Prevent 'negative' buckets from being created

[philippotto]

URL of deployed dev instance (used for testing):

• https://nonegativebrushing.webknossos.xyz

Steps to test:

• brush at negative positions (nothing should happen)
• check that no buckets are sent to the server
• note that updateSegments will still be created with negative anchor positions, but I think it's alright (the case is very unlikely anyway and shouldn't do any harm)

Issues:

• fixes Volume annotations should not save buckets with negative bucket position #7117

---

(Please delete unneeded items, merge only when none are left open)

• Updated changelog

[fm3]

Backend LGTM

[daniel-wer]

@philippotto Could you explain the fix? As far as I can see, the tracing's bounding box being negative caused "negative" buckets to be created and sent to the server. How did the tracing's bounding box become negative and shouldn't this be prevented, too?

[philippotto]

@philippotto Could you explain the fix? As far as I can see, the tracing's bounding box being negative caused "negative" buckets to be created and sent to the server. How did the tracing's bounding box become negative and shouldn't this be prevented, too?

The DataCube instantiates the BoundingBox class and stores that in this.boundingBox. That box is used to prevent buckets from being created outside of the DataCube's volume.
The BoundingBox class itself uses -Infinity and +Infinity as defaults for min and max. These defaults were partially overriden. Namely:

• when a volume task exists, min and max were overriden (--> no problems, in that case)
• otherwise, only max was overriden (with this.upperBoundary)

Thus, if no volume task existed, min was left at its default of -Infinity. To fix this, I changed the code to always intersect with min=0,0,0 and max=this.upperBoundary.

I hope, this makes sense. Otherwise, let's call about this 🤙 =)

[daniel-wer]

Thank you, that makes sense 👍 Does it make sense to use the lower boundary of the dataset instead, then? Otherwise, the behavior is inconsistent with the upper boundary being strict and the lower boundary always being at [0, 0, 0].

[philippotto]

Thank you, that makes sense +1 Does it make sense to use the lower boundary of the dataset instead, then? Otherwise, the behavior is inconsistent with the upper boundary being strict and the lower boundary always being at [0, 0, 0].

Good point! I also refactored the code a bit so that we have one bbox type less. please have another look :)

[daniel-wer]

Nice cleanup, LGTM! 👍