On Mon, 2009-11-09 at 11:26 +0100, ext Mark Brown wrote:

On Mon, Nov 09, 2009 at 10:12:31AM +0200, Eero Nurkkala wrote:

...

As far as I know, nobody fixed it. Yeah, aplay = arecord, and it occurs only with arecord to be more specific, and is present on all platforms out there:

Has anyone reported whatever the problem is? _______________________________________________

Does the below make any sense?

Call to: err = snd_pcm_mmap_begin(handle, &areas, &offset, &chunk_size); Will make the chuck_size to zero.

It's perfectly fine to be zero:

/** * \brief Application request to access a portion of direct (mmap) area * \param pcm PCM handle * \param areas Returned mmap channel areas * \param offset Returned mmap area offset in area steps (== frames) * \param frames mmap area portion size in frames (wanted on entry, contiguous available on exit) * \return 0 on success otherwise a negative error code * * It is necessary to call the snd_pcm_avail_update() function directly before * this call. Otherwise, this function can return a wrong count of available frames. * * The function should be called before a sample-direct area can be accessed. * The resulting size parameter is always less or equal to the input count of frames * and can be zero, if no frames can be processed (the ring buffer is full). * * See the snd_pcm_mmap_commit() function to finish the frame processing in * the direct areas. */ int snd_pcm_mmap_begin(snd_pcm_t *pcm, const snd_pcm_channel_area_t **areas, snd_pcm_uframes_t *offset, snd_pcm_uframes_t *frames) {

--> aplay doesn't call "snd_pcm_avail_update()" as this function suggest. I try is also, but the problem is chuck_size, when zero:

at aplay: static ssize_t pcm_read(u_char *data, size_t rcount)

returns rcount always, never reading any through readi_func!!!!

This pcm_read is called at aplay:

/* capture */ fdcount = 0; while (rest > 0) { size_t c = (rest <= (off64_t)chunk_bytes) ? (size_t)rest : chunk_bytes; size_t f = c * 8 / bits_per_frame; if (pcm_read(audiobuf, f) != f) break; if (write(fd, audiobuf, c) != c) { perror(name); exit(EXIT_FAILURE); } count -= c; rest -= c; fdcount += c; printf("here!\n"); }

--> like you see this thing is always fine:

if (pcm_read(audiobuf, f) != f) break; if (write(fd, audiobuf, c) != c) { perror(name); exit(EXIT_FAILURE); }

so it keeps on writing in the file, not reading any from ALSA. (try with strace)

-->> Aplay is buggy, period.

So the patch:

--- a/alsa-utils-1.0.21/aplay/aplay.c +++ b/alsa-utils-1.0.21/aplay/aplay-fix.c @@ -1104,7 +1104,8 @@ static void set_params(void) if (mmap_flag && verbose) { const snd_pcm_channel_area_t *areas; snd_pcm_uframes_t offset; - int i; + int i, chunk_prev = chunk_size; + snd_pcm_avail_update(handle); err = snd_pcm_mmap_begin(handle, &areas, &offset, &chunk_size); if (err < 0) { error("snd_pcm_mmap_begin problem: %s", snd_strerror(err)); @@ -1112,6 +1113,11 @@ static void set_params(void) } for (i = 0; i < hwparams.channels; i++) fprintf(stderr, "mmap_area[%i] = %p,%u,%u (%u)\n", i, areas[i].addr, areas[i].first, areas[i].step, snd_pcm_format_physical_width(hwparams.format)); + + /* Chunk size better be non-zero */ + if (!chunk_size) + chunk_size = chunk_prev; + /* not required, but for sure */ snd_pcm_mmap_commit(handle, offset, 0); }