You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Testing of prerelease 3.6.0 with new adslib from upstream #505
With #504 there is a brand new adslib version from upstream. A step that has been due far to long. All tests in the CI have passed so far. However I'm a bit reluctant to release 3.6.0 to PyPi. Could some of you guys give it some testing and provide some feedback if everything works as expected ... or not ;)
The new version has been released as a prerelease on PyPi and can be installed via
I've been testing the last couple of days, and can also say I have not found any issues and haven't had to make any changes to our implementation. Thanks for this!
Sorry, I have done a bit more testing and have some regression. It seems to be specifically for notifications: The timestamp that comes back from the notification (whether you decode it through pyads or not), is invalid and nonsensical and static cycle on cycle. I've included a small reproduction below, that sets up a cyclic notification at 1Hz
Thanks for the investigation @atoff-dynisma. This was fixed upstream by Beckhoff in e0e559c (AdsNotificationHeader: add BHF_ADS_USE_TWINCAT_ORDER), and is included in master since the adslib submodule was bumped in d5a950d.
The published pyads==3.6.0.dev1 prerelease wheel was built before this fix and does not include it.
I'll cut a new prerelease built from current master for further testing.
Activity
chrisbeardy commentedon Mar 15, 2026
Great work, ill give this a go next week
BS-Highline commentedon Apr 28, 2026
I tried it out on Linux last week and so far, I haven't seen any issues/regressions.
stlehmann commentedon Apr 28, 2026
That's great news. @BS-Highline thanks for the feedback.
atoff-dynisma commentedon Jun 4, 2026
I've been testing the last couple of days, and can also say I have not found any issues and haven't had to make any changes to our implementation. Thanks for this!
stlehmann commentedon Jun 8, 2026
Thanks for the feedback everyone. Looks like all runs well. So I'm going to drop a release to this soon.
atoff-dynisma commentedon Jun 11, 2026
Sorry, I have done a bit more testing and have some regression. It seems to be specifically for notifications: The timestamp that comes back from the notification (whether you decode it through pyads or not), is invalid and nonsensical and static cycle on cycle. I've included a small reproduction below, that sets up a cyclic notification at 1Hz
stlehmann commentedon Jun 13, 2026
Thanks for the investigation @atoff-dynisma. This was fixed upstream by Beckhoff in
e0e559c(AdsNotificationHeader: add BHF_ADS_USE_TWINCAT_ORDER), and is included inmastersince the adslib submodule was bumped ind5a950d.The published
pyads==3.6.0.dev1prerelease wheel was built before this fix and does not include it.I'll cut a new prerelease built from current master for further testing.
Generated by Claude Code
stlehmann commentedon Jun 15, 2026
@atoff-dynisma please have a look at the new prerelease 3.6.0.dev3. Things should work now as expected. If so please drop a short confirm.
atoff-dynisma commentedon Jun 16, 2026
Yep looks good now 👍
stlehmann commentedon Jun 16, 2026
With #538 new version 3.6.0 has been released. So I'll close this issue.