Re: [mpg123-users] mpg123_read() (Error reading the stream. (code 18))

[Leonard B. <leo...@gm...>]

Hi Martin

Found the issues. 

(1) The use of fread()
(2) The small buffer size from mpg123_outblock(mh) which is inadequate for ALSA (a small buffer causes ALSA to play with jitter)

To resolve (1) I have changed the code to that used in the example file “mpg123_to_wav_replaced_io.c". This uses read() in a callback arrangement instead of fread(). I haven’t tested if the callback is actually necessary, but I think not.. I assume the issue with fread() is something to do with the internal buffering it does.

To resolve (2) I have used a much larger write buffer than read buffer, like this……

    int samples = 0;    
    int scale = 16;         // Set this high enough to satisfy ALSA

    buffer_size = mpg123_outblock(mh);
    unsigned char *buffer = malloc(buffer_size);

    do {
        struct sa_audio_data_mp3_t *audio = (struct sa_audio_data_mp3_t*) malloc(sizeof (struct sa_audio_data_mp3_t));
        buffer_write_size = buffer_size * scale;
        audio->data = malloc(buffer_write_size);
        audio->data_len_bytes = 0; 

        for (int idx = 0; idx < scale; idx++) {				// Stack the in data into the out buffer by scale number of times
            err = mpg123_read(mh, buffer, buffer_size, &size);
            if (err < 0 )
                printf("Error reading (%s)\n", mpg123_strerror(mh));
            memcpy(&audio->data[audio->data_len_bytes], buffer, size);
            audio->data_len_bytes = audio->data_len_bytes + size;
        }

        audio->play_metadata = play_metadata;
        audio->chunk_count = chunk_count++;
        if (!play_queue_push(audio))
            free(audio);
        samples += size / sizeof (short);

    } while (err == MPG123_OK);

Thanks very much, nice to have it working.
Leonard


> On 23/02/2021, at 1:21 AM, Martin Guy <mar...@gm...> wrote:
> 
> On 22/02/2021, Leonard Bocock <leo...@gm... <mailto:leo...@gm...>> wrote:
>> Hi list
>> 
>> I’m stumped. This is what I see in the trace log.
>> 
>>>> in=294912 read=73728 decoded=73728 chunk=17
>>>> in=294912 read=73728 decoded=73728 chunk=18
>>>> mpg123_read() error (Error reading the stream. (code 18))
>>>> in=294912 read=73728 decoded=32256 chunk=19
>>>> fread() from file
>>>> mpg123_feed() fill mpg123 buffer
>>>> in=368640 read=73728 decoded=73728 chunk=20
>>>> in=368640 read=73728 decoded=73728 chunk=21
>>>> in=368640 read=73728 decoded=73728 chunk=22
>>>> in=368640 read=73728 decoded=73728 chunk=23
>>>> mpg123_read() error (Error reading the stream. (code 18))
>>>> in=368640 read=73728 decoded=27648 chunk=24
>> 
>> Notice the last chunk returned from mpg123_read() is truncating by a value
>> of either 27648 or 32256 with error code 18. I swapped the coding approach
>> to use mpg123_decode() but am seeing the same issue.
>> 
>> The core of the C code w/o error handling is here.
> 
> Hi Leonard
>   If you can attach the MP3 file in question or make it available
> some other way, I'll see if my use of libmpg123 reads it OK. For your
> C code, a whole example C file would be easier to debug. The only
> thing that comes to mind from a quick inspection is that mh doesn't
> know the FORMAT when mpg123_outblock is called, but I doubt it's that
> or it wouldn't work at all.
>   Lastly, to see if it's libmpg123 that's unable to read the MP3
> file, or your use of libmpg123 that needs tweaking, you could see if
> the mpg123 command reads the file OK.
> 
>    M

Leonard Bocock
leo...@gm...