On Fri, Dec 29, 2017 at 2:45 PM, Nicolas George <geo...@nsup.org> wrote:
> Aman Gupta (2017-12-29): > > From: Aman Gupta <a...@tmm1.net> > > > > This fixes a segfault when streaming from an HLS playlist which uses > > crypto, by ensuring ff_http_do_new_request will never be invoked on a > > non-http URLContext. > > > > Additionally it simplifies the demuxer by removing http.h and all the > > http specific code in open_url_keepalive. > > It it awfully specific for something quite minor. And it is messing with > the public API, that requires careful design. > I understand the public API should not be changed lightly, but I believe this approach is far superior to the mess of http specific checks currently present all over hls.c. It also makes it easier in the future to add http keepalive support to other consumers like the dash demuxer and the crypto protocol. > > I suggest to have ff_http_do_new_request() check that its argument is > really a HTTP context and return EINVAL or ENOSYS otherwise. > I'll go ahead and submit another patchset using this approach to fix the immediate segfault while we consider a larger API change. Thanks, Aman > > Regards, > > -- > Nicolas George > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
