[FFmpeg-devel] [PATCH] avformat/rtsp: check return value of read in ff_rtsp_read_reply()

[Daniel Richard G.]

On Mon, 2017 May  8 18:04+0200, wm4 wrote:
> On Mon, 08 May 2017 11:24:19 -0400
> "Daniel Richard G." <skunk at iSKUNK.ORG> wrote:
> 
> > From 477cbd18b630365d612da173201c2e4ee763d7d4 Mon Sep 17 00:00:00 2001
> > From: Daniel Richard G <skunk at iSKUNK.ORG>
> > Date: Sun, 16 Apr 2017 23:12:53 -0400
> > Subject: [PATCH] avformat/rtsp: check return value of read in ff_rtsp_read_reply()
> > 
> > Signed-off-by: Daniel Richard G <skunk at iSKUNK.ORG>
> > ---
> >  libavformat/rtsp.c | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> > 
> > diff --git a/libavformat/rtsp.c b/libavformat/rtsp.c
> > index 261e970..da962fb 100644
> > --- a/libavformat/rtsp.c
> > +++ b/libavformat/rtsp.c
> > @@ -1218,7 +1218,11 @@ start:
> >          content = av_malloc(content_length + 1);
> >          if (!content)
> >              return AVERROR(ENOMEM);
> > -        ffurl_read_complete(rt->rtsp_hd, content, content_length);
> > +        ret = ffurl_read_complete(rt->rtsp_hd, content, content_length);
> > +        if (ret != content_length) {
> > +            av_freep(&content);
> > +            return AVERROR_EOF;
> > +        }
> >          content[content_length] = '\0';
> >      }
> >      if (content_ptr)
> 
> That's probably better than before, so LGTM.
> 
> Can ffurl_read_complete() return an error value as well? If so, it
> might be better to return the error value when that happens.

Agreed; my point of reference was the 1-byte read that occurs a few
lines up, under the "parse reply (XXX: use buffers)" comment. The
calling function already assumes AVERROR_EOF given a short read.


--Daniel


-- 
Daniel Richard G. || skunk at iSKUNK.ORG
My ASCII-art .sig got a bad case of Times New Roman.