Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Derek Buitenhuis
On 12/31/2017 5:21 AM, Aman Gupta wrote: > I really don't think it makes any sense in this case, since it is expected > that the av_opt_get will fail on non-http io contexts. It doesn't matter if > the failure is due to AVERROR_OPTION_NOT_FOUND or AVERROR(ENOMEM). The only > thing that matters is

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Aman Gupta
On Sat, Dec 30, 2017 at 2:14 PM Derek Buitenhuis wrote: > On 12/30/2017 5:50 PM, Aman Gupta wrote: > > There is already a check in place to prevent strncmp from being called > with > > NULL. > > The point before still holds. > > Are people really arguing against

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Steven Liu
> 在 2017年12月31日，上午6:14，Derek Buitenhuis 写道： > >> On 12/30/2017 5:50 PM, Aman Gupta wrote: >> There is already a check in place to prevent strncmp from being called with >> NULL. > > The point before still holds. > > Are people really arguing against consistent use

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Derek Buitenhuis
On 12/30/2017 5:50 PM, Aman Gupta wrote: > There is already a check in place to prevent strncmp from being called with > NULL. The point before still holds. Are people really arguing against consistent use of error checking? Inconsistent standards of error checking are how bugs and security

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Aman Gupta
On Sat, Dec 30, 2017 at 8:34 AM Derek Buitenhuis wrote: > On 12/30/2017 4:31 PM, Nicolas George wrote: > > Does it really matter? If av_opt_get() fails for any reason, > > http_multiple will just be false, which would let the processing > > continue, only in a

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Derek Buitenhuis
On 12/30/2017 4:31 PM, Nicolas George wrote: > Does it really matter? If av_opt_get() fails for any reason, > http_multiple will just be false, which would let the processing > continue, only in a slightly degraded manner that was the norm a few > months ago. I contend that checking errors should

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Nicolas George
Derek Buitenhuis (2017-12-30): > On 12/30/2017 12:42 PM, Steven Liu wrote: > > av_opt_get(v->input, "http_version", AV_OPT_SEARCH_CHILDREN, > > _version_opt); > > c->http_multiple = http_version_opt && strncmp((const char > > *)http_version_opt, "1.1", 3) == 0; > > +

Re: [FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Derek Buitenhuis
On 12/30/2017 12:42 PM, Steven Liu wrote: > av_opt_get(v->input, "http_version", AV_OPT_SEARCH_CHILDREN, > _version_opt); > c->http_multiple = http_version_opt && strncmp((const char > *)http_version_opt, "1.1", 3) == 0; > +av_free(http_version_opt); Looks OK, but the

[FFmpeg-devel] [PATCH] avformat/hls: release mem resource to fix memleak

2017-12-30 Thread Steven Liu
fix CID: 1426991 Signed-off-by: Steven Liu --- libavformat/hls.c | 1 + 1 file changed, 1 insertion(+) diff --git a/libavformat/hls.c b/libavformat/hls.c index dccc7c7dd2..9918d1af74 100644 --- a/libavformat/hls.c +++ b/libavformat/hls.c @@ -1475,6 +1475,7 @@ reload: