fix error when loading agglomerate skeleton for single-segment agglomerate #6294
Conversation
|
I wanted to test this, but I'm not able to reproduce the error locally on master. I'm in a hybrid annotation without an editable mapping, rightclick - "Import Agglomerate Skeleton" for an agglomerate skeleton with a single node. Loads just fine 🤔 |
|
huh that is odd. And the request is going to the datastore, and a skeleton with no edges is returned? Could you give me the relevant agglomerate file? |
|
It's the test-agglomerate-file dataset with agglomerate_view_70 and segment 13 at position 61, 89, 101. |
|
you are right 🤔 That ay mean that this problem does not happen for multi-dimensional arrays like edges. I wonder what the original bug report was about, then 🧐 |
|
As indicated internally, I think the problem only happens if the edges_offsets point to the end of the edges array. Apparently, jhdf5 then tries to read 0 elements from that index which fails. I could reproduce the error on master using agglomerate 53 at position 147, 110, 124. This PR fixes the issue 🎉 |
…type * 'master' of github.com:scalableminds/webknossos: fix error when loading agglomerate skeleton for single-segment agglomerate (#6294) Editable Mappings aka Supervoxel Proofreading (#6195) Increase maximum interpolation depth to 100 (#6292) Add download modal to dataset view actions (#6283) Drop "Explorational" from info tab (#6290) Allow version history view in annotations not owned by you (#6274) Bucket loading meter (#6269) Revert "Merge "Shared Annotations" with "My annotations" (#6230)" (#6286) Merge "Shared Annotations" with "My annotations" (#6230)
In newer hdf handling code we already added zero-length checks, but this old code did not have them. So when loading single-segment agglomerate skeletons, the edges array was empty, which cannot be passed to jhdf5 as is.