[FFmpeg-devel] [PATCH] resend: add segment_path for save segment movie

Steven Liu lingjiujianke at gmail.com
Tue Oct 29 02:00:03 CET 2013


在 2013年10月28日,下午5:26,Stefano Sabatini <stefasab at gmail.com> 写道:

> [resending since I sent it privately]
> 
> In date Sunday 2013-10-27 22:55:57 +0800, Steven Liu wrote:
>> Create video file in segment_path, and *DO NOT WANT* write the path into the
>> list file, use "-segment_path path" can do it. If *WANT* write the path into the list, 
>> *DO NOT USE* "-segment_path path".
> [...]
>> ---
>> libavformat/segment.c | 15 +++++++++++++--
>> 1 file changed, 13 insertions(+), 2 deletions(-)
>> 
>> diff --git a/libavformat/segment.c b/libavformat/segment.c
>> index 05e29d4..c0e3ca5 100644
>> --- a/libavformat/segment.c
>> +++ b/libavformat/segment.c
>> @@ -76,6 +76,8 @@ typedef struct {
>>    AVIOContext *list_pb;  ///< list file put-byte context
>>    char *time_str;        ///< segment duration specification string
>>    int64_t time;          ///< segment duration
>> +  char *path;            ///< parent path for segment create
>> +  char *seg_file_path;   ///< segment file create full path
>> 
>>    char *times_str;       ///< segment times specification string
>>    int64_t *times;        ///< list of segment interval specification
>> @@ -167,6 +169,14 @@ static int set_segment_filename(AVFormatContext *s)
>>        return AVERROR(EINVAL);
>>    }
>>    av_strlcpy(seg->cur_entry.filename, oc->filename, sizeof(seg->cur_entry.filename));
>> +
>> +  /* If no -segment_path option, use the oc->filename */
>> +  if (!seg->path) {
>> +      seg->seg_file_path = av_asprintf("%s", oc->filename);
>> +  } else {
>> +      seg->seg_file_path = av_asprintf("%s/%s", seg->path, oc->filename);
>> +  }
>> +
>>    return 0;
>> }
>> 
>> @@ -188,7 +198,7 @@ static int segment_start(AVFormatContext *s, int write_header)
>>    if ((err = set_segment_filename(s)) < 0)
>>        return err;
>> 
>> -   if ((err = avio_open2(&oc->pb, oc->filename, AVIO_FLAG_WRITE,
>> +  if ((err = avio_open2(&oc->pb, seg->seg_file_path, AVIO_FLAG_WRITE,
>>                          &s->interrupt_callback, NULL)) < 0)
>>        return err;
>> 
>> @@ -599,7 +609,7 @@ static int seg_write_header(AVFormatContext *s)
>>        goto fail;
>> 
>>    if (seg->write_header_trailer) {
>> -        if ((ret = avio_open2(&oc->pb, oc->filename, AVIO_FLAG_WRITE,
>> +       if ((ret = avio_open2(&oc->pb, seg->seg_file_path, AVIO_FLAG_WRITE,
>>                              &s->interrupt_callback, NULL)) < 0)
>>            goto fail;
> 
> I don't think this approach is correct. Indeed we should distinguish
> the filename where the file is written (oc->filename) from the
> filename which is displayed on the output list file.
But the oc->filename is displayed on the output list file with the full absolute path,
some scene just want create the mpegts files into the absolute path and display 
the mpegts files name in the list file. So, this maybe can not support on ffmpeg now,
lots of people just use ffmpeg and some split tools to support this function.
I think this can support on ffmpeg or segment.c, now this patch come :-)
> 
> Note that your approach won't work for example on Windows, where "/"
> is not recognized as a file separator.
This is a problem, i think i will fix it. :-)
> 
> I see two solutions.
> 
> 1. we *always* print on the list file the segment file basename (the
>   filename stripped of the prefix path), as it is done in hlsenc.c
hlsenc is just can be used at the HLS, for real time play AV. There only have few AV segments
displayed in the list file. there have a scene for transform from the realtime AV stream into local 
and save all the mpegts name into the list file for display. this function maybe must use segment.c
because this function hls can not support.
> 2. we allow to specify a pattern for the name of the file written on
>   the list file (probably overkill), or at least allow to specify a
>   prefix.
Yes, just want to implementation this way, i want to support this :-)
Let me try to send a new patch later :-)
> 
> What do you think is good enough for you?



Steven





More information about the ffmpeg-devel mailing list