMHO, then i really think that wildcard support for the command-line tools of FLAC, is a really important feature missing.

Anyway, i just wanted to fill this feature request to you, and also say many thank's for all your great work on the FLAC format.

CU, Martin.

I feel your pain Martin! this has come up before, have you seen the other discussions on it? I'm hesitant to add non-portable stuff to fix a bug in the windows shell that has been there for like 20+ years. some of the easy workarounds are too dangerous to be acceptable:
http://mail.python.org/pipermail/python-de...une/000397.html

Martin, that's not a FLAC thing: it's a Windows thing. Wildcards work fine in Linux.

Wow ! A 20-year old bug ! Amazing ! Thanks Bill !

Thank's for your reply Josh No, i actually haden't seen any previous discussions about this before, but i perfectly understand your reasoning for not implementing it

Thank you anyways


CU, Martin.

@Heliologue

Yes, i know that, mate

Tested that script, works like a charm. Many thanks. Now I only have to run that script once for every new version of FLAC. However, sometimes I get this error:

"ERROR: Input file <insert filename here> has an ID3v2 tag."

What do I do about this? I've tried foobar's "Rewrite file tags", but to no avail.

I'm using FLAC 1.2.1, if it matters.

Try the Mp3tag program, it may be possible to remove the felonious ID3 tag with it.

Thanks, I'll try. Is there no other way around this, to make it automatic? Preferably by including something in the script?

Guys, wildcards are done by your shell. It is a shell thing. It is not a Linux thing or a Windows thing. It is a shell thing.

If you write *.flac shell finds all the files ending with flac and passes these filenames to flac.

Putting a functionality like this into FLAC would be very bad idea. Possibly will cause problems. For example if the filename passed onto FLAC has these metacharacters, then we have a problem. Frankly escaping them would be awfully unusual and probably difficult, too for anyone used to how command line applications should work.

Regards,

Triza