Closed pixelzoom closed 3 years ago
Something is a bit weird here. This problem is no longer showing up in CT, as can be seen in the screenshot below. It also doesn't occur when I test with the multi-touch query parameters on my local machine. However, after some investigation, I'm able to duplicate the problem manually on the master version. Here is the sequence:
I'll fix this, but this situation makes me suspect that the multi-touch testing isn't working correctly, and I'll need to follow up on that.
Here's a screenshot to show that the sim with the errant code still in place is passing the multi-touch automated testing:
I updated to the most recent drag handling and forwarding approaches, and can no longer reproduce the problem manually. Also, I didn't see any issues after two minutes each of fuzz and multi-touch-fuzz testing. I'll leave this open to verify that CT is clear, and I'll also mark for dev meeting to discuss whether something is wrong with multi-touch fuzz testing, see above.
CT was clear over the weekend.
I think you still have a potential problem here. You’re calling userControlledProperty.unlink
in 2 places in BiomoleculeCreatorNode. So it’s possible that userControlledPropertyObserver
may not be a listener when reset
is called -- or at best, it's unclear if userControlledPropertyObserver
is guaranteed to be an observer when reset
is called. To be safe, in reset
I would check userControlledProperty.hasListener
before doing the unlink
.
I think the reason CT looks green is just because it usually isnt running much fuzzing at the moment, can see here that multitouch fuzzing especially is not running:
Maybe the priorities of certain tests can be adjusted to make sure that important fuzzing is always run before moving on to next set of SHAs.
Above, @pixelzoom said:
I think you still have a potential problem here. You’re calling userControlledProperty.unlink in 2 places in BiomoleculeCreatorNode...
I didn't want to add the suggested hasListener
test since the two code paths that do the unlink
are intended to be mutually exclusive. I stated this in a Slack discussion, and @pixelzoom recommended adding an assert to make sure that this assumption is always true, so that's what I've done.
Dev Meeting 3/13: @jonathanolson is looking into the issues with multitouch fuzz testing on CT.
CT has continued to be clear during this time, and we've discussed the issue with CT in dev meeting, so I'm going to close.
In 2/4/21 dev meeting, we established that this may be due to CT multi-touch fuzz testing (
?fuzzPointers=2
) being enabled for phetsims/aqua#106.