Acquisition thread gets sporadically blocked after multiple acquisitions

[skalarproduktraum]

It seems that the acquisition thread gets blocked after several acquisitions, which looks like a deadlock issue, see attached log.

hang.log

[jondaniels]

Maybe resolved? Haven't seen recently. Let's all keep an eye out for.

[bls337]

The next version of the plugin has better lifecycle management, I'm not sure if it will fix this issue, but when you close the plugin it makes sure to set the frame/model to null.

Previously the objects would still be allocated in memory even when you closed the plugin, because the SciJava MenuPlugin needs to hold onto a reference. Hopefully this will help with any "ghost" activity and prevent previous sessions from causing problems.

[oleksiievetsno]

I have checked and this issue is still valid
I get any error after a couple of acquisitions, I need to restart the plugin.

[bls337]

I tried to replicate this problem here, but I haven't been able to. I have 2 Hamamatsu Fusion cameras so that may be the difference. I'm curious if you see this issue when you use only one camera, or does it only happen when both cameras are active?

How many acquisitions does it take before you get a hang? Are you getting any errors?

[bls337]

I would be curious to see if there is a difference in the acquisition timing settings files that each plugin outputs.

[bls337]

Pull Request #427 should help with this issue. The finish() method could throw and exception which would prevent currentAcquisition_ from being set to null. Now there is an inner finally block that always runs.

[bls337]

There has been a lot of work on the acquisition engine lifecycle that should help.