From: Derek Lesho Subject: Re: [PATCH 3/6] mf/session: Fix backwards handling of MF_TOPONODE_NOSHUTDOWN_ON_REMOVE. Message-Id: Date: Thu, 13 Jan 2022 11:27:56 -0500 In-Reply-To: <07b70e6e-178e-98db-86d3-294dde431a22@codeweavers.com> References: <20220104173527.77306-1-dlesho@codeweavers.com> <20220104173527.77306-3-dlesho@codeweavers.com> <07b70e6e-178e-98db-86d3-294dde431a22@codeweavers.com> On 1/13/22 10:51, Nikolay Sivov wrote: > I unsuccessfully tried to test this with simple topology of real > source -> sample grabber using source type (no conversion), and > Close(), or clearing topology does not trigger shutdown, ever. > session->Shutdown() always shuts down actual sink when it's set as > activation object, and again, never shuts down if it was set directly. > > This is pretty weird, I'll try some more tests. Maybe just > SetTopology() does not transition session state enough, and it has to > start playback even if just for a little while. It seems wrong if this > automatic shutdown only happened when playback completes by reaching > its end time. > > Derek, have you tried to test this in some way? I've not yet tested this, but that is quite strange behavior you've described.  I'll try using mftrace with a media engine that runs multiple topologies to try and get some insight here later today.