Hi there again,
OK, I've tested the idea of swallowing the startup frames. Unfortunately, for each frame consumed this way, a new one is apparently put into the queue, so when I finally get the correct picture, there are more legitimate frames waiting, which are then delivered upon the next request - and are outdated again.
I've attached a small Qt based test program that exhibits the problem. Just click on the image to issue a new request. The code in capture.cpp is a stripped down version of the actual project code, with lots of debug output stuffed in while investigating, but some generalizations moved out to cater for the rather specialized requirements of the project and this demonstration. No guarantee that it doesn't crash with other cameras or setups...
Again, after a change of scene, it takes five clicks until an updated image appears. Hopefully this can be reproduced on other tester's systems.
Ulrich
OK, I've tested the idea of swallowing the startup frames. Unfortunately, for each frame consumed this way, a new one is apparently put into the queue, so when I finally get the correct picture, there are more legitimate frames waiting, which are then delivered upon the next request - and are outdated again.
I've attached a small Qt based test program that exhibits the problem. Just click on the image to issue a new request. The code in capture.cpp is a stripped down version of the actual project code, with lots of debug output stuffed in while investigating, but some generalizations moved out to cater for the rather specialized requirements of the project and this demonstration. No guarantee that it doesn't crash with other cameras or setups...
Again, after a change of scene, it takes five clicks until an updated image appears. Hopefully this can be reproduced on other tester's systems.
Ulrich
Statistics: Posted by uh_comm — Mon Feb 16, 2026 9:36 pm