Thanks for checking it out. I might be doing something odd or did something odd at some poing. I am new to both Onstep and EKOS.

I'll reset the Onstep firmware and see if it resolves the issue.

Hello all ...

This bug may be related to an issue that is inside Ekos (not INDI), that causes status changes to be immediate, and not reflect what is actually going on.

The above log seems to confirm that: it changed from slewing to tracking in merely one second.

This issue was discussed in this thread . I do not recall a definitive resolution yet.

One poster on the OnStep group identified where the issue is, but does not have a solution yet.
 

Could someone please check the status of the INDI property EQUATORIAL_EOD_COORD is always IPS_BUSY when the mount slews?

Here the explanation how Ekos recognizes the mount's tracking and slewing states:

There is one thread running that regularly reads the data of the INDI driver showing the current RA/DEC coordinates and its status. Simplified, IPS_IDLE --> mount is idle, IPS_OK --> tracking, IPS_BUSY --> slewing.

Now the question is how should we detect from EKOS that a slew is finished. Currently, we detect this that after a slew command, the mount shows the tracking state - and this is recognized by the thread above. When a slew is issued, Ekos sets the mount state to "slewing".

The problem occurs when the above mentioned thread reads the mount in parallel and obtains the tracking state BEFORE the INDI driver starts slewing. As a result, Ekos assumes that the slew has already finished while it hasn't even started.

But in this case, the time difference in the log is more than a second - that's why I think we do not have this problem here.

That's why I would like to check whether the OnStep INDI driver properly reports it's state

I have another message tread going that details exactly this - I have been hunting in code for who is setting that state and where and it eventually comes down to one block of code in mount.cpp and the check of m_status which is in a wrong state for sure. - I still cannot find it - have located the state changes to slewing but once that is set the other thread somewhere flips it back - I am tempted to throw a wait() into the block of code that sets the slewing state (it is right after the actual slew command) to give a second or 2 before setting slewing state in hopes it sticks - it is obvious that the driver is returning a steady stream of RA/DEC changes, so whatever is looking at that output should know but in the mount/driver there is a very small lag as the ramp up inthe physical mount starts slowly - I suspect THAT is what is giving thread #1 the chance to think the slew is done as the RA/DEC has not yet changed.....thinking of this perhaps the polling frequency could dropped via indi configuration forcing a slow down of the data stream - if BOTH threads have do wait perhaps the two will sync vesus the real state of RA/DEC and thus know a change has happed and set slewing!....hmmmmm...

Its tricky There is a thread running in Mount that calls updateTelescopeCoords() once a second. The place where Mount detects a mount status change slewing --> tracking (and subsequently sets the current position) is at mount.cpp::658 ff:A few lines above at line 652, the scope status is read from INDI scope client:
The INDI scope client (inditelescope.cpp) determines the scope status by the status of the INDI property EQUATORIAL_EOD_COORD (see inditelescope.cpp::1286 ff).

This client listens to the connection to the INDI service for the scope. The INDI service informs its clients by pushing updates for INDI properties. In the case of a OnStep scope, this is lx200_OnStep.cpp from the libindi project. This happens through LX200_OnStep::ReadScopeStatus() (lx200_OnStep.cpp::1804 ff), which is regularly called through Telescope::TimerHit() (see indibase/inditelescope.cpp::1592 in the libindi project).

The frequency of TimerHit() is set through the polling period set for the INDI service.

HTH Wolfgang

It has to be a change in the past ~ several months, since it was working fine before.

For me, I use Jasem's PPAs for Ubuntu, and I noticed the 'immediate state transitions' issue only recently (past couple of months). Not sure how soon or how often changes in git make it to the PPAs though.

I tried to check the git log for kstars/ekos/mount, and I see two changes for meridian flips as follows. But checking the commits there is no smoking gun there.

commit e32f6399c20266c50e73280ddce1616428487fa2
Author: Wolfgang Reissenberger <This email address is being protected from spambots. You need JavaScript enabled to view it.>
Date: Tue May 4 13:26:15 2021 +0000

Bugfix meridian flip for lower culmination


commit 581c7ab7729be674eff87772e3c52f85f0317eea
Author: Wolfgang Reissenberger <This email address is being protected from spambots. You need JavaScript enabled to view it.>
Date: Wed Mar 10 15:38:41 2021 +0000

Meridian flip test case simulating current behavior

Currently, a guiding deviation while the mount is waiting for capture completed
leads to the result that after the meridian flip is completed, neither guiding
nor capturing is restarted.

This is not appropriate, since guiding deviations lead to restarting captures
and not to aborts. This should remain true if a meridian flip happens inbetween.

I'm quite sure that the problem lies somewhere in the state update from the INDI service, that's why I asked for the checks about the status of EQUATORIAL_EOD_COORD. I could assist in an online session, but I need somebody with a OnStep mount and a debugging environment.

What kind of debugging environment?  

Linux + QT Creator

I have now been testing changes in polling time with no effect so far but I can reproduce this consistently - I am on my 3rd attempt now and will post logs when it is done - from what is can see, the polling changes simple move along the time between MF slewing started and flip completed - this behavior does actually stop the full slew - the mount stops and resumes tracking at some half baked position after about 2 seconds (with the added wind-down of the slew to tracking - so perhaps 6 seconds of motion)

Jamie,

An simpler and easier test scenario may be enabling the notification for end of slew (a popup will show). Then when you start a slew, the popup will show instantly saying the the slew is done, whereas in reality, the slew has just begun and has not reached the destination coordinates yet.

FYI latest test - increased the polling to 2000 ms and I get exactly the same MF_complete 1 second after slewing started BUT the mount was able to physically do the full flip whihc was NOT the case with 500, 1000, 1500. So regardless of polling the actor that is resetting m_state to tracking is not being impacted by polling. I need to find out where that tracking is being pulled from - will be checking the new logs - I am suspecting teh state is coming form a default or local variable that is just sitting at tracking - there is a LOT of cross-variable checking going on here