Commit 21b1f08e authored by Paul B Mahol's avatar Paul B Mahol

avfilter/avf_avectorscope: switch to activate

parent 77aacdb0
...@@ -28,6 +28,7 @@ ...@@ -28,6 +28,7 @@
#include "libavutil/opt.h" #include "libavutil/opt.h"
#include "libavutil/parseutils.h" #include "libavutil/parseutils.h"
#include "avfilter.h" #include "avfilter.h"
#include "filters.h"
#include "formats.h" #include "formats.h"
#include "audio.h" #include "audio.h"
#include "video.h" #include "video.h"
...@@ -69,6 +70,7 @@ typedef struct AudioVectorScopeContext { ...@@ -69,6 +70,7 @@ typedef struct AudioVectorScopeContext {
int mirror; int mirror;
unsigned prev_x, prev_y; unsigned prev_x, prev_y;
AVRational frame_rate; AVRational frame_rate;
int nb_samples;
} AudioVectorScopeContext; } AudioVectorScopeContext;
#define OFFSET(x) offsetof(AudioVectorScopeContext, x) #define OFFSET(x) offsetof(AudioVectorScopeContext, x)
...@@ -208,12 +210,8 @@ static int config_input(AVFilterLink *inlink) ...@@ -208,12 +210,8 @@ static int config_input(AVFilterLink *inlink)
{ {
AVFilterContext *ctx = inlink->dst; AVFilterContext *ctx = inlink->dst;
AudioVectorScopeContext *s = ctx->priv; AudioVectorScopeContext *s = ctx->priv;
int nb_samples;
nb_samples = FFMAX(1024, ((double)inlink->sample_rate / av_q2d(s->frame_rate)) + 0.5); s->nb_samples = FFMAX(1, ((double)inlink->sample_rate / av_q2d(s->frame_rate)) + 0.5);
inlink->partial_buf_size =
inlink->min_samples =
inlink->max_samples = nb_samples;
return 0; return 0;
} }
...@@ -365,6 +363,28 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *insamples) ...@@ -365,6 +363,28 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
return ff_filter_frame(outlink, av_frame_clone(s->outpicref)); return ff_filter_frame(outlink, av_frame_clone(s->outpicref));
} }
static int activate(AVFilterContext *ctx)
{
AVFilterLink *inlink = ctx->inputs[0];
AVFilterLink *outlink = ctx->outputs[0];
AudioVectorScopeContext *s = ctx->priv;
AVFrame *in;
int ret;
FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
ret = ff_inlink_consume_samples(inlink, s->nb_samples, s->nb_samples, &in);
if (ret < 0)
return ret;
if (ret > 0)
return filter_frame(inlink, in);
FF_FILTER_FORWARD_STATUS(inlink, outlink);
FF_FILTER_FORWARD_WANTED(outlink, inlink);
return FFERROR_NOT_READY;
}
static av_cold void uninit(AVFilterContext *ctx) static av_cold void uninit(AVFilterContext *ctx)
{ {
AudioVectorScopeContext *s = ctx->priv; AudioVectorScopeContext *s = ctx->priv;
...@@ -377,7 +397,6 @@ static const AVFilterPad audiovectorscope_inputs[] = { ...@@ -377,7 +397,6 @@ static const AVFilterPad audiovectorscope_inputs[] = {
.name = "default", .name = "default",
.type = AVMEDIA_TYPE_AUDIO, .type = AVMEDIA_TYPE_AUDIO,
.config_props = config_input, .config_props = config_input,
.filter_frame = filter_frame,
}, },
{ NULL } { NULL }
}; };
...@@ -397,6 +416,7 @@ AVFilter ff_avf_avectorscope = { ...@@ -397,6 +416,7 @@ AVFilter ff_avf_avectorscope = {
.uninit = uninit, .uninit = uninit,
.query_formats = query_formats, .query_formats = query_formats,
.priv_size = sizeof(AudioVectorScopeContext), .priv_size = sizeof(AudioVectorScopeContext),
.activate = activate,
.inputs = audiovectorscope_inputs, .inputs = audiovectorscope_inputs,
.outputs = audiovectorscope_outputs, .outputs = audiovectorscope_outputs,
.priv_class = &avectorscope_class, .priv_class = &avectorscope_class,
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment