Is there a known issue with flat calibration and setting the time to infinity?
It is not clear why it happens.
Steps to recreate:
1. Set sub to Flat
2. Set ADU to 12000
3. Set exposure to 1 sec
3. Start capture
Some times the calibration works, sometimes now. Editing the exposure to reset the value and redoing until a value allows it to work, was the work around.
Once the sequence was created and saved, rerunning the sequence still resulted in 2 of the seven filters needing to have exposures played with.
I do not know what camera you use, but with my ASI1600 I had a similar issue where it couldn't figure out the right exposure time for the ADU I wanted. I set the exposure to start at 0.5 seconds and the problem never resurfaced.
I added more debug logging for this feature in the next KStars release so hopeully we'll know what's going on. Basically, after the 2nd capture, it tries to do llsq/polynomial fitting for the data to get the optimal exposure, so something wrong went there. When we have the logs in the future we might know why and fix it.
I had those problems too. Yesterday I took flats set the count to 22000 and 0.5s and ekos failed to find the best exposure after some iterations.
I have set it to the lowest possible value (0.001s) and on the second try it worked. From what I have seen in my system, when the defined exposure (0.5s) is higher than the predicted exposure (in my case 0.027s) ekos has some kind of difficulty to find it. When it's lower than expected it works fine, it goes up until it finds the correct one.
In my case I was using an Atik 420M.
Thanks for the log. This is why I stress on logs because now I identified the issue. General descriptions of the problem can sometimes be next to useless if not accompanied by log. Description + Log = FTW!
So the issue is that at 1 second, the flat image was already saturated and the code didn't deal with saturated case so it was including them in the calculations when it should not. I pushed the fix to Git and might make a release tomorrow in the PPA.