Since the last nightly update two nights ago, the scheduler is now failing to take an image when called up within the scheduler. As a result, it tries to solve a non-existing frame, resulting in failure. It makes no difference if I am using the online or the offline solver, since there is no image to solve. Both fail.

Note, this problem only occurs when using the scheduler now. The solver works just fine when asked to solve and slew to a target that was first manually selected.

Attached a screenshot that illustrates the problem. Note the scheduler failing to solve a field, while there is no image.



Has anybody else experienced this?

Jo

I hate to bring this back to everyone's attention, but if you need to use the scheduler, this bug is an absolute deal breaker. Basically, it renders the scheduler completely unusable.

Eric and Jasem, can any of the changes that were made to fix the missing pixels problem in the ZWO cameras explain this problem? Is this problem also restricted to ZWO ASI cameras or has anyone else experienced the same?

Thanks for your comments on this.

Jo

Would the new Library

INDI Library v1.7.6 Released. (Yesterday)

have fixed that problem?

I have not yet tried that out, but it would be good to know whether changes were made or whether I would be better off reverting to the last stable version of KStars. There are just too few clear nights here at the moment to spend observation time with purging the nightly version and reinstalling the stable version in a rush.

Thanks, Jo

I use self-compiled GIT head versions of KStars and INDI and have been imaging using scheduler, including aligning, over the weekend, yesterday and today without issues so it would be good to have logs of the failing case as it doesn't seem to be deterministic.

Apologies, the link to the logs was posted in a different thread.

www.indilib.org/forum/ekos/4517-can-not-....html?start=48#35559

Reposted here now:

www.dropbox.com/sh/jsu99h7geo6sgss/AACUm...76w67XR5ETf5gUa?dl=0


Which version of KStars have you compiled and been using? The problem does not occur with the stable version, only with the nightly build from a few days ago. That version fixed the lost-pixel problem (discussed in the other thread).

I usually compile a new version daily before the session so I have probably used same-ish version as you have too. I briefly checked the logs you sent but didn't find debug output for alignment at all though the trixel log spam (seems to be something related to star catalogs) made it quite hard to find anything Do you have Scheduler and Alignment logs turned on in Ekos logging settings like the attached image shows?

Yes. All boxes are checked, except for the FITS one.

Bottomline is the alignment module fails to capture an image, but ONLY when called from the scheduler. It proceeds straight to solving without an image to solve. That then affects obviously offline solving as well as online solving. In both cases, the solver fails. However, it works just fine when I am solving a target manually. It also realigns just fine after a meridian flip is commanded by the Capture module. So its not as if the Alignment modules is systematically broken, the problem is that when called from within the scheduler it skips the part where it is supposed to capture a frame.

It always worked before, though.

That is a new bug. Perhaps it has gotten fixed in the latest nightly version. I installed that this morning, but have not yet had a chance to test it.

Judging by the near total silence this post has received, I take it nobody else is experiencing this problem?

Does that mean that it has been corrected in the latest build and this was a transient fluke?

I have not been able to test the latest build yet, since the weather here has not been cooperative, but will do so, of course, as soon as the skies clear up.

In the meantime, anyone else using the scheduler from the nightly build (not the stable one), just let me know your experience. Send me a message if you don't want to post.

Thanks, Jo

I just have updated my two mini-pcs with the latest KStars 3.1.0 releases so if this is part of the release then I will see it soon. Possibly tomorrow night but that, as usual, depends on the weather. I'll let you know as soon as I know more.

Wouter

Thanks, Wouter! Greatly appreciate your help with sorting out this problem.

Best

Jo

Last night I wasn’t able to test this despite it being a perfectly clear night grrr but tonight I was despite it being cloudy. What I did is simply connect all my gear indoors, do a fake slew to M 42 and try to solve an image. All done with the scheduler where solving the image was part of centering the object. It all worked flawlessly, apart from the fact of course that there were no stars in the image. But the solver does receive an image when called from the scheduler, which was what this was about, right?


CS, Wouter

Thanks, Wouter, and yes, that was the problem.

It is possible that that problem only affect that particular nightly build and that it got corrected since. Most likely, because I would have expected some similar reports otherwise.

I have also updated KStars in the meantime to the most recent build (yesterday). So it is good to know that it will now work again. Just too few clear nights around here at the moment to waste the scarce observing time on troubleshooting and purging installations.

Thanks so much!

Jo