[1.x] RTSP SETUP does not handle errors

[AWilco]

What version of Janus is this happening on?
52f86d5

When connecting to an RTSP source that doesn't respond with a 200 OK request, Janus assumes it did get a 200 OK and continues.

https://github.com/meetecho/janus-gateway/blob/db3300d9263b353b247cb3d9316cb1e87e83dc1a/src/plugins/janus_streaming.c#L7558-7570

My understanding is that this is because the code is not retrieved from the most recent request, so is still using the value from the original DESCRIBE request. The fix would be to get the code after checking that the request was sent ok (or assume the request was OK given we just succeeded in our DESCRIBE request).

      res = curl_easy_perform(curl);
		if(res != CURLE_OK) {
			JANUS_LOG(LOG_ERR, "Couldn't send SETUP request: %s\n", curl_easy_strerror(res));
			g_strfreev(parts);
			curl_easy_cleanup(curl);
			g_free(curldata->buffer);
			g_free(curldata);
			if(video_fds.fd != -1) close(video_fds.fd);
			if(video_fds.rtcp_fd != -1) close(video_fds.rtcp_fd);
			if(audio_fds.fd != -1) close(audio_fds.fd);
			if(audio_fds.rtcp_fd != -1) close(audio_fds.rtcp_fd);
			return -5;
		}
                res = curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &code);
                if(code != 200) {
			JANUS_LOG(LOG_ERR, "Couldn't get SETUP code: %ld\n", code);
			g_strfreev(parts);
			curl_easy_cleanup(curl);
			g_free(curldata->buffer);
			g_free(curldata);
			if(video_fds.fd != -1) close(video_fds.fd);
			if(video_fds.rtcp_fd != -1) close(video_fds.rtcp_fd);
			if(audio_fds.fd != -1) close(audio_fds.fd);
			if(audio_fds.rtcp_fd != -1) close(audio_fds.rtcp_fd);
			return -5;
		}

Similarly this would need to be added at line 7745.
I can create a PR for this if requested.

[lminiero]

Makes sense, we're already using curl_easy_getinfo to check the response to a DESCRIBE to I guess we should indeed use it for other requests too (and not just SETUP). No need for a PR as you highlighted the problem clearly, I'll prepare a commit for master shortly (even though I guess we'll have to make the same fix in 0.x as well).

[lminiero]

Should be fixed now, thanks for spotting this!