Hi,
we have finally put our production system to 1.0.0beta1 using alsa without any
errors anymore.
To close this thread I just want to summarize the needed steps:
1) install the int overflow patch before configure, make, make install
Index: alsa_stubs.c
===================================================================
--- alsa_stubs.c (révision 7697)
+++ alsa_stubs.c (copie de travail)
@@ -114,7 +114,11 @@
return ans;
}
-static void check_for_err(int ret)
+/* Here, we use the type snd_pcm_sframes_t
+ * which is long usually and makes sure we
+ * avoid long to int overflow in the read/write
+ * function's return value. */
+static void check_for_err(snd_pcm_sframes_t ret)
{
if (ret>= 0)
return;
@@ -296,7 +300,7 @@
int ofs = Int_val(ofs_);
snd_pcm_t *handle = Pcm_handle_val(handle_);
char *buf;
- int ret;
+ snd_pcm_sframes_t ret;
if (ofs + len * Frame_size_val(handle_)> caml_string_length(dbuf))
caml_invalid_argument("buffer");
@@ -320,7 +324,7 @@
int ofs = Int_val(ofs_);
snd_pcm_t *handle = Pcm_handle_val(handle_);
char *buf;
- int ret;
+ snd_pcm_sframes_t ret;
if (ofs + len * Frame_size_val(handle_)> caml_string_length(sbuf))
caml_invalid_argument("buffer");
@@ -346,7 +350,7 @@
snd_pcm_t *handle = Pcm_handle_val(handle_);
char **buf;
int c;
- int ret;
+ snd_pcm_sframes_t ret;
/* TODO: check the size of dbuf */
buf = malloc(chans * sizeof(char*));
@@ -377,7 +381,7 @@
snd_pcm_t *handle = Pcm_handle_val(handle_);
char **buf;
int c;
- int ret;
+ snd_pcm_sframes_t ret;
/* TODO: check the size of sbuf */
buf = malloc(chans * sizeof(char*));
@@ -408,7 +412,7 @@
snd_pcm_t *handle = Pcm_handle_val(handle_);
float **buf;
int c, i;
- int ret;
+ snd_pcm_sframes_t ret;
/* TODO: check the size of dbuf */
buf = malloc(chans * sizeof(float*));
@@ -440,7 +444,7 @@
snd_pcm_t *handle = Pcm_handle_val(handle_);
float **buf;
int c, i;
- int ret;
+ snd_pcm_sframes_t ret;
/* TODO: check the size of fbuf */
buf = malloc(chans * sizeof(float*));
@@ -472,7 +476,7 @@
snd_pcm_t *handle = Pcm_handle_val(handle_);
double **buf;
int c, i;
- int ret;
+ snd_pcm_sframes_t ret;
/* TODO: check the size of fbuf */
buf = malloc(chans * sizeof(double*));
@@ -504,7 +508,7 @@
snd_pcm_t *handle = Pcm_handle_val(handle_);
double **buf;
int c, i;
- int ret;
+ snd_pcm_sframes_t ret;
/* TODO: check the size of fbuf */
buf = malloc(chans * sizeof(double*));
2) create a virtual alsa device in
/etc/asound.conf
pcm.liq-test {
type plug
slave { pcm "hw:0,0" }
}
3) use the created alsa device in source scripts
input.alsa(device="pcm.liq-test")
overrun occures at
44100 hz * 16 bit * 2 (stereo)
4294967296÷(44100×16×2)
12*3600 secs =43200 secs
Am 01.10.2010 18:31, schrieb David Baelde:
> Good news! I should have mentioned it earlier, but Romain pointed out
> that my patch is useless. So what really happened is that not using
> alsa-pulse you simply didn't have that weird error. And there was no
> attempt to recover. Hence the signal should be clean.
>
> We could try to re-prepare instead of recovering when such errors
> happen. But, since you are the only one to have experienced it and you
> just found the cause and fixed the problem, we might simply leave our
> code as it is.
>
> Cheers,
>
> David
>
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_feb
_______________________________________________
Savonet-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/savonet-users
