On Fri Sep 8 11:51:16 2023 +0000, Rémi Bernon wrote:
I see, this is very unfortunate IMO... Looks like it's coming from the GstAggregator class. I still think we should avoid the condition variable, but I don't really have any good suggestion to workaround that. A busy loop might be better, assuming it will quickly EOS, but it's a bit ugly. Alternatively I had thought about implementing a GTask pool with Win32 threads, which would also solve some problems we have with logging, but I don't know if it's a good idea. It would let us use Win32 sync primitives for notifications instead though, so less ugly but a lot more code so I'm not sure it's worth it. Then, as we already have condition variables elsewhere, I can probably live with it in this case, if you think it's better.
Hi Rémi, I've asked you this before in 3606, but in what way are condition variables broken? Can you please explain?