[FFmpeg-devel] [PATCH 01/10] avformat/segment: Add segment_write_temp option

[softworkz .]

Hi Marton,


> -----Original Message-----
> From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of
> Marton Balint
> Sent: Freitag, 13. Juni 2025 22:38
> To: FFmpeg development discussions and patches <ffmpeg-
> devel at ffmpeg.org>
> Subject: Re: [FFmpeg-devel] [PATCH 01/10] avformat/segment: Add
> segment_write_temp option
> 
> 
> 
> On Fri, 13 Jun 2025, softworkz wrote:
> 
> > From: softworkz <softworkz at hotmail.com>
> >
> > Allows to write segments as temp files (.tmp) which
> > are renamed on completion.
> >
> > Signed-off-by: softworkz <softworkz at hotmail.com>
> > ---
> > libavformat/segment.c | 30 +++++++++++++++++++++++++++---
> > 1 file changed, 27 insertions(+), 3 deletions(-)
> >
> > diff --git a/libavformat/segment.c b/libavformat/segment.c
> > index 65323ec678..04e973a198 100644
> > --- a/libavformat/segment.c
> > +++ b/libavformat/segment.c
> > @@ -121,6 +121,7 @@ typedef struct SegmentContext {
> >     int   break_non_keyframes;
> >     int   write_empty;
> >
> > +    int segment_write_temp; ///< write segments as temp files
> and rename on completion
> >     int use_rename;
> >     char temp_list_filename[1024];
> >
> > @@ -226,6 +227,15 @@ static int
> set_segment_filename(AVFormatContext *s)
> >              seg->entry_prefix ? seg->entry_prefix : "",
> >              av_basename(oc->url));
> >
> > +    // Write segment as a temp file and rename on completion
> > +    if(seg->segment_write_temp) {
> > +        av_strlcatf(buf, sizeof(buf), ".tmp");
> > +        char *temp_name = av_strdup(buf);
> > +        if (!temp_name)
> > +            return AVERROR(ENOMEM);
> 
> You should use av_asprintf() directly instead of strlcatf() +
> av_strdup()

I could be wrong, but I was thinking that this way is more efficient,
because in the av_strdup case, the size of the needed allocation is  
already known, while I would suppose that av_asprintf() needs to 
allocate a larger size of memory because it cannot predict the eventual
size after applying all the formats. Am I on a wrong track?


> > +        ff_format_set_url(oc, temp_name);
> > +    }
> > +
> >     return 0;
> > }
> >
> > @@ -372,7 +382,7 @@ static int segment_end(AVFormatContext *s,
> int write_trailer, int is_last)
> >             SegmentListEntry *entry = av_mallocz(sizeof(*entry));
> >             if (!entry) {
> >                 ret = AVERROR(ENOMEM);
> > -                goto end;
> > +                goto fail;
> >             }
> >
> >             /* append new element */
> > @@ -393,7 +403,7 @@ static int segment_end(AVFormatContext *s,
> int write_trailer, int is_last)
> >             }
> >
> >             if ((ret = segment_list_open(s)) < 0)
> > -                goto end;
> > +                goto fail;
> >             for (entry = seg->segment_list_entries; entry; entry
> = entry->next)
> >                 segment_list_print_entry(seg->list_pb, seg-
> >list_type, entry, s);
> >             if (seg->list_type == LIST_TYPE_M3U8 && is_last)
> > @@ -450,7 +460,20 @@ static int segment_end(AVFormatContext *s,
> int write_trailer, int is_last)
> >         }
> >     }
> >
> > -end:
> > +    ff_format_io_close(oc, &oc->pb);
> > +
> > +    // Now rename the .tmp file to its actual name.
> > +    if (seg->segment_write_temp) {
> > +        char *final_filename = av_strndup(oc->url, strlen(oc-
> >url) - 4);
> > +        if (!final_filename)
> > +            return AVERROR(ENOMEM);
> 
> goto fail?

At that point,  ff_format_io_close(oc, &oc->pb) has 
already been called.

Here's the full block (after the patch):


    ff_format_io_close(oc, &oc->pb);

    // Now rename the .tmp file to its actual name.
    if (seg->segment_write_temp) {
        char *final_filename = av_strndup(oc->url, strlen(oc->url) - 4);
        if (!final_filename)
            return AVERROR(ENOMEM);
        ret = ff_rename(oc->url, final_filename, s);
        av_free(final_filename);
    }

    return ret;

fail:
    ff_format_io_close(oc, &oc->pb);

    return ret;
}


Thanks,
sw