From: David Woodhouse <dwmw@amazon.co.uk>
Date: Sun, 5 Apr 2020 23:04:14 +0000 (+0100)
Subject: Cope with screen==NULL in screen_appsink_new_sample()
X-Git-Tag: v1.4~10
X-Git-Url: https://www.infradead.org/git/?a=commitdiff_plain;h=ceee47d7b186b250fc6fc6625ae4adb1305773b0;p=pidgin-chime.git

Cope with screen==NULL in screen_appsink_new_sample()

I saw a crash with this once, in a long sequence of sharing from A to B
and then B to A testing screen sharing startup/teardown.

It looksl like a race condition when we change to the no_appsink_callbacks
but the callback is already in progress, and the cbdata gets set to NULL.
---

diff --git a/chime/chime-call-screen.c b/chime/chime-call-screen.c
index b038a43..939fc24 100644
--- a/chime/chime-call-screen.c
+++ b/chime/chime-call-screen.c
@@ -379,6 +379,9 @@ static GstFlowReturn screen_appsink_new_sample(GstAppSink* self, gpointer data)
 	if (!sample)
 		return GST_FLOW_OK;
 
+	if (!screen)
+		return GST_FLOW_ERROR;
+
 	if (screen->state == CHIME_SCREEN_STATE_SENDING && screen->viewer_present) {
 		GstBuffer *buffer = gst_sample_get_buffer(sample);
 		gsize len = gst_buffer_get_size(buffer);