Hy suggested offline that I try to reproduce this with the Simulator profile and I managed to do so. Thanks for the suggestion Hy. I can reproduce it with the Simulator once I install GSC. It is easy to do:

1) start the simulators
2) do a manual auto-focus with these settings:

• In the Settings tab choose Full Field with inner 25% and outer 80%
• In the Process tab choose Detection: SEP, Algorithm: Polynomial, Tolerance: 5%

Leave all other settings at their default values

3) schedule a job in the CCD tab with these settings:

• Exposure: 30 sec
• Count: 5
• Refocus every: 1 minutes

4) add the job to the Sequence Queue and press the Start Sequence button on the lower right of the Sequence Queue

The job will take 2 images and will then focus. It is easy to see that all focus statistics are exactly the same. Same HFR, same number of stars.

This is on an x86_64 mini-pc with Ubuntu Mate 20.04 and the stable KStars-bleeding of today (April 25, 2020).


Wouter

I just tried with 15 10 sec images and rescheduling every 1 minutes with no annulus. Again, the same, last, image of the manual auto-focus sequence was used for all images. I then restarted KStars completely and tried again without a manual auto-focus first. I tried twice, once with the polynomial algorithm and once with the linear algorithm. In both cases KStars tried to take an image when the auto-focus routine was triggered and then crashed since no image was retrieved.

ElCorazon suggested offline that this may be architecture related since he doesn't have any issues on his RPi4 running 32 bit Ubuntu. I’ll try to install some new 32-bit Ubuntu version in a virtual machine and see if I can verify that it does work there.


Wouter

Hmm only KStars version 3.0.0 is available for Ubuntu 18.04 i386. Sorry, this is not a good way to try and reproduce this.


Wouter

Wouter,

Given the number of variations you have tried and the fact that others haven't seen the same issue across multiple architectures (RPi, i386 and amd64), there must be something unique to your set up or your tests.

One variable besides computing architecture between our set up is I use an arduino-based Moonlite compliant focuser.

You are right, it is related to one unique setup in my tests. I just had a Zoom session with Hy and we found the cause. I have the FITS Viewer disabled in the KStars settings. Once I enable it, all works as expected. As a matter of fact, as soon as Hy disables it on his macbook, the bug manifests itself there as well.

So, as a workaround I will enable the FITS Viewer until this has been fixed.

Thanks once again for the Zoom session Hy!!!


Clear skies, Wouter

Glad to see it's sorted now. If you were using SEP, not having the FITS viewer may have been why there were no messages about the stars SEP found. My log has many rows with this information.

Wow, thanks for the workaround Wouter! That’s awesome news. I’ve been using RPI4 with the latest SM OS 1.5.1 and I get the same EXACT results from all of your tests! I mean exact. Even the test where KStars crashes if you go straight to an automated autofocus routine. I’m so glad there’s finally a workaround. Maybe I can catch up on sleep now!!!

-WES


Sent from my iPhone using Tapatalk

Glad to see that it is working for you as well now, Wes!


Clear skies, Wouter

Great work, Hy!

FWIW, the fix for this problem is now in the latest software, and very likely in the latest nightly builds. You should not be seeing this issue anymore, even if you don't run Fitsviewer.

The bug was introduced along with the fix for the old memory leak issue on Feb 23, and it has been in there since then until the last day or two.
If you do see these symptoms on future software, please post.