snd_pcm_readi returns -11
what does -11 mean as an error code?
documentation mentions -EBADFD, -EPIPE, -ESTRPIPE, but neither specifies
what those stand for (as in what patterns of bits) nor says anything about -11

I have better ways of generating -11, so I would rather call snd_pcm_readi
to read bytes... what magic is required for that to happen, and why do
I get -11?

An answer (to what magic...) would be a MINIMAL C program that actually
gets snd_pcm_readi to work in such a manner that the bytes read are stored
in a file (and when I cat the file, I really do see the bytes).

---

Clarification of Question by ignorant-ga on 12 Jan 2005 22:58 PST

I had been unable to use snd_pcm_state() to determine the PCM state
because the documentation I was reading mispelled macro names which
resulted in an inability to compile.

I am past that now and have discovered that once the state is
"SND_PCM_STATE_PREPARED" then executing snd_pcm_start() does manage to get
snd_pcm_readi to read bytes (sort of), but the data seems to be garbage;
the signal into line in (of the sound card) is from a guitar, but the
captured .wav file (produced via snd_pcm_readi) plays as nothing but static.

---

Clarification of Question by ignorant-ga on 12 Jan 2005 23:23 PST

By the way, the misspelling was "SND_PCM_STATE_PREPARE" 
(http://www.alsa-project.org/alsa-doc/alsa-lib/pcm.html)
but it needs to be "SND_PCM_STATE_PREPARED".

moving on, I should be able to learn enough with a debugger 
(hope springs eternal) if I could get aplay.c to compile
( -gdwarf-2 -g3 ), BUT where do I find aconfig.h?

---

Clarification of Question by ignorant-ga on 13 Jan 2005 07:44 PST

I know why I couldn't find aconfig.h
In alsa-utils-1.0.8/include one finds aconfig.h.in
Evidently aconfig.h is a generated file.

moving on,
gcc  -gdwarf-2 -g3 -o aplay aplay.c  -I/usr/local/include -lm -lasound 
is successful, so now maybe I can sort out my problems with capture by
using the debugger to probe this example app.

---

Clarification of Question by ignorant-ga on 13 Jan 2005 17:09 PST

Ok, so looking at the source for aplay.c indicates that if one does

cp aplay arecord

then I should be able to record; man arecord indicates that 

arecord -d 10 -f cd -t wav -D copy foobar.wav

will record.  Of course it doesn't, and generates error
messages instead:

ALSA lib pcm.c:1972:(snd_pcm_open_noupdate) Unknown PCM copy
arecord: main:507: audio open error: No such file or directory

Googling around indicates that one should leave out the -D option,
but (you guessed it) whereas

arecord -d 10 -f cd -t wav  foobar.wav 

does produce a wave file, it contains nothing but silence.  I tried 
successfully emerged media-sound/alsaplayer (I run gentoo), but of course
that also produces only wave files of silence.

My original question -- and goal -- was to capture with snd_pcm_readi,
but now I will settle for someone telling me how to make arecord actually
record!

---

Clarification of Question by ignorant-ga on 13 Jan 2005 17:48 PST

ok, so alsamixer displays a variety of inputs for capture,
and since I am plugged into line, I naively supposed enabling
line for capture (via the space bar) was appropriate...
silly me:  I need the input I am plugged into to show up
as being DISABLED, and I need something called "mix mono"
to be enabled... then arecord works (damn strange world!).

So now perhaps my debugging adventure (tracing the execution
of arecord -- which is simply aplay by a different name)
will shed light on snd_pcm_readi

If anyone is interested, feel free to beat me to the punch
(i.e., make a minimal C program that gets snd_pcm_readi to work)
and the money is yours.

If not, when I eventually succeed I'll post an answer.

---

Clarification of Question by ignorant-ga on 15 Jan 2005 12:18 PST

As you will see in the code below, there are a mess of calls in pre_record() of
the form snd_pcm_...  Honestly, I have little clue about why; they seem to be
needed to make recording work.  If anyone can give an in depth explanation of
what is going on with them, the money is yours --- and i'm increasing it to $20

Following is a reasonably "minimal" program that records (I promised to post):

#define _GNU_SOURCE
#include <stdlib.h>
#include <alsa/asoundlib.h>
#include "formats.h"

static struct 
{
  snd_pcm_format_t format;
  unsigned int     channels;
  unsigned int     rate;
} Hwparams = {SND_PCM_FORMAT_S16_LE, 2, 44100};

static snd_pcm_uframes_t Frames;
static size_t            Bytes_per_frame;
static snd_pcm_t        *Handle;

void pcm_read_frames(u_char *d)
{
  long r, s = Frames;

  while (s > 0) {
    if ((r = snd_pcm_readi(Handle,d,s)) > 0) {
      s -= r;
      d += r * Bytes_per_frame;
    } else if (r == -EAGAIN || ((r < s) && !(r < 0)))
      snd_pcm_wait(Handle,1000);
  }
}

#define Sz(x) sizeof(x)

void pre_record(int fd, long bytes)
{
  long                 buffer_time = 500000, period_time = buffer_time/4; 
  snd_pcm_uframes_t    buffer_frames, buffer_size, period_frames, xfer_align;
  WaveChunkHeader      wch, ch = {WAV_FMT,16};
  WaveHeader           h;
  WaveFmtBody          f;
  snd_pcm_hw_params_t *params; 
  snd_pcm_sw_params_t *swparams;

  snd_pcm_open(&Handle,"default",SND_PCM_STREAM_CAPTURE,0);

  wch.type     = WAV_DATA;
  wch.length   = bytes & 0x7fffffff;
  h.magic      = WAV_RIFF;
  h.length     = bytes+Sz(WaveHeader)+Sz(WaveChunkHeader)+Sz(WaveFmtBody)+Sz(WaveChunkHeader)-8;
  h.type       = WAV_WAVE;
  f.format     = WAV_PCM_CODE;
  f.modus      = Hwparams.channels;
  f.sample_fq  = Hwparams.rate;
  f.byte_p_spl = Hwparams.channels*snd_pcm_format_physical_width(Hwparams.format)/8;
  f.byte_p_sec = f.byte_p_spl*Hwparams.rate;
  f.bit_p_spl  = 16;
    
  write(fd, &h , Sz(WaveHeader)     );
  write(fd, &ch, Sz(WaveChunkHeader));
  write(fd, &f , Sz(WaveFmtBody)    );
  write(fd, &wch, Sz(WaveChunkHeader));

  snd_pcm_hw_params_alloca(&params);
  snd_pcm_sw_params_alloca(&swparams);
  snd_pcm_hw_params_any(Handle,params);
  snd_pcm_hw_params_set_access(Handle,params,SND_PCM_ACCESS_RW_INTERLEAVED);
  snd_pcm_hw_params_set_format(Handle,params,Hwparams.format);
  snd_pcm_hw_params_set_channels(Handle,params,Hwparams.channels);
  snd_pcm_hw_params_set_rate_near(Handle,params,&Hwparams.rate, 0);
  snd_pcm_hw_params_set_period_time_near(Handle,params,(unsigned *)&period_time,0);
  snd_pcm_hw_params_set_buffer_time_near(Handle,params,(unsigned *)&buffer_time,0);
  snd_pcm_hw_params(Handle,params);
  snd_pcm_hw_params_get_period_size(params,&Frames,0);
  snd_pcm_hw_params_get_buffer_size(params,&buffer_size);
  snd_pcm_sw_params_current(Handle,swparams);
  snd_pcm_sw_params_get_xfer_align(swparams,&xfer_align);
  snd_pcm_sw_params_set_sleep_min(Handle,swparams,0);
  snd_pcm_sw_params_set_avail_min(Handle,swparams,Frames);
  snd_pcm_sw_params_set_start_threshold(Handle,swparams,1);
  snd_pcm_sw_params_set_stop_threshold(Handle,swparams,buffer_size);
  snd_pcm_sw_params_set_xfer_align(Handle,swparams,xfer_align);
  snd_pcm_sw_params(Handle,swparams);
}

void post_record(int fd, long bytes)
{
  WaveChunkHeader cd          = {WAV_DATA, bytes & 0x7fffffff};
  off64_t         seek        = Sz(WaveHeader) + Sz(WaveChunkHeader) +
Sz(WaveFmtBody);
  u_int           wave_size   = bytes + 2*Sz(WaveChunkHeader) +
Sz(WaveFmtBody) + 4;

  lseek64(fd,4   ,SEEK_SET); write(fd, &wave_size, 4                  );
  lseek64(fd,seek,SEEK_SET); write(fd, &cd       , Sz(WaveChunkHeader));
  close(fd);
  snd_pcm_close(Handle);
  snd_config_update_free_global();
}

main(int argc, char *argv[])
{
  char *audiobuf;  long bytes, c, fd, time;  sscanf(argv[2],"%d",&time);

  fd    = open64(argv[1], O_WRONLY|O_CREAT|O_TRUNC, 0644);
  bytes = time*snd_pcm_format_size(Hwparams.format,
Hwparams.rate*Hwparams.channels);

  pre_record(fd,bytes);  // side effect : sets Frames

  Bytes_per_frame = Hwparams.channels*snd_pcm_format_physical_width(Hwparams.format)/8;
  audiobuf        = malloc(Frames*Bytes_per_frame);
  c               = bytes;

  do
    pcm_read_frames(audiobuf);
  while (0 < (c -= write(fd,audiobuf,Frames*Bytes_per_frame)));

  post_record(fd,bytes);
  free(audiobuf);
  return 0;
}
// gcc  -gdwarf-2 -g3 -o m m.c  -I/usr/local/include -lm -lasound