encode: rewrite half of it

The main change is that we wait with opening the muxer ("writing
headers") until we have data from all streams. This fixes race
conditions at init due to broken assumptions in the old code.

This also changes a lot of other stuff. I found and fixed a few API
violations (often things for which better mechanisms were invented, and
the old ones are not valid anymore). I try to get away from the public
mutex and shared fields in encode_lavc_context. For now it's still
needed for some timestamp-related fields, but most are gone. It also
removes some bad code duplication between audio and video paths.

4 files changed, 557 insertions, 706 deletions

diff --git a/common/common.h b/common/common.h
index 14a9973371..dfb2ba7b7a 100644
--- a/common/common.h
+++ b/common/common.h
@@ -106,4 +106,15 @@ char *mp_tprintf_buf(char *buf, size_t buf_size, const char *format, ...)
 
 char **mp_dup_str_array(void *tctx, char **s);
 
+// We generally do not handle allocation failure of small malloc()s. This would
+// create a large number of rarely tested code paths, which would probably
+// regress and cause security issues. We prefer to fail fast.
+// This macro generally behaves like an assert(), except it will make sure to
+// kill the process even with NDEBUG.
+#define MP_HANDLE_OOM(x) do {   \
+        assert(x);              \
+        if (!(x))               \
+            abort();            \
+    } while (0)
+
 #endif /* MPLAYER_MPCOMMON_H */
diff --git a/common/encode.h b/common/encode.h
index c29cb3bc67..fcf4a8317e 100644
--- a/common/encode.h
+++ b/common/encode.h
@@ -54,17 +54,16 @@ struct encode_opts {
     char **remove_metadata;
 };
 
-// interface for mplayer.c
-struct encode_lavc_context *encode_lavc_init(struct encode_opts *options,
-                                             struct mpv_global *global);
-void encode_lavc_free(struct encode_lavc_context *ctx);
+// interface for player core
+struct encode_lavc_context *encode_lavc_init(struct mpv_global *global);
+bool encode_lavc_free(struct encode_lavc_context *ctx);
 void encode_lavc_discontinuity(struct encode_lavc_context *ctx);
 bool encode_lavc_showhelp(struct mp_log *log, struct encode_opts *options);
 int encode_lavc_getstatus(struct encode_lavc_context *ctx, char *buf, int bufsize, float relative_position);
-void encode_lavc_expect_stream(struct encode_lavc_context *ctx, int mt);
+void encode_lavc_expect_stream(struct encode_lavc_context *ctx,
+                               enum stream_type type);
 void encode_lavc_set_metadata(struct encode_lavc_context *ctx,
                               struct mp_tags *metadata);
-void encode_lavc_set_video_fps(struct encode_lavc_context *ctx, float fps);
 void encode_lavc_set_audio_pts(struct encode_lavc_context *ctx, double pts);
 bool encode_lavc_didfail(struct encode_lavc_context *ctx); // check if encoding failed
 
diff --git a/common/encode_lavc.c b/common/encode_lavc.c
index 905c30dc54..6f8378ac7f 100644
--- a/common/encode_lavc.c
+++ b/common/encode_lavc.c
@@ -21,6 +21,7 @@
  */
 
 #include <libavutil/avutil.h>
+#include <libavutil/timestamp.h>
 
 #include "config.h"
 #include "encode_lavc.h"
@@ -28,6 +29,7 @@
 #include "common/global.h"
 #include "common/msg.h"
 #include "common/msg_control.h"
+#include "options/m_config.h"
 #include "options/m_option.h"
 #include "options/options.h"
 #include "osdep/timer.h"
@@ -35,6 +37,44 @@
 #include "mpv_talloc.h"
 #include "stream/stream.h"
 
+struct encode_priv {
+    struct mp_log *log;
+
+    // --- All fields are protected by encode_lavc_context.lock
+
+    bool failed;
+
+    struct mp_tags *metadata;
+
+    AVFormatContext *muxer;
+
+    bool header_written; // muxer was initialized
+
+    struct mux_stream **streams;
+    int num_streams;
+
+    // Temporary queue while muxer is not initialized.
+    AVPacket **packets;
+    int num_packets;
+
+    // Statistics
+    double t0;
+
+    long long abytes;
+    long long vbytes;
+
+    unsigned int frames;
+    double audioseconds;
+};
+
+struct mux_stream {
+    int index;                      // index of this into p->streams[]
+    struct encode_lavc_context *ctx;
+    enum AVMediaType codec_type;
+    AVRational encoder_timebase;    // packet timestamps from encoder
+    AVStream *st;
+};
+
 #define OPT_BASE_STRUCT struct encode_opts
 const struct m_sub_options encode_config = {
     .opts = (const m_option_t[]) {
@@ -71,53 +111,23 @@ const struct m_sub_options encode_config = {
     },
 };
 
-static bool value_has_flag(const char *value, const char *flag)
-{
-    bool state = true;
-    bool ret = false;
-    while (*value) {
-        size_t l = strcspn(value, "+-");
-        if (l == 0) {
-            state = (*value == '+');
-            ++value;
-        } else {
-            if (l == strlen(flag))
-                if (!memcmp(value, flag, l))
-                    ret = state;
-            value += l;
-        }
-    }
-    return ret;
-}
-
-#define CHECK_FAIL(ctx, val) \
-    if (ctx && (ctx->failed)) { \
-        MP_ERR(ctx, \
-               "Called a function on a %s encoding context. Bailing out.\n", \
-               ctx->failed ? "failed" : "finished"); \
-        return val; \
-    }
+static void write_remaining_packets(struct encode_lavc_context *ctx);
 
-#define CHECK_FAIL_UNLOCK(ctx, val) \
-    if (ctx && (ctx->failed)) { \
-        MP_ERR(ctx, \
-               "Called a function on a %s encoding context. Bailing out.\n", \
-               ctx->failed ? "failed" : "finished"); \
-        pthread_mutex_unlock(&ctx->lock); \
-        return val; \
-    }
-
-int encode_lavc_available(struct encode_lavc_context *ctx)
+struct encode_lavc_context *encode_lavc_init(struct mpv_global *global)
 {
-    CHECK_FAIL(ctx, 0);
-    return ctx && ctx->avc;
-}
+    struct encode_lavc_context *ctx = talloc_ptrtype(NULL, ctx);
+    *ctx = (struct encode_lavc_context){
+        .global = global,
+        .options = mp_get_config_group(ctx, global, &encode_config),
+        .priv = talloc_zero(ctx, struct encode_priv),
+        .log = mp_log_new(ctx, global->log, "encode"),
+    };
+    pthread_mutex_init(&ctx->lock, NULL);
 
-struct encode_lavc_context *encode_lavc_init(struct encode_opts *options,
-                                             struct mpv_global *global)
-{
-    struct encode_lavc_context *ctx;
-    const char *filename = options->file;
+    struct encode_priv *p = ctx->priv;
+    p->log = ctx->log;
+
+    const char *filename = ctx->options->file;
 
     // STUPID STUPID STUPID STUPID avio
     // does not support "-" as file name to mean stdin/stdout
@@ -131,109 +141,46 @@ struct encode_lavc_context *encode_lavc_init(struct encode_opts *options,
             !strcmp(filename, "pipe:1")))
         mp_msg_force_stderr(global, true);
 
-    ctx = talloc_zero(NULL, struct encode_lavc_context);
-    pthread_mutex_init(&ctx->lock, NULL);
-    ctx->log = mp_log_new(ctx, global->log, "encode-lavc");
-    ctx->global = global;
     encode_lavc_discontinuity(ctx);
-    ctx->options = options;
-
-    ctx->avc = avformat_alloc_context();
-
-    if (ctx->options->format) {
-        char *tok;
-        const char *in = ctx->options->format;
-        while (*in) {
-            tok = av_get_token(&in, ",");
-            ctx->avc->oformat = av_guess_format(tok, filename, NULL);
-            av_free(tok);
-            if (ctx->avc->oformat)
-                break;
-            if (*in)
-                ++in;
-        }
-    } else {
-        ctx->avc->oformat = av_guess_format(NULL, filename, NULL);
-    }
 
-    if (!ctx->avc->oformat) {
-        encode_lavc_fail(ctx, "format not found\n");
-        return NULL;
-    }
+    p->muxer = avformat_alloc_context();
+    MP_HANDLE_OOM(p->muxer);
 
-    ctx->avc->url = av_strdup(filename);
-
-    mp_set_avdict(&ctx->foptions, ctx->options->fopts);
-
-    if (ctx->options->vcodec) {
-        char *tok;
-        const char *in = ctx->options->vcodec;
-        while (*in) {
-            tok = av_get_token(&in, ",");
-            ctx->vc = avcodec_find_encoder_by_name(tok);
-            av_free(tok);
-            if (ctx->vc && ctx->vc->type != AVMEDIA_TYPE_VIDEO)
-                ctx->vc = NULL;
-            if (ctx->vc)
-                break;
-            if (*in)
-                ++in;
-        }
+    if (ctx->options->format && ctx->options->format[0]) {
+        ctx->oformat = av_guess_format(ctx->options->format, filename, NULL);
     } else {
-        ctx->vc = avcodec_find_encoder(av_guess_codec(ctx->avc->oformat, NULL,
-                                                      ctx->avc->url, NULL,
-                                                      AVMEDIA_TYPE_VIDEO));
+        ctx->oformat = av_guess_format(NULL, filename, NULL);
     }
 
-    if (ctx->options->acodec) {
-        char *tok;
-        const char *in = ctx->options->acodec;
-        while (*in) {
-            tok = av_get_token(&in, ",");
-            ctx->ac = avcodec_find_encoder_by_name(tok);
-            av_free(tok);
-            if (ctx->ac && ctx->ac->type != AVMEDIA_TYPE_AUDIO)
-                ctx->ac = NULL;
-            if (ctx->ac)
-                break;
-            if (*in)
-                ++in;
-        }
-    } else {
-        ctx->ac = avcodec_find_encoder(av_guess_codec(ctx->avc->oformat, NULL,
-                                                      ctx->avc->url, NULL,
-                                                      AVMEDIA_TYPE_AUDIO));
-    }
-
-    if (!ctx->vc && !ctx->ac) {
-        encode_lavc_fail(ctx, "neither audio nor video codec was found\n");
-        return NULL;
+    if (!ctx->oformat) {
+        MP_FATAL(ctx, "format not found\n");
+        goto fail;
     }
 
-    /* taken from ffmpeg unchanged
-     * TODO turn this into an option if anyone needs this */
+    p->muxer->oformat = ctx->oformat;
 
-    ctx->avc->max_delay = 0.7 * AV_TIME_BASE;
-
-    ctx->abytes = 0;
-    ctx->vbytes = 0;
-    ctx->frames = 0;
-
-    if (options->video_first)
-        ctx->video_first = true;
-    if (options->audio_first)
-        ctx->audio_first = true;
+    p->muxer->url = av_strdup(filename);
+    MP_HANDLE_OOM(p->muxer->url);
 
     return ctx;
+
+fail:
+    p->failed = true;
+    encode_lavc_free(ctx);
+    return NULL;
 }
 
 void encode_lavc_set_metadata(struct encode_lavc_context *ctx,
                               struct mp_tags *metadata)
 {
+    struct encode_priv *p = ctx->priv;
+
+    pthread_mutex_lock(&ctx->lock);
+
     if (ctx->options->copy_metadata) {
-        ctx->metadata = metadata;
+        p->metadata = mp_tags_dup(ctx, metadata);
     } else {
-        ctx->metadata = talloc_zero(ctx, struct mp_tags);
+        p->metadata = talloc_zero(ctx, struct mp_tags);
     }
 
     if (ctx->options->set_metadata) {
@@ -242,7 +189,7 @@ void encode_lavc_set_metadata(struct encode_lavc_context *ctx,
         for (int n = 0; kv[n * 2]; n++) {
             MP_VERBOSE(ctx, "setting metadata value '%s' for key '%s'\n",
                        kv[n*2 + 0], kv[n*2 +1]);
-            mp_tags_set_str(ctx->metadata, kv[n*2 + 0], kv[n*2 +1]);
+            mp_tags_set_str(p->metadata, kv[n*2 + 0], kv[n*2 +1]);
         }
     }
 
@@ -251,152 +198,52 @@ void encode_lavc_set_metadata(struct encode_lavc_context *ctx,
         // Remove all user-provided metadata tags
         for (int n = 0; k[n]; n++) {
             MP_VERBOSE(ctx, "removing metadata key '%s'\n", k[n]);
-            mp_tags_remove_str(ctx->metadata, k[n]);
+            mp_tags_remove_str(p->metadata, k[n]);
         }
     }
+
+    pthread_mutex_unlock(&ctx->lock);
 }
 
-int encode_lavc_start(struct encode_lavc_context *ctx)
+bool encode_lavc_free(struct encode_lavc_context *ctx)
 {
-    AVDictionaryEntry *de;
-
-    if (ctx->header_written < 0)
-        return 0;
-    if (ctx->header_written > 0)
-        return 1;
-
-    CHECK_FAIL(ctx, 0);
-
-    if (ctx->expect_video && ctx->vcc == NULL) {
-        if (ctx->avc->oformat->video_codec != AV_CODEC_ID_NONE ||
-            ctx->options->vcodec)
-        {
-            encode_lavc_fail(ctx,"no video stream succeeded - invalid codec?\n");
-            return 0;
-        }
-    }
-
-    if (ctx->expect_audio && ctx->acc == NULL) {
-        if (ctx->avc->oformat->audio_codec != AV_CODEC_ID_NONE ||
-            ctx->options->acodec)
-        {
-            encode_lavc_fail(ctx, "no audio stream succeeded - invalid codec?\n");
-            return 0;
-        }
-    }
-
-    ctx->header_written = -1;
+    bool res = true;
+    if (!ctx)
+        return res;
 
-    if (!(ctx->avc->oformat->flags & AVFMT_NOFILE)) {
-        MP_INFO(ctx, "Opening output file: %s\n", ctx->avc->url);
+    struct encode_priv *p = ctx->priv;
 
-        if (avio_open(&ctx->avc->pb, ctx->avc->url, AVIO_FLAG_WRITE) < 0) {
-            encode_lavc_fail(ctx, "could not open '%s'\n",
-                             ctx->avc->url);
-            return 0;
-        }
+    if (!p->failed && !p->header_written) {
+        MP_FATAL(p, "no data written to target file\n");
+        p->failed = true;
     }
 
-    ctx->t0 = mp_time_sec();
+    write_remaining_packets(ctx);
 
-    MP_INFO(ctx, "Opening muxer: %s [%s]\n",
-            ctx->avc->oformat->long_name, ctx->avc->oformat->name);
+    if (!p->failed && p->header_written) {
+        if (av_write_trailer(p->muxer) < 0)
+            MP_ERR(p, "error writing trailer\n");
 
-    if (ctx->metadata) {
-        for (int i = 0; i < ctx->metadata->num_keys; i++) {
-            av_dict_set(&ctx->avc->metadata,
-                ctx->metadata->keys[i], ctx->metadata->values[i], 0);
-        }
-    }
+        MP_INFO(p, "video: encoded %lld bytes\n", p->vbytes);
+        MP_INFO(p, "audio: encoded %lld bytes\n", p->abytes);
 
-    if (avformat_write_header(ctx->avc, &ctx->foptions) < 0) {
-        encode_lavc_fail(ctx, "could not write header\n");
-        return 0;
+        MP_INFO(p, "muxing overhead %lld bytes\n",
+                (long long)(avio_size(p->muxer->pb) - p->vbytes - p->abytes));
     }
 
-    for (de = NULL; (de = av_dict_get(ctx->foptions, "", de,
-                                      AV_DICT_IGNORE_SUFFIX));)
-    {
-        MP_WARN(ctx, "ofopts: key '%s' not found.\n", de->key);
+    if (avio_closep(&p->muxer->pb) < 0 && !p->failed) {
+        MP_ERR(p, "Closing file failed\n");
+        p->failed = true;
     }
-    av_dict_free(&ctx->foptions);
 
-    ctx->header_written = 1;
-    return 1;
-}
+    avformat_free_context(p->muxer);
 
-void encode_lavc_free(struct encode_lavc_context *ctx)
-{
-    unsigned i;
-
-    if (!ctx)
-        return;
-
-    if (ctx->avc) {
-        if (ctx->header_written > 0)
-            av_write_trailer(ctx->avc);  // this is allowed to fail
-
-        if (ctx->vcc) {
-            if (ctx->twopass_bytebuffer_v) {
-                char *stats = ctx->vcc->stats_out;
-                if (stats)
-                    stream_write_buffer(ctx->twopass_bytebuffer_v,
-                                        stats, strlen(stats));
-            }
-            avcodec_free_context(&ctx->vcc);
-        }
-
-        if (ctx->acc) {
-            if (ctx->twopass_bytebuffer_a) {
-                char *stats = ctx->acc->stats_out;
-                if (stats)
-                    stream_write_buffer(ctx->twopass_bytebuffer_a,
-                                        stats, strlen(stats));
-            }
-            avcodec_free_context(&ctx->acc);
-        }
-
-        for (i = 0; i < ctx->avc->nb_streams; i++) {
-            av_free(ctx->avc->streams[i]->info);
-            av_free(ctx->avc->streams[i]);
-        }
-        ctx->vst = NULL;
-        ctx->ast = NULL;
-
-        if (ctx->twopass_bytebuffer_v) {
-            free_stream(ctx->twopass_bytebuffer_v);
-            ctx->twopass_bytebuffer_v = NULL;
-        }
-
-        if (ctx->twopass_bytebuffer_a) {
-            free_stream(ctx->twopass_bytebuffer_a);
-            ctx->twopass_bytebuffer_a = NULL;
-        }
-
-        MP_INFO(ctx, "vo-lavc: encoded %lld bytes\n",
-               ctx->vbytes);
-        MP_INFO(ctx, "ao-lavc: encoded %lld bytes\n",
-               ctx->abytes);
-        if (ctx->avc->pb) {
-            MP_INFO(ctx, "muxing overhead %lld bytes\n",
-                   (long long) (avio_size(ctx->avc->pb) - ctx->vbytes
-                                                        - ctx->abytes));
-            avio_close(ctx->avc->pb);
-        }
-
-        av_free(ctx->avc);
-        ctx->avc = NULL;
-    }
+    res = !p->failed;
 
     pthread_mutex_destroy(&ctx->lock);
     talloc_free(ctx);
-}
 
-void encode_lavc_set_video_fps(struct encode_lavc_context *ctx, float fps)
-{
-    pthread_mutex_lock(&ctx->lock);
-    ctx->vo_fps = fps;
-    pthread_mutex_unlock(&ctx->lock);
+    return res;
 }
 
 void encode_lavc_set_audio_pts(struct encode_lavc_context *ctx, double pts)
@@ -409,399 +256,230 @@ void encode_lavc_set_audio_pts(struct encode_lavc_context *ctx, double pts)
     }
 }
 
-static void encode_2pass_prepare(struct encode_lavc_context *ctx,
-                                 AVDictionary **dictp,
-                                 AVStream *stream,
-                                 AVCodecContext *codec,
-                                 struct stream **bytebuf,
-                                 const char *prefix)
+// called locked
+static void write_remaining_packets(struct encode_lavc_context *ctx)
 {
-    if (!*bytebuf) {
-        char buf[1024];
-        AVDictionaryEntry *de = av_dict_get(ctx->voptions, "flags", NULL, 0);
-
-        snprintf(buf, sizeof(buf), "%s-%s-pass1.log", ctx->avc->url, prefix);
-
-        if (value_has_flag(de ? de->value : "", "pass2")) {
-            if (!(*bytebuf = stream_open(buf, ctx->global))) {
-                MP_WARN(ctx, "%s: could not open '%s', "
-                       "disabling 2-pass encoding at pass 2\n", prefix, buf);
-                codec->flags &= ~AV_CODEC_FLAG_PASS2;
-                av_dict_set(dictp, "flags", "-pass2", AV_DICT_APPEND);
-            } else {
-                struct bstr content = stream_read_complete(*bytebuf, NULL,
-                                                           1000000000);
-                if (content.start == NULL) {
-                    MP_WARN(ctx, "%s: could not read '%s', "
-                           "disabling 2-pass encoding at pass 1\n",
-                           prefix, ctx->avc->url);
-                } else {
-                    content.start[content.len] = 0;
-                    codec->stats_in = content.start;
-                }
-                free_stream(*bytebuf);
-                *bytebuf = NULL;
-            }
+    struct encode_priv *p = ctx->priv;
+
+    if (!p->header_written && !p->failed)
+        return; // wait until muxer initialization
+
+    for (int n = 0; n < p->num_packets; n++) {
+        AVPacket *pkt = p->packets[n];
+        MP_TARRAY_REMOVE_AT(p->packets, p->num_packets, 0);
+
+        if (p->failed) {
+            av_packet_free(&pkt);
+            continue;
         }
 
-        if (value_has_flag(de ? de->value : "", "pass1")) {
-            if (!(*bytebuf = open_output_stream(buf, ctx->global))) {
-                MP_WARN(ctx,
-                    "%s: could not open '%s', disabling "
-                    "2-pass encoding at pass 1\n",
-                    prefix, ctx->avc->url);
-                av_dict_set(dictp, "flags", "-pass1", AV_DICT_APPEND);
-            }
+        struct mux_stream *s = p->streams[pkt->stream_index];
+
+        pkt->stream_index = s->st->index;
+        assert(s->st == p->muxer->streams[pkt->stream_index]);
+
+        av_packet_rescale_ts(pkt, s->encoder_timebase, s->st->time_base);
+
+        switch (s->st->codecpar->codec_type) {
+        case AVMEDIA_TYPE_VIDEO:
+            p->vbytes += pkt->size;
+            p->frames += 1;
+            break;
+        case AVMEDIA_TYPE_AUDIO:
+            p->abytes += pkt->size;
+            p->audioseconds += pkt->duration
+                * (double)s->st->time_base.num
+                / (double)s->st->time_base.den;
+            break;
+        }
+
+        if (av_interleaved_write_frame(p->muxer, pkt) < 0) {
+            MP_ERR(p, "Writing packet failed.\n");
+            p->failed = true;
         }
     }
+
+    p->num_packets = 0;
 }
 
-int encode_lavc_alloc_stream(struct encode_lavc_context *ctx,
-                             enum AVMediaType mt,
-                             AVStream **stream_out,
-                             AVCodecContext **codec_out)
+// called locked
+static void maybe_init_muxer(struct encode_lavc_context *ctx)
 {
-    AVDictionaryEntry *de;
+    struct encode_priv *p = ctx->priv;
 
-    *stream_out = NULL;
-    *codec_out = NULL;
+    if (p->header_written || p->failed)
+        return;
 
-    CHECK_FAIL(ctx, -1);
+    // Check if all streams were initialized yet. We need data to know the
+    // AVStream parameters, so we wait for data from _all_ streams before
+    // starting.
+    for (int n = 0; n < p->num_streams; n++) {
+        if (!p->streams[n]->st) {
+            int max_num = 50;
+            if (p->num_packets > max_num) {
+                MP_FATAL(p, "no data on stream %d, even though other streams "
+                         "produced more than %d packets.\n", n, p->num_packets);
+                goto failed;
+            }
+            return;
+        }
+    }
 
-    if (ctx->header_written)
-        return -1;
+    if (!(p->muxer->oformat->flags & AVFMT_NOFILE)) {
+        MP_INFO(p, "Opening output file: %s\n", p->muxer->url);
 
-    if (ctx->avc->nb_streams == 0) {
-        // if this stream isn't stream #0, allocate a dummy stream first for
-        // the next call to use
-        if (mt == AVMEDIA_TYPE_VIDEO && ctx->audio_first) {
-            MP_INFO(ctx, "vo-lavc: preallocated audio stream for later use\n");
-            ctx->ast = avformat_new_stream(
-                ctx->avc, NULL);  // this one is AVMEDIA_TYPE_UNKNOWN for now
-        }
-        if (mt == AVMEDIA_TYPE_AUDIO && ctx->video_first) {
-            MP_INFO(ctx, "ao-lavc: preallocated video stream for later use\n");
-            ctx->vst = avformat_new_stream(
-                ctx->avc, NULL);  // this one is AVMEDIA_TYPE_UNKNOWN for now
+        if (avio_open(&p->muxer->pb, p->muxer->url, AVIO_FLAG_WRITE) < 0) {
+            MP_FATAL(p, "could not open '%s'\n", p->muxer->url);
+            goto failed;
         }
     }
 
-    // already have a stream of that type (this cannot really happen)?
-    switch (mt) {
-    case AVMEDIA_TYPE_VIDEO:
-        if (ctx->vcc != NULL)
-            return -1;
-        if (ctx->vst == NULL)
-            ctx->vst = avformat_new_stream(ctx->avc, NULL);
-        break;
-    case AVMEDIA_TYPE_AUDIO:
-        if (ctx->acc != NULL)
-            return -1;
-        if (ctx->ast == NULL)
-            ctx->ast = avformat_new_stream(ctx->avc, NULL);
-        break;
-    default:
-        encode_lavc_fail(ctx, "requested invalid stream type\n");
-        return -1;
-    }
+    p->t0 = mp_time_sec();
 
-    if (ctx->timebase.den == 0) {
-        AVRational r;
+    MP_INFO(p, "Opening muxer: %s [%s]\n",
+            p->muxer->oformat->long_name, p->muxer->oformat->name);
 
-        if (ctx->options->fps > 0)
-            r = av_d2q(ctx->options->fps, ctx->options->fps * 1001 + 2);
-        else if (ctx->options->autofps && ctx->vo_fps > 0) {
-            r = av_d2q(ctx->vo_fps, ctx->vo_fps * 1001 + 2);
-            MP_INFO(ctx, "option --ofps not specified "
-                "but --oautofps is active, using guess of %u/%u\n",
-                (unsigned)r.num, (unsigned)r.den);
-        } else {
-            // we want to handle:
-            //      1/25
-            //   1001/24000
-            //   1001/30000
-            // for this we would need 120000fps...
-            // however, mpeg-4 only allows 16bit values
-            // so let's take 1001/30000 out
-            r.num = 24000;
-            r.den = 1;
-            MP_INFO(ctx, "option --ofps not specified "
-                "and fps could not be inferred, using guess of %u/%u\n",
-                (unsigned)r.num, (unsigned)r.den);
+    if (p->metadata) {
+        for (int i = 0; i < p->metadata->num_keys; i++) {
+            av_dict_set(&p->muxer->metadata,
+                p->metadata->keys[i], p->metadata->values[i], 0);
         }
-
-        if (ctx->vc && ctx->vc->supported_framerates)
-            r = ctx->vc->supported_framerates[av_find_nearest_q_idx(r,
-                    ctx->vc->supported_framerates)];
-
-        ctx->timebase.num = r.den;
-        ctx->timebase.den = r.num;
     }
 
-    switch (mt) {
-    case AVMEDIA_TYPE_VIDEO:
-        if (!ctx->vc) {
-            if (ctx->avc->oformat->video_codec != AV_CODEC_ID_NONE ||
-                ctx->options->vcodec) {
-                encode_lavc_fail(ctx, "vo-lavc: encoder not found\n");
-            }
-            return -1;
-        }
-        ctx->vcc = avcodec_alloc_context3(ctx->vc);
+    AVDictionary *opts = NULL;
+    mp_set_avdict(&opts, ctx->options->fopts);
 
-        // Using codec->time_base is deprecated, but needed for older lavf.
-        ctx->vst->time_base = ctx->timebase;
-        ctx->vcc->time_base = ctx->timebase;
+    if (avformat_write_header(p->muxer, &opts) < 0) {
+        MP_FATAL(p, "Failed to initialize muxer.\n");
+        p->failed = true;
+    } else {
+        mp_avdict_print_unset(p->log, MSGL_WARN, opts);
+    }
 
-        ctx->voptions = NULL;
+    av_dict_free(&opts);
 
-        mp_set_avdict(&ctx->voptions, ctx->options->vopts);
+    if (p->failed)
+        goto failed;
 
-        de = av_dict_get(ctx->voptions, "global_quality", NULL, 0);
-        if (de)
-            av_dict_set(&ctx->voptions, "flags", "+qscale", AV_DICT_APPEND);
+    p->header_written = true;
 
-        if (ctx->avc->oformat->flags & AVFMT_GLOBALHEADER)
-            av_dict_set(&ctx->voptions, "flags", "+global_header", AV_DICT_APPEND);
+    write_remaining_packets(ctx);
+    return;
 
-        encode_2pass_prepare(ctx, &ctx->voptions, ctx->vst, ctx->vcc,
-                             &ctx->twopass_bytebuffer_v,
-                             "vo-lavc");
-        *stream_out = ctx->vst;
-        *codec_out = ctx->vcc;
-        return 0;
+failed:
+    p->failed = true;
+}
 
-    case AVMEDIA_TYPE_AUDIO:
-        if (!ctx->ac) {
-            if (ctx->avc->oformat->audio_codec != AV_CODEC_ID_NONE ||
-                ctx->options->acodec) {
-                encode_lavc_fail(ctx, "ao-lavc: encoder not found\n");
-            }
-            return -1;
-        }
-        ctx->acc = avcodec_alloc_context3(ctx->ac);
+// called locked
+static struct mux_stream *find_mux_stream(struct encode_lavc_context *ctx,
+                                          enum AVMediaType codec_type)
+{
+    struct encode_priv *p = ctx->priv;
 
-        // Using codec->time_base is deprecated, but needed for older lavf.
-        ctx->ast->time_base = ctx->timebase;
-        ctx->acc->time_base = ctx->timebase;
+    for (int n = 0; n < p->num_streams; n++) {
+        struct mux_stream *s = p->streams[n];
+        if (s->codec_type == codec_type)
+            return s;
+    }
 
-        ctx->aoptions = 0;
+    return NULL;
+}
 
-        mp_set_avdict(&ctx->aoptions, ctx->options->aopts);
+void encode_lavc_expect_stream(struct encode_lavc_context *ctx,
+                               enum stream_type type)
+{
+    struct encode_priv *p = ctx->priv;
 
-        de = av_dict_get(ctx->aoptions, "global_quality", NULL, 0);
-        if (de)
-            av_dict_set(&ctx->aoptions, "flags", "+qscale", AV_DICT_APPEND);
+    pthread_mutex_lock(&ctx->lock);
 
-        if (ctx->avc->oformat->flags & AVFMT_GLOBALHEADER)
-            av_dict_set(&ctx->aoptions, "flags", "+global_header", AV_DICT_APPEND);
+    enum AVMediaType codec_type = mp_to_av_stream_type(type);
 
-        encode_2pass_prepare(ctx, &ctx->aoptions, ctx->ast, ctx->acc,
-                             &ctx->twopass_bytebuffer_a,
-                             "ao-lavc");
+    // These calls are idempotent.
+    if (find_mux_stream(ctx, codec_type))
+        goto done;
 
-        *stream_out = ctx->ast;
-        *codec_out = ctx->acc;
-        return 0;
+    if (p->header_written) {
+        MP_ERR(p, "Cannot add a stream during encoding.\n");
+        p->failed = true;
+        goto done;
     }
 
-    // Unreachable.
-    return -1;
+    struct mux_stream *dst = talloc_ptrtype(p, dst);
+    *dst = (struct mux_stream){
+        .index = p->num_streams,
+        .ctx = ctx,
+        .codec_type = mp_to_av_stream_type(type),
+    };
+    MP_TARRAY_APPEND(p, p->streams, p->num_streams, dst);
+
+done:
+    pthread_mutex_unlock(&ctx->lock);
 }
 
-int encode_lavc_open_codec(struct encode_lavc_context *ctx,
-                           AVCodecContext *codec)
+// Signal that you are ready to encode (you provide the codec params etc. too).
+// This returns a muxing handle which you can use to add encodec packets.
+// Can be called only once per stream. info is copied by callee as needed.
+static struct mux_stream *encode_lavc_add_stream(struct encode_lavc_context *ctx,
+                                                 struct encoder_stream_info *info)
 {
-    AVDictionaryEntry *de;
-    int ret;
-
-    CHECK_FAIL(ctx, -1);
-
-    switch (codec->codec_type) {
-    case AVMEDIA_TYPE_VIDEO:
-        MP_INFO(ctx, "Opening video encoder: %s [%s]\n",
-                ctx->vc->long_name, ctx->vc->name);
-
-        if (ctx->vc->capabilities & AV_CODEC_CAP_EXPERIMENTAL) {
-            codec->strict_std_compliance = FF_COMPLIANCE_EXPERIMENTAL;
-            MP_WARN(ctx, "\n\n"
-                       "           ********************************************\n"
-                       "           **** Experimental VIDEO codec selected! ****\n"
-                       "           ********************************************\n\n"
-                       "This means the output file may be broken or bad.\n"
-                       "Possible reasons, problems, workarounds:\n"
-                       "- Codec implementation in ffmpeg/libav is not finished yet.\n"
-                       "     Try updating ffmpeg or libav.\n"
-                       "- Bad picture quality, blocks, blurriness.\n"
-                       "     Experiment with codec settings (--ovcopts) to maybe still get the\n"
-                       "     desired quality output at the expense of bitrate.\n"
-                       "- Slow compression.\n"
-                       "     Bear with it.\n"
-                       "- Crashes.\n"
-                       "     Happens. Try varying options to work around.\n"
-                       "If none of this helps you, try another codec in place of %s.\n\n",
-                   ctx->vc->name);
-        }
+    struct encode_priv *p = ctx->priv;
 
-        ret = avcodec_open2(codec, ctx->vc, &ctx->voptions);
-        if (ret >= 0)
-            ret = avcodec_parameters_from_context(ctx->vst->codecpar, codec);
-
-        // complain about all remaining options, then free the dict
-        for (de = NULL; (de = av_dict_get(ctx->voptions, "", de,
-                                          AV_DICT_IGNORE_SUFFIX));)
-            MP_WARN(ctx, "ovcopts: key '%s' not found.\n",
-                   de->key);
-        av_dict_free(&ctx->voptions);
-
-        break;
-    case AVMEDIA_TYPE_AUDIO:
-        MP_INFO(ctx, "Opening audio encoder: %s [%s]\n",
-                ctx->ac->long_name, ctx->ac->name);
-
-        if (ctx->ac->capabilities & AV_CODEC_CAP_EXPERIMENTAL) {
-            codec->strict_std_compliance = FF_COMPLIANCE_EXPERIMENTAL;
-            MP_WARN(ctx, "\n\n"
-                       "           ********************************************\n"
-                       "           **** Experimental AUDIO codec selected! ****\n"
-                       "           ********************************************\n\n"
-                       "This means the output file may be broken or bad.\n"
-                       "Possible reasons, problems, workarounds:\n"
-                       "- Codec implementation in ffmpeg/libav is not finished yet.\n"
-                       "     Try updating ffmpeg or libav.\n"
-                       "- Bad sound quality, noise, clicking, whistles, choppiness.\n"
-                       "     Experiment with codec settings (--oacopts) to maybe still get the\n"
-                       "     desired quality output at the expense of bitrate.\n"
-                       "- Slow compression.\n"
-                       "     Bear with it.\n"
-                       "- Crashes.\n"
-                       "     Happens. Try varying options to work around.\n"
-                       "If none of this helps you, try another codec in place of %s.\n\n",
-                   ctx->ac->name);
-        }
+    pthread_mutex_lock(&ctx->lock);
 
-        ret = avcodec_open2(codec, ctx->ac, &ctx->aoptions);
-        if (ret >= 0)
-            ret = avcodec_parameters_from_context(ctx->ast->codecpar, codec);
-
-        // complain about all remaining options, then free the dict
-        for (de = NULL; (de = av_dict_get(ctx->aoptions, "", de,
-                                          AV_DICT_IGNORE_SUFFIX));)
-            MP_WARN(ctx, "oacopts: key '%s' not found.\n",
-                   de->key);
-        av_dict_free(&ctx->aoptions);
-
-        break;
-    default:
-        ret = -1;
-        break;
+    struct mux_stream *dst = find_mux_stream(ctx, info->codecpar->codec_type);
+    if (!dst) {
+        MP_ERR(p, "Cannot add a stream at runtime.\n");
+        p->failed = true;
+        goto done;
+    }
+    if (dst->st) {
+        // Po