Hi Jim,

Yes, I know copr.fedorainfracloud.org/coprs/xsnrg/, of course

All my Raspberry PIs run under Fedora.
This distro works really fine on PIs, i've been using it for years, and I love setting up my network both the old way AND with nmcli) ...

That would be nice to have those 'bleedings' (with the appropriate fedora-repoxx) under Raspberry Pi..

Last time I checked, I dindn't find them.

- Marc

Addition:  If I downgrade only libasi, in my case to 1.9.0-21_gd9a5e309 from March 11, everything runs fine.  That is the old SDK, 1.16.3.0.  So it obviously is an issue of the ZWO SDK?
 

Libasi package also contains the udev rule file, which has that usbfs 256MB memory line in the old version, but not in new one.

This is true, but it still exists. It was just moved to INDIlib, and should have the same effect. This change has been working for me now for  some time. Jasem moved it there because many devices require it, so indilib is a good central place to manage it.
I still suspect changes in the SDK/driver from ZWO as having some differences to figure out.

Jim

 Ah, it's in 99-indi_auxiliary.rules now.  Was wondering, too.  But I had checked that the setting was at 256, so that was/is not the issue.
But if things work with the previous SDK and not with the latest, it's either a bug in the SDK, or some change in the API (or responses) the is not reflected by the INDI code.  The SDK changelog however is not very helpful, just referring to 'bug fixes'
 

Addendum: I had not checked the syslog so far. It actually says
(I also noticed that the new udev rules no longer contain the chmod to 0666 for ASI devives; but it also crashes when run as root....)

Running ASI1600mm, ASI120mm mini, EFW, EAF, building from git on RPI4. No issues (anymore).

I did have an issue with the ASI1600mm where it would fail to capture, and the image from the ASI120mm would occationally pop up in the preview instead.
Deleted xml-files from .indi, and set everything up from scratch again. After deleting the xml-files, I saw that the ekos driver setting for both ASI120mm and ASI1600mm were both set to main scope. I set ASI120mm to guide scope, and everything were working again. Not in front of kstars at the moment, so tick boxes could have different names.

Yup, SDK regression isn't out of the question, though it seems to be either camera or architecture specific as my ASI120MM-S with ARMv7 and 178MC with x64 I test with both work fine. API hasn't changed, at least the header files were identical when I updated the SDK and they usually haven't changed existing things, just added new ones.

In the latest SDK, the problem with damaged frames during long exposure is fixed.
Probably it required more library modification and maybe something got corrupted.
I still offer cooperation for ZWO, unfortunately without success.

Just for confirmation:
With latest, you refer to 1.17, yes? So that one fixes the original issue in this thread, but causes crashes for some?

Yes, I mean 1.17.
I am not getting detailed information from ZWO, I am trying to establish cooperation with them. I am working on the libASICamera2Boost library. Unfortunately, each camera has a different implementation and it's hard for me to keep developing without the original library source.

In original library (1.17), there are probably larger double buffers used for communication with the camera.
The usbfs_memory_mb value should be at least '6 x Resolution'. For now, these are guesses, I need to analyze the library more.
Anyway, to rule out a memory problem, I suggest temporarily increasing the usbfs_memory_mb value and testing indi with different ZWO libraries.
For example:

The image buffer is not the only thing that is broken.

Yesterday I used an installation from March 17 that had worked perfectly last month after I manually increased the usbfs memory in the asi.rules file. Unfortunately, I had not turned off unattended-upgrades for the Pi (running Ubuntu 20.04) and after that had completed, the indi server crashed every time I called it up. It would crash even when just starting the Moonlite driver for my focuser, so unrelated to the ZWO issue.

With the system already broken, I figured upgrading everything from the nightly repo might help. It did, indi server no longer crashed after that, but now my ZWO-EFW was missing again! Moving my old asi.rules and efw.rules files back into the /etc/udev/rules.d folder did not fix it this time.

Is there a solution on the horizon for this, or is the only practical solution right now to go back to the stable version from March 1?

Jo